WO2013018160A1

WO2013018160A1 - Valuation assistance system and valuation assistance program

Info

Publication number: WO2013018160A1
Application number: PCT/JP2011/067468
Authority: WO
Inventors: 光一矢野
Original assignee: アクアエンタープライズ株式会社
Priority date: 2011-07-29
Filing date: 2011-07-29
Publication date: 2013-02-07

Abstract

The invention is a valuation assistance system that enables a user to more precisely, accurately and simply assess a value disparity between products or services before purchase. The invention is provided with: a valuation server that determines the statistically significant difference for a value disparity between items to investigate, and outputs a result; and a user terminal. The valuation server comprises: a valuation information input unit that receives input of valuation data related to the items to investigate, and stores said data in a valuation history database; a valuation history information acquisition unit that extracts valuation data from the valuation history database, for each valuation indicator, for each of multiple user-selected items to investigate; a statistical calculator that determines the statistically significant difference between the items to investigate, using a statistical calculation based on ranking data computed by a predetermined procedure on the basis of extracted valuation data; and a valuation result output unit that outputs statistical calculation result information to the user terminal, along with recommendation information related to a value disparity assessment standard selected on the basis of information indicating the statistically significant difference between the items to investigate.

Description

Value evaluation support system and value evaluation support program

The present invention relates to a technology for trading systems for products and services using a computer, and in particular, when there are a plurality of compatible products and services, provides information related to the difference in values as judgment materials to a user. The present invention relates to an effective technology applied to a value evaluation support system and a value evaluation support program.

There are many trading systems for products and services using computers (hereinafter sometimes referred to as “products, etc.”), but there are no individual differences in the value of products handled by such systems, or In general, a user can determine a product to be purchased by estimating a difference in value from detailed information presented attached to the product.

For example, in the case of purchasing a certain kind of notebook computer through Internet sales, it is assumed that the notebook PC with any serial number has no individual difference and has the same value. Depending on the model, there may be cases where the battery has been in stock for a long time and the battery life of the attached battery is short, but it is assumed that such individual differences are not common.

Therefore, when there are multiple types of notebook computers that meet the user's requirements, a method for making the user understand the difference in value that exists between models, etc. by presenting their detailed information and price. Has been taken. For example, by comparing and displaying manufacturer names, you can understand the difference in value such as quality, popularity, brand power, etc. among manufacturers, by displaying CPU performance, HDD capacity, display resolution, etc. The method is to make the user guess the value difference in performance and quality.

On the other hand, in some fields, the value of a specific evaluation target is calculated and evaluated from parametric data (data assuming a specific distribution such as a normal distribution with respect to the distribution of the population) using a dedicated system or program. That is also done.

For example, Japanese Patent Laying-Open No. 2005-322089 (Patent Document 1) describes a system for quantitatively evaluating a corporate brand value by statistical calculation (average, variance, trend, etc.) using parametric data such as sales. Yes. Japanese Patent Laying-Open No. 2010-286997 (Patent Document 2) describes a system that calculates an evaluation value indicating the value of an evaluation target article based on a deviation in the number of appearances of words included in the evaluation target article. ing. Japanese Patent Laid-Open No. 2002-299668 (Patent Document 3) describes a system that quantitatively evaluates the value of a photovoltaic power generation facility based on performance values such as the amount of solar radiation and power generation cost.

On the other hand, a method for calculating and evaluating a difference in value to be evaluated by statistical calculation using non-parametric data (data that does not assume that a population distribution follows a specific distribution) has been proposed. For example, Japanese Patent Laid-Open No. 10-260995 (Patent Document 4) inputs rank data provided by a plurality of testers to a plurality of samples, and a Friedman test for determining the consistency of the input rank data. If it is determined that there is a significant difference in the consistency of rank data, Wilcoxon's signed rank sum test is performed to determine the rank relationship of each sample. A rank evaluation system is described that quantifies and clarifies significant differences in relationships.

Japanese Patent Laid-Open No. 2005-322089 JP 2010-286997 A JP 2002-299668 A Japanese Patent Laid-Open No. 10-260995

When a user tries to purchase a product etc., the difference in value among the individual products etc. exists in various viewpoints, and the value evaluation standard varies depending on the user to be evaluated. Accordingly, for example, a method in which the user estimates the value of a product or the like extracted by a search or the like by a product trading system cannot be judged with high accuracy. Therefore, in reality, the user may understand the original value of the purchased product at the time of use rather than at the time of purchase. For example, when using it, the user notices that the product was worse than expected. There is also a case.

For example, in the case of the above-described notebook computer, there is a case where the battery life reaches 3 years after purchase, and when it is asked about the battery life at the nearest computer shop, it is usually informed that it is over 5 years. It is desirable that the original value of a product etc. be understood more accurately at the time of purchase. As a similar case, for example, you bought clothes through internet sales, but the texture and color of the fabric is different from what you imagined from the description or photo, or you searched for a dental clinic or lawyer office on the internet. There were a lot of hits, but there are cases where the photos and text information (including user reviews) are not good or bad. Introducing and selling products using a computer often causes such inconvenience because the user cannot confirm the actual product on site before purchase.

On the other hand, it is possible to use the computer system to calculate, evaluate and present the value difference of the product, etc., so that the user can use it as a judgment material for selection when purchasing the product. Here, the evaluation criteria (indicators) for the value of products and the like vary depending on the user, and there are some that are greatly influenced by human senses such as appearance, satisfaction, and likability. Some of these indicators include, for example, metric values according to normal distribution, count values according to binomial distribution, defect number data according to Poisson distribution, and classification data and rank data with unknown distribution. To do. In addition, there is variation in the index data individually, and some of the numerical values themselves have low reliability due to the low accuracy of the data measurement method.

It is necessary to statistically process such a wide variety of data according to certain uniform standards. However, in the conventional techniques for calculating and evaluating the value etc. using the parametric data described in Patent Documents 1 to 3, for example, the parametric index value is assumed under certain assumptions for the nonparametric information. The method of statistical calculation is applied to the parametric data after being replaced with or converted into.

Here, it is mathematically impossible to correctly convert non-parametric data with a small amount of information into parametric data with a large amount of information, and inappropriate information will be added. For this reason, even if accurate information processing is performed by a computer, there is a problem that accuracy and reliability of a result obtained are insufficient. In addition, since each system performs replacement / conversion to parametric data under special assumptions, it can be applied only to specific evaluation objects corresponding to the respective assumptions.

On the other hand, as in the technique described in Patent Document 4, for example, it is conceivable to calculate / evaluate the difference in the value of goods etc. by statistical calculation using nonparametric data. Here, with the technique described in Patent Document 4 and the like, it is possible to surely perform statistical calculation, but the disclosure is only up to the point of obtaining statistical calculation result information, and how to obtain the obtained information. It is not disclosed up to the point of use / utilization. That is, the prior art is merely a statistical calculation system. For example, based on information on the difference in value of products obtained as a result of statistical calculation, what kind of recommendation is recommended when a user selects a product, etc. It does not disclose specific contents of subsequent useful information processing such as whether to go and support.

In view of the above problems, an object of the present invention is to enable a user to more accurately, accurately, and easily determine a difference in value of goods and services that are widely sold and sold at the time of purchase. It is to provide a value evaluation support system and a value evaluation support program that support the above.

The above and other objects and novel features of the present invention will be apparent from the description of this specification and the accompanying drawings.

Of the inventions disclosed in this application, the outline of typical ones will be briefly described as follows.

A value evaluation support system according to a representative embodiment of the present invention provides statistics on a difference in value between a plurality of survey objects whose values are represented by one or more evaluation indexes including ambiguous data. A value evaluation support system having a value evaluation server that determines a significant difference and outputs a determination result, and a user terminal connected to the value evaluation server via a network, having the following features It is.

That is, the value evaluation server receives an input of evaluation data that is an evaluation result of the value for each evaluation index related to each survey target, and stores the evaluation data in an evaluation history database, and the use An evaluation history information acquisition unit that extracts the evaluation data from the evaluation history database for each of the plurality of investigation targets selected based on conditions specified by the user via a user terminal Have.

Further, based on the evaluation data extracted by the evaluation history information acquisition unit, rank data is calculated according to a predetermined procedure, and statistical significance between the respective survey targets is calculated by statistical calculation based on the rank data. The statistical calculation unit to be determined and the statistical calculation result by the statistical calculation unit are combined with the information of the statistical calculation result, and are defined in advance based on the statistically significant difference information determined by the statistical calculation unit. An evaluation result output unit that outputs the recommended information related to the determination criteria for the difference in value between the survey targets to the user terminal. is there.

The present invention can also be applied to a program that causes a computer to function as the above-described value evaluation support system.

Among the inventions disclosed in the present application, effects obtained by typical ones will be briefly described as follows.

According to a typical embodiment of the present invention, it becomes possible for a user to more accurately, accurately, and easily determine a difference in value of goods and services that are widely sold and sold at the time of purchase. .

It is the figure which showed the outline | summary about the structural example of the value evaluation assistance system which is Embodiment 1 of this invention. It is the flowchart which showed the outline | summary about the example of the flow of the main process in Embodiment 1 of this invention. It is the flowchart which showed the outline | summary about the example of the flow of the price investigation process in Embodiment 1 of this invention. It is the figure which showed the example of the menu screen for the user in Embodiment 1 of this invention to select or input goods etc. It is the flowchart which showed the outline | summary about the example of the flow of the value evaluation process in Embodiment 1 of this invention. It is the flowchart which showed the outline | summary about the example of the flow of the statistical process in Embodiment 1 of this invention. It is the flowchart which showed the outline | summary about the example of the flow of the statistical process in Embodiment 1 of this invention. It is the figure shown about the example which determines a significant difference by statistical calculation from the value of the evaluation parameter | index for every two investigation object in Embodiment 1 of this invention. It is the figure shown about another example which determines a significant difference by statistical calculation from the value of the evaluation parameter | index for every two investigation object in Embodiment 1 of this invention. It is the figure shown about another example which determines a significant difference by statistical calculation from the value of the evaluation parameter | index for every two investigation object in Embodiment 1 of this invention. It is the figure shown about the example which determines a significant difference by statistical calculation from the value of the evaluation parameter | index for every three investigation object in Embodiment 1 of this invention. It is the figure shown about another example which determines a significant difference by statistical calculation from the value of the evaluation parameter | index for every three investigation object in Embodiment 1 of this invention. It is the figure shown about another example which determines a significant difference by statistical calculation from the value of the evaluation parameter | index for every three investigation object in Embodiment 1 of this invention. It is the figure which showed the example which represented the value of the goods etc. and the value of the significant evaluation index | index in tabular form for every investigation object in Embodiment 1 of this invention. (A), (b) is the figure which showed the example which represented the relationship between the average value and price of a significant evaluation index with respect to each investigation object in Embodiment 1 of this invention with the graph. It is the flowchart which showed the outline | summary about the example of the flow of the significant difference determination process between investigation objects in Embodiment 1 of this invention. It is the flowchart which showed the outline | summary about the example of the flow of the significant difference determination process between investigation objects in Embodiment 1 of this invention. It is the figure shown about the example which determines the significant difference between the 3 investigation object in Embodiment 1 of this invention by statistical calculation from the value of a significant evaluation index. It is the figure shown about the example which determines the significant difference between the 5 investigation object in Embodiment 1 of this invention by statistical calculation from the value of a significant evaluation index. It is the figure shown about another example which determines the significant difference between the 3 investigation object in Embodiment 1 of this invention by the statistical calculation from the value of a significant evaluation index. It is the figure shown about the example which determines the significant difference between the 2 investigation object in Embodiment 1 of this invention by statistical calculation from the value of a significant evaluation index. It is the figure shown about another example which determines the significant difference between the 2 investigation object in Embodiment 1 of this invention by the statistical calculation from the value of a significant evaluation index. It is the figure shown about another example which determines the significant difference between the 2 investigation object in Embodiment 1 of this invention by the statistical calculation from the value of a significant evaluation index. It is the flowchart which showed the outline | summary about the example of the flow of the result output process in Embodiment 1 of this invention. It is the figure which showed the example of the table | surface expressing the determination content of the result output process in Embodiment 1 of this invention. It is the flowchart which showed the outline | summary about the example of the flow of the evaluation log | history information recording process in Embodiment 1 of this invention. It is the figure which showed the example about the content of the questionnaire shown to the user in Embodiment 1 of this invention. It is the figure shown about the example of the table which defined the evaluation data conversion rule and the rule of ranking for every kind of numerical value of the evaluation index in Embodiment 1 of this invention. It is the figure which showed the outline | summary about the structural example of the value evaluation assistance system which is Embodiment 2 of this invention. It is the flowchart which showed the outline | summary about the example of the flow of the main processing in Embodiment 2 of this invention. It is the flowchart which showed the outline | summary about the example of the flow of the price and value investigation process in Embodiment 2 of this invention. It is the flowchart which showed the outline | summary about the example of the flow of the matrix creation process in Embodiment 2 of this invention. It is the figure which showed the example in the case of deleting the column or row | line with missing data from the matrix table of the combination of a seller and a sales agent in Embodiment 2 of this invention. It is the figure which showed the example in the case of deleting the column or row | line with missing data from the matrix table of the combination of a seller and a sales agent in Embodiment 2 of this invention. It is the figure shown about the example of the estimate request | requirement transmitted to the seller's goods provision system in Embodiment 2 of this invention. It is the figure shown about the example of the request for quotation transmitted to the sales agent system of the sales agent in Embodiment 2 of this invention. It is the figure shown about the example of the estimate which the seller's goods provision system in Embodiment 2 of this invention produces. It is the figure shown about the example which determines the significant difference between two types of investigation object groups each having five investigation object in Embodiment 2 of this invention from the value of an evaluation parameter | index by statistical calculation. The figure which showed the example of the combination of the test | inspection means in case the deletion rate in Embodiment 2 of this invention exceeds a threshold value, and the number S of investigation objects is larger than the number of investigation objects of a matrix table after deleting all missing values. It is. (A) And (b) is the figure which showed the example which represented the relationship between the price and value of goods etc. for every investigation object in Embodiment 2 of this invention with a tabular form and a graph. It is the flowchart which showed the outline | summary about the example of the flow of the result output process in Embodiment 2 of this invention. It is the figure which showed the example of the table | surface expressing the determination content of the result output process in Embodiment 2 of this invention. It is the figure which showed the example of the part which defined the property of the evaluation data for every some investigation object, and the combination of statistical calculation information among the supplementary explanation setting tables in Embodiment 3 of this invention. It is the figure which showed the example of the part which defined the content of the supplementary explanation corresponding to the property of the evaluation data for every investigation object, and the combination of statistical calculation information among the supplementary explanation setting tables in Embodiment 3 of this invention. It is the figure which showed the example of the table | surface which defined the pattern of the explanatory note of the supplementary description in Embodiment 3 of this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Note that components having the same function are denoted by the same reference symbols throughout the drawings for describing the embodiment, and the repetitive description thereof will be omitted.

<Overview>
The evaluation indicators that indicate the value of products that are widely sold and sold generally include, for example, appearance quality, initial quality, freshness, condition, performance, durability, reliability, rarity, satisfaction, popularity, and favorable sensitivity. There are various things such as comfort, name recognition, brand power, safety, and compliance rate.

In general, new industrial products, etc., are sold with products with small variations between individuals that have passed a predetermined high quality standard, but some of them are out of quality standards, or within the standard but with good quality. There are no products.

On the other hand, for example, primary industrial products such as agricultural products and services related to people and weather, etc., have a large variation in quality unlike industrial products. Generally, conditions such as price are determined by using. For example, the price of cabbage is determined by using the weight (size), product type, and the like as indicators. However, since freshness, scratches, color gloss, and the like, which are important value indicators in agricultural products, differ from individual to individual, users generally check and purchase those indicators at a store. Therefore, it can be said that these products are not suitable for buying and selling using a computer system.

Also, some indicators that indicate the value of products, etc., are indicators that are greatly influenced by human senses, such as appearance quality, satisfaction, and favorable sensitivity. For example, indicators such as the quality of service provided by flight attendants and the deliciousness of in-flight meals that may be taken into consideration when purchasing an air ticket have different evaluation criteria. For this reason, such a value index is not presented to the user in the air ticket trading system using a computer. Thus, for example, it is possible to make a reservation for the same seat on multiple competing routes, and to determine which ticket is more valuable if the prices are the same. The current situation is that it is not possible to provide appropriate information.

In this way, the value of goods, etc. is expressed by various types of indicators. For example, these indicators include metric values such as length according to normal distribution, count values such as defect rate according to binomial distribution, Poisson, etc. There are parametric data such as count values such as the number of defects according to the distribution, and non-parametric data such as classification data and rank data whose distribution is unknown. Further, as described above, there are variations in the index data, as well as the reliability of the numerical value itself because the accuracy of the data measurement method is low.

In order to statistically process such a wide variety of evaluation index data according to a certain standard, the value evaluation support system according to an embodiment of the present invention is a non-standard method for handling rank data as a statistical processing technique. Use parametric method. Among many evaluation indexes, an evaluation index determined to have a significant difference by a test using a non-parametric method can be determined as a factor indicating a difference in value of a product or the like. Therefore, the total value of these factors is considered to indicate the difference in value of the product etc., and the user selects the product etc. more accurately from the relationship between the value difference and the price. It becomes possible.

Furthermore, since each user has various value standards, in the value evaluation support system according to an embodiment of the present invention, a significant evaluation index, price, It is recommended to select the most suitable product from the relationship. Also, users who place importance on insignificant evaluation indices are encouraged to select products etc. from price alone. In this way, the user evaluates the difference in the value of the product etc. based on the dialogue between the user and the computer, such as the presentation of the evaluation index that the user places importance on and the recommendation of the selection criteria for the optimum product etc. This provides a method for selecting optimal products and the like.

In the conventional technology, there are a large number of evaluation indexes representing value, and in order to understand the value of products, etc., it was necessary to make a judgment after taking into account all evaluation indexes. However, in the value evaluation support system which is one embodiment of the present invention, an evaluation index having a significant difference can be extracted from a large number of evaluation indices using a statistical test method. Therefore, the user can more accurately understand the difference in the value of the product etc. based on the small number of evaluation indexes that have been narrowed down. In addition, the user can interact with the computer on how to select a product etc. from the evaluation index and price narrowed down by the statistical calculation, and can specify from among the options, for example. Never get lost.

* In principle, evaluation index data used for statistical calculation is obtained by conducting a questionnaire survey to users who use computer system trading after using, using or consuming goods or services. It can also be obtained by a questionnaire survey for a third party who does not use the product or the like or a seller who sells the product or the like. In addition, a method other than a questionnaire may be used, and for example, it can be obtained by a questionnaire or an experience report. By means of a questionnaire survey or the like, the user inputs, for example, evaluation points (0 to 100 points) and the like for a plurality of evaluation indexes representing the value of the product or the like. Here, parametric data such as evaluation points are converted into rank data that is non-parametric data during statistical processing, and then a significant difference is determined.

<Embodiment 1>
The value evaluation support system according to the first embodiment of the present invention uses a computer system as an example for buying and selling airline tickets, and when the airline tickets are bought and sold between a seller and a user, The difference in the value of each air ticket (airline company) that meets the conditions is evaluated, and based on this, a standard for selecting an airline company that purchases the airline ticket is recommended to the user.

In this embodiment, the buying and selling of an air ticket is taken as an example, but the product to be sold is not limited to an air ticket, and the seller is naturally not limited to an airline. For example, an accommodation facility management company such as a hotel may sell accommodation services. In addition, it may be a case in which various products such as electrical appliances, automobiles, precious metals, daily goods clothes, and other various services are sold.

In the present embodiment, Wilcoxon's of two survey subjects (for example, airlines 1 and 2) is compared with one evaluation index (for example, satisfaction) to determine whether or not there is a statistically significant difference. The test method shall be used. In addition, when comparing a plurality of evaluation indices for two survey subjects, Wilcoxon's signed rank test shall be used. In addition, the Kruskal-Wallis test is used when comparing one evaluation index for three or more survey targets. In addition, the Friedman test shall be used when comparing multiple evaluation indices for three or more survey targets.

[System configuration]
FIG. 1 is a diagram showing an outline of a configuration example of a value evaluation support system 1 according to the first embodiment of the present invention. The value evaluation support system 1 is configured such that a value evaluation server 10, a plurality of product etc. providing systems 20 (20a to 20d in the example of FIG. 1), and a user terminal 30 are connected to a network 40 and can communicate with each other. have.

The product etc. providing system 20 is an information processing system for sellers of products etc. to sell products etc. In this embodiment, for example, it is a computer system for selling airline tickets at each airline. is there. The product etc. providing system 20 stores information for selling products etc. (air tickets) in a product etc. content database (DB) 201 (201a to 201d in the example of FIG. 1). The product content DB 201 includes, for example, various information necessary for sales of products such as an airline name, departure place, departure time, arrival place, arrival time, seat number, model, price, passenger personal information, and the like.

The value evaluation server 10 is composed of a server device or the like that is operated and managed by a business operator or an information search service provider that mediates sales of goods and the like (air ticket) between a user and a seller (airline company). It is a computer system that stores information about users and sellers and information evaluated by users or sellers regarding products and the like. In addition, when a product or the like is traded between a user and a seller, information on the product etc. that meets the conditions specified by the user is collected from each product etc. providing system 20, and the difference in their values Based on this, a standard for selecting a product or the like (or a seller who sells the product or the like) is recommended to the user.

The value evaluation server 10 includes a server device having a general configuration. For example, a product etc. information acquisition unit 11, an evaluation history information acquisition unit 12, a statistical calculation unit 13, and an evaluation result output unit 14 implemented by a software program. And each part such as a questionnaire processing part 15. Each of these units can be implemented as, for example, a web application that runs on a web server program (not shown). For example, the storage device such as an HDD or a memory on the value evaluation server 10 or the value evaluation server 10 can read the unit. It is stored in a storage medium such as an optical disk. You may hold | maintain in the other file server etc. which can be accessed via networks, such as the internet.

Further, the value evaluation server 10 has data such as a seller DB 101, a user DB 102, a product etc. DB 103, an evaluation history DB 104, and a questionnaire DB 105, which are made up of databases and file tables.

The seller DB 101 is a table that holds information related to the seller (airline company) and the merchandise provision system 20 of the seller. For example, account information such as IDs and passwords used in the value evaluation support system 1, sales, etc. Various information such as a person's name, location and other attribute information, and seller's characteristic information (for example, in the case of an airline, information related to airports, hotels, land traffic, accident insurance, etc.) are included. The information in the seller DB 101 is registered in advance by, for example, the administrator of the value evaluation server 10.

The user DB 102 is a table that holds information related to users, and includes, for example, various information such as account information such as IDs and passwords used in the value evaluation support system 1 and attribute information such as user names. . The information in the user DB 102 is assumed to be registered in advance by each user or the administrator of the value evaluation server 10. The product etc. DB 103 is a table that holds information related to the product etc. extracted from each product etc. providing system 20, for example, attribute information such as product name, model, spec, release date, price, product image, description, etc. Various information such as sentences is included.

The evaluation history DB 104 is a table that accumulates and holds information related to questionnaire results (value evaluation results) for products and the like evaluated and input by the user. For example, each question (evaluation index) of the questionnaire and Response results (evaluation data), response reception date (entry date), respondent (entrant), information on the respondent's user terminal 30, various information such as the product subject to the questionnaire, seller, manufacturer name, etc. Is included.

The questionnaire DB 105 is a table that holds information related to the contents of a questionnaire to be filled in when a user purchases or uses a product, for example, various questions associated with a product, a seller, or a manufacturer name. And various questions for entering information such as the respondent, the reply date, the information on the user terminal 30 of the respondent, and the like. The information in the questionnaire DB 105 is registered in advance by the administrator of the value evaluation server 10 or the like.

In addition, the product etc. information acquisition unit 11 requests each product etc. providing system 20 based on the conditions specified by the user to obtain information on the relevant products etc., or from each product etc. providing system 20. A suitable product etc. is searched and extracted, and information relating to the obtained product etc. is stored in the product etc. DB 103. The evaluation history information acquisition unit 12 extracts, for each evaluation index, history information of corresponding evaluation data (questionnaire results) from the evaluation history DB 104 for each product held in the product etc. DB 103.

The statistical calculation unit 13 implements calculation formulas and algorithms for performing various statistical calculations including a non-parametric method, and calculates a difference in value for each product held in the product etc. DB 103. This includes data such as various test tables and random number tables for performing statistical calculations. In mounting, for example, a statistical calculation function or various libraries included in an existing application program such as spreadsheet software can be used as appropriate.

The evaluation result output unit 14 compiles and outputs the statistical results calculated and evaluated by the statistical calculation unit 13 and information on the value difference of the product, etc., and determines the selection criteria for the appropriate product etc. It is output to the user as recommended information related to selection criteria for products and the like. When the user purchases a product or the like and uses it, the questionnaire processing unit 15 acquires the corresponding questionnaire information from the questionnaire DB 105 and presents it to the user to carry out the questionnaire. In addition, evaluation information is input by acquiring information related to a questionnaire result (evaluation data) input by the user and storing it in the evaluation history DB 104.

The user terminal 30 is an information processing terminal that is operated by a user who purchases a product such as an airline ticket, and includes, for example, a personal computer, a mobile terminal such as a mobile phone or a smartphone. A digital TV, game machine, karaoke machine, or the like capable of bidirectional communication may be used. The user terminal 30 can execute various functions by accessing a web server program (not shown) on the value evaluation server 10 or the product providing system 20 via a web browser (not shown), for example. The network 40 is a communication network composed of, for example, a wired / wireless LAN (Local Area Network) line, a satellite line, a telephone line, an optical line, and the like. A typical example is the Internet network.

[Processing flow (main processing)]
FIG. 2 is a flowchart showing an outline of an example of the flow of main processing in the value evaluation server 10. First, an initial registration process for registering seller and user information is performed (S01). Here, for example, the administrator or operator of the value evaluation server 10 may initially register such information with respect to each DB, or a seller who does not have a user ID used in the value evaluation support system 1 or uses it. Initial registration may be performed when there is an access from a person. At this time, for example, registration information input to the DB such as personal information other than the user ID is received from the seller or user who performs registration, and after the information is officially registered in the DB, the user ID is issued. Register a password.

Next, in response to a usage request from the user, a price survey process is performed to obtain information including prices related to products that meet the conditions by inquiring each product etc. providing system 20 (S02). Next, for the products obtained in step S02, historical information of past value evaluation is extracted and statistically processed, thereby statistically determining whether or not there is a significant difference in the value of these products. In accordance with the result, a selection criterion for an appropriate product or the like is determined based on the result, and a value evaluation process recommended to the user is performed (S03). Further, for a user who selects and purchases a product etc., a questionnaire survey on the value of the purchased product etc. is performed, and an evaluation history information recording process for recording information related to the result is performed (S04). The process ends.

[Processing flow (price survey processing)]
FIG. 3 is a flowchart showing an outline of an example of the flow of the price survey process (step S02 in FIG. 2) in the value evaluation server 10. First, a user who wants to purchase a product etc. operates a homepage screen such as a Web browser on the user terminal 30, a service menu screen, a standby screen, etc., and accesses the value evaluation server 10 to make a usage request. Perform (S201). Upon receiving the use request, the value evaluation server 10 outputs a menu screen for selecting or inputting a product or the like to the user terminal 30 by the product etc. information acquisition unit 11 or the like (S202).

FIG. 4 is a diagram showing an example of a menu screen for the user to select or input a product or the like. In the example of FIG. 4, a site for purchasing a ticket for using a transportation means is taken as an example. For example, the user selects a desired transportation means such as an airplane, a ship, a train, etc. from a pull-down menu of the transportation means. . In addition, the name of the company or organization having the selected means of transportation is displayed in the pull-down menu of the transportation facility name, and the user selects the target company or organization. Similarly, the flight name of the selected company or group is displayed in the pull-down menu, and the user selects the target flight.

Travel options are displayed on the pull-down menu for items such as round trip, one-way, and round trip, and are selected by the user. For example, if the user selects a round trip, as shown in the example of FIG. 4, whether the departure place (departure airport), departure date, departure time, and arrival place (arrival airport) for the outbound route and the return route are selected from the pull-down menu. A field to be specified by direct input is displayed. When tour is selected as the travel mode, it is possible to input a plurality of routes such as

routes

1, 2, 3,.

Fare class can be selected from the pull-down menu such as first, business, economy, etc., and the number of travelers can be selected from the pull-down menu or directly enter the desired number. Search options include search options such as non-stop flights and cheapest order.For example, if you select non-stop flights, only non-stop flights will be searched, and if you select the cheapest order, the search results will be displayed in order from the cheapest. Is displayed. It is possible to select multiple options. When no option is selected, for example, flights other than non-stop flights are displayed in order from the highest to the lowest.

When the user selects or inputs necessary information via the menu screen as shown in FIG. 4, the user terminal 30 requests the value evaluation server 10 for information such as a product that satisfies the condition ( S203). Upon receiving the request, the value evaluation server 10 transmits a search request for a product or the like satisfying the condition to the product etc. providing system 20 of all the sellers registered in advance by the product etc. information acquisition unit 11 (S204). ).

The merchandise providing system 20 of each seller that has received the search request extracts information on the merchandise that satisfies the condition from the merchandise content DB 201 and responds to the value evaluation server 10 (S205). The product etc. information acquisition unit 11 of the value evaluation server 10 is configured to be able to directly access the product etc. content DB 20 of each product etc. providing system 20, and the product etc. information acquisition unit 11 meets the conditions. May be directly searched and extracted.

The product etc. information acquisition unit 11 that has acquired or extracted information on the product etc. that meets the conditions stores the acquired product etc. information in the product etc. DB 103 (S206), and ends the price survey process. The product information stored in the product etc. DB 103 includes, for example, departure place (departure airport), departure date, scheduled departure time, arrival place (arrival airport), arrival date, estimated arrival time, required time, aircraft name, model It includes items such as airline name, flight number, number of remaining seats, number of remaining seats, meal menu, passenger information input form, and fare. Note that the above-described transportation means searching method is a known means that has already been carried out by airline ticket sales etc. on the Internet, and the same method can be used in this embodiment.

[Processing flow (value evaluation processing)]
FIG. 5 is a flowchart showing an outline of an example of the flow of value evaluation processing (step S03 in FIG. 2) in the value evaluation server 10. First, the value evaluation server 10 is each survey object extracted in the price survey process of FIG. 3 by the evaluation history information acquisition unit 12 (that is, each product acquired in step S206 of FIG. 3). Then, the evaluation history information is extracted from the evaluation history DB 104 (S301).

The evaluation history information is information in which the purchaser's evaluation for each survey target (product or the like) is recorded, and is information recorded in the evaluation history DB 104 in an evaluation history information recording process (step S04 in FIG. 2) described later. Therefore, the evaluation history information acquisition unit 12 searches the evaluation history information recorded in the evaluation history DB 104 in the past, thereby evaluating the evaluation history associated with each product acquired in step S206 of FIG. Information will be extracted.

Evaluation history information is evaluation information evaluated by a user who purchased a product after purchasing, using, or consuming the product, and includes various evaluation indexes. Evaluation indicators include, for example, overall satisfaction, comfort, luxury of in-flight meals, attitudes of cabin attendants, ease of use of in-flight audio, etc., satisfaction of video / music provided, etc. It is. Each evaluation index is expressed by, for example, evaluation points (0 to 100 points).

Next, the value evaluation server 10 determines whether or not the number of survey targets is 2 or more (S302), and if it is less than 2, the process directly proceeds to step S340. Here, as described above, the number of survey targets is the number of types of airline tickets acquired in step S206 of FIG. 3 in the case of the present embodiment, and one type of all airlines in step S205 of FIG. When the ticket is extracted and transmitted to the value evaluation server 10, the number of airlines may be used in place of the number of types of airline tickets (that is, the survey target may be each airline).

In step S302, when the number of investigation targets is 2 or more, next, a loop process for repeatedly performing the process for each of all the evaluation indexes included in the evaluation history information acquired in step S301 is started. In each loop process, first, the number of data of evaluation indexes to be processed in the loop is counted for each investigation object, and it is determined for each investigation object whether or not it is equal to or more than a preset lower limit value (S303). If it is less than the lower limit, it is determined that there is no significant difference because there is not enough data for the evaluation index, and the process proceeds to the next evaluation index in a loop process. At this time, for example, when displaying the result by the process described later, a message such as “There is not enough data because there is not enough data” may be added. Here, for example, the lower limit value is a value obtained by adding two or more numbers to the number of data generally determined that sufficient accuracy cannot be obtained at the time of statistical calculation, and does not exceed the upper limit value described later. Set a value (eg 6).

If it is determined in step S303 that the number of data for all survey targets is equal to or greater than the lower limit value, it is next determined whether or not each counted data number is equal to or less than a preset upper limit value (S304). Here, as the upper limit value, for example, a value obtained by subtracting 1 from the minimum number that is generally recognized to be equivalent to the number of populations (the number of populations) (for example, 600,000) is set. If the number of data is less than or equal to the upper limit value, the process proceeds to step S306 with all data as processing targets (that is, all data is extracted as samples).

On the other hand, if the number of data exceeds the upper limit value in step S303, the upper limit number of data is extracted as a sample in order from the latest data among the target evaluation index data (S305), and the process proceeds to step S306. . Here, the sample is extracted in order from the latest one, but other extraction methods may be used. Moreover, although the upper limit number is used as the number to be extracted, an appropriate number can be appropriately determined as long as it is a number equal to or less than the upper limit number (and a number equal to or greater than the lower limit value).

In step S306, a random number table or the like held in advance in the value evaluation server 10 is used, and the target evaluation index is smaller than a predetermined number, for example, the number of data extracted from the sample data extracted for each survey target. A number of (lower limit value-1) or more data is randomly extracted as samples (S306). By the processing in steps S303 to S306, the processing target is limited to the number of data within a range in which significant difference determination can be performed without reducing the accuracy of statistical calculation performed in step S320, which will be described later, and inappropriate evaluation as a sample of statistical calculation Preventing extraction of index data. As a result, the calculation load of the value evaluation server 10 can be reduced, and the processing speed of statistical calculation can be improved while ensuring sufficient accuracy.

The method of extracting the sample for statistical calculation is not limited to the method of steps S303 to S306, but instead, for example, the number of evaluation index data to be processed in each loop processing is counted for each survey target. Randomly extract the number of data less than the counted data and less than the minimum number recognized as a population and more than the statistically inferior accuracy from the counted data by survey target It is also possible. In this method, if the number of counted data is less than the minimum value that can be recognized as a population, the sample is substantially a random extraction from the sample, and the counted number of data is greater than or equal to the minimum value that is recognized as a population. Is essentially a random extraction from the population. Thus, if two samples with different extraction sources are equally treated as randomly extracted samples and used for statistical calculation, the statistical accuracy may be lowered.

Next, the value evaluation server 10 uses the statistical calculation unit 13 to perform a statistical process for performing a predetermined statistical calculation on the processing target data of the target evaluation index to determine a significant difference between the evaluation indexes (S320). 6 and 7 are flowcharts showing an overview of an example of the flow of statistical processing in step S320 of FIG. The statistical calculation unit 13 first determines whether or not the number of survey targets is 2 (S3201). When the number of survey targets is not 2 (that is, the number of survey targets is 3 or more), as will be described later, an estimation regarding each population is performed by Kruskal-Wallis test. On the other hand, when the number of objects to be investigated is two, the two populations are estimated by the Wilcoxon test.

[Wilcoxon test]
First, it is determined whether or not the data recorded for each survey target for the target evaluation index has the same value (S3202). That is, it is determined whether there is an identical value in the entire set of evaluation index data (evaluation data) possessed by each of the two survey targets. Instead of determining the presence or absence of the same value, the presence or absence of the same rank may be determined for the rank data calculated from the evaluation data by the process described later.

FIG. 8 is a diagram showing an example in which a significant difference is determined by statistical calculation from the values of the evaluation indices for each of the two survey targets. In the table of FIG. 8, the evaluation data conversion rules (hereinafter referred to as “evaluation”) defined in the table shown in the example of FIG. Satisfaction level (0 to 100 points), which is an evaluation index that has been converted (hereinafter may be described as “evaluation data conversion”) based on “data conversion rules”) Has been. That is, the evaluation points in FIG. 8 are evaluation points for use in the analysis by statistical calculation. For example, the evaluation data conversion rule defined in the definition table in FIG. 28 (for example, “100− (Original evaluation score) "is converted.

FIG. 8 shows an example in which data of the same value does not exist for all the evaluation data. In this case, in FIG. 6, the minimum value of the number n _i of evaluation data possessed by each survey object is a predetermined threshold ( For example, it is determined whether it is less than 15) (S3203). In the example of FIG. 8, since the number n ₁ = 6 of the evaluation data of the airline _{1 and} the number n ₂ = 5 of the evaluation data of the airline 2 and the minimum value is n ₂ = 5, the threshold (15) Is less than.

In this case, next, the rank in the set of the entire evaluation data for each evaluation data and the sum (rank rank) of the ranks for each survey target are calculated (S3204). In the example of FIG. 8, information on ranks and rank sums calculated for each evaluation data is also written in the table. Next, the amount of W is calculated based on the ranking data (S3205). In the present embodiment, the rank sum of the survey targets with the smaller number of evaluation data among the survey targets is set as the W amount. In the example of FIG. 8, since n ₁ > n ₂ , the rank sum 25 of the airline 2 is set as the W amount. When the number of evaluation data is the same, the rank sum of any one of the survey targets is the W amount. It is not always necessary to use the sum of ranks of the survey targets with the smaller number of evaluation data as the amount of W, and the rank sum of the survey targets with the larger number may be used. In this case, not W amount but W ′ amount is described.

Next, the lower limit value W _L (α / 2) and the upper limit value W _U (α / 2) of the predetermined significance level α are identified from the Wilcoxon test table held in advance by the statistical calculation unit 16 and the like. _W and _{L (α} / 2) and _{W U} (α / 2) that is, compares the W amount calculated in step S3205 (S3206). Note that α in the Wilcoxon test table is α / 2 = 0.025 when a significant difference is determined by a two-sided test with a significance level of 5%. Further, m in the number of data (m, n) is a value with the smaller number of evaluation data, and n ₂ = 5 in the example of FIG. Further, n is a value with the larger number of evaluation data, and n ₁ = 6 in the example of FIG. Therefore, W _L (0.025) = 18 and W _U (0.025) = 42 are obtained from the Wilcoxon test table.

In the Wilcoxon test table, a type that obtains a significant point (limit value) from the number of data (m, n) and the significance level α and a significance probability (P value) from the number of data (m, n) and the amount of W are obtained. There are types, but the former test table is used for comparison here. The latter test table is used when calculating the P value in the selection of supplementary explanation items to be described later.

Next, based on the comparison result in step S3206, it is determined whether or not there is a significant difference between the survey targets (S3207), and the statistical processing is terminated. Here, when W _L (α / 2) <W amount <W _U (α / 2), it is determined that there is no significant difference, and W amount ≦ W _L (α / 2) or W _U When (α / 2) ≦ W amount, it is determined that there is a significant difference. In the example of FIG. 8, since {W _L (0.025) = 18} <(W amount = 25) <{W _U (0.025) = 42}, it is determined that there is no significant difference. To do.

If there is no value to compare in the Wilcoxon test table because the number of evaluation data n _i in each survey target is small, it cannot be said that there is a significant difference because there is not enough data. Make a decision. At this time, when the result is displayed by the processing described later, for example, a message such as “There is not enough data and it cannot be said that there is a significant difference as a result” may be added. Thereby, even when there is no value to be compared in the test table, a significant difference can be determined. In addition, since the value of n is very large, if there is no value to be compared with the Wilcoxon test table, the process of step S3210 and later described below may be performed instead of the process of step S3206 and subsequent steps.

On the other hand, if the minimum value of the number of evaluation data is greater than or equal to the threshold (15) in step S3203, the rank of each evaluation data and the rank sum for each survey target are calculated by the same process as in steps S3204 and S3205 described above. Then, the amount of W is calculated (S3208, S3209). Further, the u ₀ amount is calculated by the following equation (S3210).

You may use the following formula | equation which performed the correction called continuous correction which improves statistical accuracy.

FIG. 9 is a diagram illustrating another example in which a significant difference is determined by statistical calculation from the values of the evaluation indices for each of two survey targets. In the table of FIG. 9, as in FIG. 8, the evaluation data conversion is performed for the two airlines (airlines 1 and 2) to be investigated based on the evaluation data conversion rule shown in FIG. Satisfaction as an evaluation index is indicated by evaluation points (0 to 100 points). In FIG. 9, there is no data having the same value for all evaluation data, and the minimum number of evaluation data is the number n ₁ = 15 of evaluation data of the airline 1, which is equal to or greater than the threshold (15). Yes. Moreover, the rank calculated about each evaluation data and the information of the rank sum for every investigation object are written together in the table | surface. At this time, the amount of W = 292 and the number of data (m, n) = (15,16) from the sum of ranks of the airline 1 which has the smaller number of evaluation data. , U ₀ amount = 2.06.

After that, the limit value u (α) is specified from the predetermined significance level α of the normal distribution table held in advance by the statistical calculation unit 16 or the like, and the obtained u (α) and u ₀ obtained in step S3210. The absolute value of the quantity is compared (S3211). When a significant difference is determined by a two-sided test with a significance level of 5%, a limit value u (0.05) = 1.96 is obtained from the normal distribution table.

The normal distribution table includes a type that handles one-side rejection area and a type that handles both-side rejection areas. In the present embodiment, the normal distribution table that handles the latter-side rejection areas is used. The normal distribution table includes a type for obtaining a significance point (limit value) from the significance level α and a type for obtaining a significance probability (P value) from the absolute value of the u ₀ quantity. The former table is used for comparison. The latter table is used when calculating the P value in the selection of supplementary explanation items to be described later.

Next, based on the comparison result in step S3211, it is determined whether or not there is a significant difference between the survey targets (S3212), and the statistical processing is terminated. Here, it is determined that there is a significant difference when the absolute value of the u ₀ quantity is not less than the limit value u (α), and it is determined that there is no significant difference when the absolute value is less than the limit value u (α). To do. In the example of FIG. 9, since {| u ₀ | = 2.06}> {u (0.05) = 1.96}, it is determined that there is a significant difference.

If there are evaluation data having the same value in step S3202 in FIG. 6, the rank of each evaluation data is calculated in the same manner as in steps S3204 and S3208 described above (S3213). Since the same rank is calculated, an average rank is assigned to the same rank (S3214). The average rank is also called an intermediate rank. The average rank is calculated by the equation {(a + 1) + (a + t _j )} / 2. Here, a is the previous rank of the same rank, and j is a group number when the ranks of evaluation data groups having the same rank are numbered in ascending order. T _j is the number of evaluation data of the j-th group.

FIG. 10 is a diagram showing another example in which a significant difference is determined by statistical calculation from the values of evaluation indexes for two survey targets. In the table of FIG. 10, satisfaction that is an evaluation index obtained by performing evaluation data conversion based on the evaluation data conversion rule shown in FIG. Degrees are indicated by evaluation points (0 to 100 points). FIG. 10 shows an example in the case where data of the same value exists among all the evaluation data, and the rank calculated for each evaluation data and the information of the rank sum for each survey target are also shown in the table. .

Here, in the table of FIG. 10, there are two pieces of evaluation data having the first rank, and three pieces of evaluation data having the fifth rank (t ₁ = 2 and t ₂ = 3). When the average rank is calculated for these two groups by the above formula, the average rank of the group ranked first is 1.5, and the average rank of the group ranked fifth is 6. The table of FIG. 10 also shows the average rank including these calculation results and the rank sum information.

Next, based on the average rank data calculated in step S3214, the amount of W is calculated by the same processing as in steps S3205 and S3209 described above. For distinction, this is referred to as W * amount (S3215). Further, the u ₀ * amount is calculated by the following equation. (S3216).

Ｊ where j is an integer of 1 to the number g of evaluation data groups having the same rank.

In the example of FIG. 10, the W * amount = 12, and from Equation 3, u ₀ * amount = −2.00. Thereafter, the limit value u (α) is specified from the predetermined significance level α in the normal distribution table by the same processing as in step S3211 described above, and u (α) obtained is obtained, and u ₀ * obtained in step S3216. The absolute value of the quantity is compared (S3217). Next, based on the comparison result in step S3217, it is determined whether or not there is a significant difference between the survey targets (S3218), and the statistical processing is terminated. Here, it is determined that there is a significant difference when the absolute value of the u ₀ * quantity is equal to or greater than the limit value u (α), and it cannot be said that there is a significant difference when the absolute value is less than the limit value u (α). judge. In the example of FIG. 10, in a two-sided test with a significance level of 5%, {| u ₀ * | = 2.00}> {u (0.05) = 1.96}, so it is determined that there is a significant difference. .

When the number of data (m, n) in the Wilcoxon test with the same value is less than a predetermined threshold (for example, m <5, which is generally considered to be inferior in the accuracy of statistical calculation), or the same rank The maximum value of the ratio {t _j / (m + n)} of the number t _j of evaluation data in each group and the total number of data m + n is equal to or greater than a predetermined threshold (for example, the maximum that the statistical calculation accuracy is generally inferior) When value ≧ 0.7), it is not necessary to perform statistical calculation because sufficient accuracy cannot be ensured. In this case, it may be determined that there is no significant difference because there is not enough data. At this time, when the result is displayed by the processing described later, for example, a message such as “There is not enough data and it cannot be said that there is a significant difference as a result” may be added. Further, after executing the statistical calculation, for example, a message such as “the accuracy of the result of the statistical calculation is not high” may be added.

[Kruskal Wallis test]
If the number of survey objects in step S3201 in FIG. 6 is 3 or more, the process proceeds to FIG. 7 to perform Kruskal-Wallis test. First, as in step S3202 in FIG. It is determined whether there is an evaluation data having the same value in the entire set (S3219 in FIG. 7). Instead of determining the presence or absence of the same value, the presence or absence of the same rank may be determined for the rank data calculated from the evaluation data by the process described later.

FIG. 11 is a diagram showing an example in which a significant difference is determined by statistical calculation from the values of evaluation indices for each of three survey targets. In the table of FIG. 11, satisfaction that is an evaluation index obtained by performing evaluation data conversion on the three airlines (airlines 1 to 3) to be investigated based on the evaluation data conversion rule shown in FIG. 28 described later. Degrees are indicated by evaluation points (0 to 100 points). FIG. 11 shows an example of the case where there is no data having the same value for all evaluation data. In this case, the total number N of evaluation data in each survey target is less than a predetermined threshold (for example, 15). It is determined whether or not (S3220). For the number of surveyed 3 above, the data the total number N as the sum of the number of evaluation data of each study _is represented by _{_{n 1 + n 2 + n 3}} + ... + n i + ... n k (k is investigated Number of objects).

In the example of FIG. 11, the number of evaluation data n ₁ = 4 of the airline ₁ , the number of evaluation data n ₂ = 4 of the airline ₂ , the number of evaluation data n ₃ = 4 of the airline 3, and the total number of data N Since = 12, it is less than the threshold value (15). In this case, next, the minimum value n _MIN is specified for the number n _i of the evaluation data of each investigation target (S3221). Then, it is judged whether the investigation with evaluation data for the number above the minimum value _{n MIN} of the specified number of data in step S3221 (S3222). When there is a survey object having a number of evaluation data exceeding the minimum value, a minimum number of evaluation data is randomly extracted from the evaluation data of the survey object using a random number table or the like held in advance in the value evaluation server 10 (S3223). As a result, the number n _i of evaluation data for each survey target is unified to the minimum value n _MIN .

Thereafter, by the same processing as in step S3204 in FIG. 6 and the like, the rank in the set of the entire evaluation data for each evaluation data and the sum (rank sum) of the ranks for each survey target are calculated (S3224). In the example of FIG. 11, information on ranks and rank sums calculated for each evaluation data is also written in the table. Next, based on this ranking data, the KW amount is calculated by the following equation (S3225).

Here, i is an integer from 1 to the number of survey targets k, and R _i is the sum of ranks in each survey target.

Next, from the Kruskal-Wallis test table (when the number of repetitions is the same) held in advance by the statistical calculation unit 16 or the like, the number of evaluation data of the investigation object unified to a predetermined significance level α, the number of investigation objects k, and the minimum value A significant point kw (α) corresponding to n _MIN is specified, and the obtained kw (α) is compared with the KW amount calculated in step S3225 (S3226). In the example of FIG. 11, the KW amount is 6.62 according to the above equation 4, and when the significant difference is determined by the two-sided test with the significance level of 5%, the number of repetitions is k = 3 and n _MIN = 4. Kw (0.05) = 5.69 is obtained from the Kruskal-Wallis test table in the case of being equal (when the number of evaluation data of each investigation object is equal).

If the Kruskal-Wallis test table when the number of repetitions is not equal is stored in advance in the statistical calculation unit 16 or the like, the processing in steps S3221 to S3223 in FIG. 7 is not necessary, and the number of evaluation data for each survey target The KW amount can be calculated in step S3225 without unifying the minimum values. As a result, the amount of KW can be calculated without reducing the number of samples, and it is possible to simultaneously improve the reliability of statistical results and reduce the computer calculation load by simplifying the calculation steps. In this case significant point kw (alpha) is significance level alpha, chosen from those corresponding to the survey number k and evaluation data number n _i of each study.

In addition, the Kruskal-Wallis test table includes a type that obtains a significant point (limit value) from the number k of survey targets, the number of data n _MIN or _ni , and the significance level α, the number of survey targets k, and the number of data n _MIN. Alternatively, there is a type that obtains a significance probability (P value) from the n _i and the KW amount, but the former test table is used for comparison here. The latter test table is used when calculating the P value in the selection of supplementary explanation items to be described later.

Next, based on the comparison result in step S3226, it is determined whether there is a significant difference between the survey targets (S3227), and the statistical processing is terminated. Here, it is determined that there is a significant difference when the KW amount ≧ kw (α), and it is determined that there is no significant difference when the KW amount <kw (α). In the example of FIG. 11, since (KW amount = 6.62)> {kw (0.05) = 5.69}, it is determined that there is a significant difference.

If there is no value to be compared in the Kruskal-Wallis test table because the number n _i of the evaluation data in each survey target is small, it cannot be said that there is a significant difference because there is not enough data. Judgment is made. At this time, when the result is displayed by the processing described later, for example, a message such as “There is not enough data and it cannot be said that there is a significant difference as a result” may be added. Thereby, even when there is no value to be compared in the test table, a significant difference can be determined.

In step S3220 in FIG. 7, when the total number N of data is equal to or greater than a predetermined threshold (for example, 15), the rank of each evaluation data and the rank sum for each survey target are calculated by the same processing as in steps S3224 and S3225 described above. To calculate the KW amount (S3228, S3229). FIG. 12 is a diagram showing another example in which a significant difference is determined by statistical calculation from the values of the evaluation indices for each of three survey targets. In the table of FIG. 12, satisfaction is an evaluation index obtained by performing evaluation data conversion for the three airlines (airlines 1 to 3) to be investigated based on the evaluation data conversion rule shown in FIG. Degrees are indicated by evaluation points (0 to 100 points). In FIG. 12, there is no data having the same value for all evaluation data, and the total number of evaluation data N = 16, which is equal to or greater than the threshold (15). Moreover, the rank calculated about each evaluation data and the information of the rank sum for every investigation object are written together in the table | surface. At this time, the amount of KW is 5.11 from the above equation (4).

Next, a predetermined significance level α and a limit value χ ² (φ, α) given with a degree of freedom φ = k−1 are specified from the chi-square distribution table previously held by the statistical calculation unit 16 and the like. The obtained χ ² (φ, α) is compared with the KW amount calculated in step S 3229 (S 3230). In the example of FIG. 12, the limit value χ ² (2,0.05) = 5.99 is obtained from the chi-square distribution table from the significance level α = 0.05 and the degree of freedom φ = 3-1 = 2. The chi-square distribution table includes a type that obtains a significant point (limit value) from a degree of freedom φ and a significance level α, and a significance probability (P value) from statistically calculated values such as the degree of freedom φ and the amount of KW. However, the former test table is used for comparison here. The latter test table is used when calculating the P value in the selection of supplementary explanation items to be described later.

Next, based on the comparison result in step S3230, it is determined whether there is a significant difference between the survey targets (S3231), and the statistical processing is terminated. Here, it is determined that there is a significant difference when the KW amount ≧ χ ² (φ, α), and it is determined that there is no significant difference when the KW amount <χ ² (φ, α). . In the example of FIG. 12, since (KW amount = 5.11) <{χ ² (2,0.05) = 5.99}, it is determined that there is no significant difference.

If there is an evaluation data having the same value in step S3219 in FIG. 7, the ranking of each evaluation data is calculated (S3232) as in steps S3224 and S3228 described above, and the evaluation data having the same value is further calculated. Since the same rank is calculated, an average rank is assigned to the same rank by the same process as step S3214 in FIG. 6 (S3233).

FIG. 13 is a diagram showing another example in which a significant difference is determined by statistical calculation from the values of evaluation indexes for each of three survey targets. In the table of FIG. 13, the satisfaction is an evaluation index obtained by performing evaluation data conversion on the three airlines (airlines 1 to 3) to be investigated based on the evaluation data conversion rule shown in FIG. Degrees are indicated by evaluation points (0 to 100 points). FIG. 13 shows an example in the case where data of the same value exists among all the evaluation data, and the rank calculated for each evaluation data and the information of the rank sum for each survey target are also shown in the table. .

Here, in the table of FIG. 13, there are three pieces of evaluation data having the third rank, and two pieces of evaluation data having the seventh rank (t ₁ = 3, t ₂ = 2). When the average rank is calculated for these two groups by the formula used in step S3214 of FIG. 6, the average rank of the group ranked third is 4, and the average rank of the group ranked seventh is 7. 5 The table of FIG. 13 also shows the average rank including these calculation results and the rank sum information.

Next, based on the average rank data calculated in step S3233, the KW amount is calculated by the same processing as in steps S3225 and S3229 described above. For distinction, this is described as KW * amount (S3234). Further, the KW ′ amount is calculated by the following equation. (S3235).

Here, j is an integer from 1 to the number g of evaluation data groups having the same rank.

In the example of FIG. 13, KW * amount = 1.42 from the above equation 4, and KW ′ amount = 1.48 from the equation 5. Thereafter, the limit value χ ² (φ, α) is specified from the predetermined significance level α of the chi-square distribution table and the degree of freedom φ = k−1 by the same processing as in step S3230, and the obtained χ ² (Φ, α) is compared with the amount of KW ′ obtained in step S3235 (S3236). Next, based on the comparison result in step S3236, it is determined whether there is a significant difference between the survey targets (S3237), and the statistical processing is terminated. Here, it is determined that there is a significant difference when KW ′ amount ≧ χ ² (φ, α), and it cannot be said that there is a significant difference when KW ′ amount <χ ² (φ, α). judge. In the example of FIG. 13, in the two-sided test with a significance level of 5%, (KW ′ amount = 1.48) <{χ ² (2,0.05) = 5.99}. Judge that you can not say.

Note that the number of survey targets k in the Kruskal-Wallis test with the same value or the minimum value n _MIN of the number of evaluation data n _i of each survey target is less than a predetermined threshold (for example, generally the accuracy of statistical calculation is When k <3, which is considered inferior, or the minimum value n _MIN <6), it is not necessary to perform statistical calculation because sufficient accuracy cannot be secured. In this case, it may be determined that there is no significant difference because there is not enough data. At this time, when the result is displayed by the processing described later, for example, a message such as “There is not enough data and it cannot be said that there is a significant difference as a result” may be added. Further, after executing the statistical calculation, for example, a message such as “the accuracy of the result of the statistical calculation is not high” may be added.

In the statistical calculation described above, the Kruskal-Wallis test is used when the number of survey targets k ≧ 3. However, when applied when k = 2, it matches the Wilcoxon test of the two-sided test. Therefore, in this embodiment, the Kruskal-Wallis test can be applied even when k = 2. However, the accuracy of the test using the chi-square distribution table is inferior. Moreover, it does not agree with the Wilcoxon test of the one-sided test described later.

When the statistical processing in step S320 in FIG. 5 is completed through the above processing, the series of processing in steps S303 to S320 is repeated for the next evaluation index by loop processing in the processing flow in FIG. When the series of processing in steps S303 to S320 is repeated for all evaluation indexes, the loop processing is terminated, and then it is determined whether all the evaluation indexes have a significant difference (whether there is a significant evaluation index). (S307) If it cannot be said that all the evaluation indices are significantly different (there is no significant evaluation index), the process proceeds to step S340.

If there is at least one significant evaluation index in step S307, the evaluation result output unit 14 of the value evaluation server 10 can compare the price of the product etc. with the value of the significant evaluation index for each survey target. Is created (S308). FIG. 14 is a diagram illustrating an example in which the price of a product and a significant evaluation index are represented in a tabular format for each survey target. In FIG. 14, for each of the three survey targets (airlines 1 to 3), a significant difference is determined for each index in the evaluation history information recorded in the evaluation history DB 104 by the statistical processing in step S320. An example is shown in which there is a significant difference in indicators (“quality of AV equipment”, “attitude of CA”, “luxury of meals”). For each airline, the price of the ticket on the desired route and the average value of the evaluation points of each evaluation index in the evaluation history information are displayed in a list so that they can be compared. As the average value, for example, the evaluation in the table of FIG. As the average value of the points, a value obtained by performing inverse conversion based on an evaluation data conversion rule shown in FIG. 28 described later (that is, an average value of values of original evaluation data that is not subjected to evaluation data conversion) can be used. .

The example of FIG. 14 shows a case where the evaluation index is a metric value called an evaluation point. However, if the evaluation index is not a metric value but a count value such as a probability or the number of defects, before calculating the average value. A conversion for approximating the measured value may be performed. For example, logit transformation or inverse sine transformation can be used for probability, and square root transformation or logarithmic transformation can be used for the number of defects. Since the percentage (probability) or the number of defects follows a binomial distribution or a Poisson distribution, respectively, data may be converted to a normal distribution. Thereby, it can be regarded as data according to a normal distribution like the measurement value. The percentage can be normalized by using a logit transform or an inverse sine transform after being expressed as a probability.

The logit transformation is represented by L (P) = ln {P / (1-P)}. Here, P is a probability and L (P) is a logit. The inverse sine transformation is represented by Sin ⁻¹ √P. The square root transformation is represented by √λ, and the logarithmic transformation is represented by lnλ. Here, λ is the number of defects. If x and n in x / n (where x is the total number of occurrences or defects and n is the number of trials or the number of trials) are known, continuous correction is performed as a data conversion formula. Add to the normal approximation better. The logistic transformation or inverse sine transformation continuous correction is expressed by P * = (x + 0.5) / (n + 1), and the value of P * is used in place of P in the above-mentioned logit transformation or inverse sine transformation expression. That's fine. Further, continuous correction of square root transformation or logarithmic transformation is represented by λ * = (x + 0.5) / n, and λ * may be used in place of λ in the above square root transformation or logarithmic transformation formula.

FIG. 15 is a diagram showing an example of a graph representing the relationship between the average value of the significant evaluation index and the price for each survey target. FIG. 15A shows an example in which each airline is represented as a radar chart with three evaluation indexes and prices as axes. In addition, instead of the average value of the evaluation data in FIG. 14 and FIG. 15 (in this case, the original evaluation score that is not subjected to the evaluation data conversion), the values after the evaluation data conversion are ranked by processing such as step S3204 in FIG. Later averages may be used. For example, an average value of ranks in the table of FIG. 11 can be used. In the examples of FIGS. 14 and 15, the values of the three significant evaluation indices are all evaluation points, and the unit or sign of the values are indicated by the same value evaluation standard. Either of the average values of ranks may be used. However, when the value of the evaluation index is not indicated by the same value evaluation standard (for example, two evaluation points and one defect number), an average value rank is used.

However, if the evaluation data is an original evaluation point that is not subjected to evaluation data conversion, it is necessary to convert the evaluation data in ascending order because the evaluation is performed in descending order with respect to the original evaluation point. In descending order, 100 points are ranked first (ranked minimum value) and 0 points are ranked lowest (ranked maximum value), so the center side in the radar chart of FIG. Since it becomes a point, it converts in order to correct | amend this. Instead of converting the rank from descending to ascending order, the scale display of the radar chart may be changed from ascending to descending order.

For example, in the case of an evaluation index that has a ranking rule different from the evaluation score, such as the number of defects, the ranking is performed in ascending order with respect to the original number of defects that does not perform evaluation data conversion. Although the scale display of is the same as the ranking of the evaluation points, when displaying the value of the number of defects on the radar chart, the scale display is reversed from that of the evaluation points (that is, the same scale display as that of the ranking). For example, the scale display of the radar chart can be determined using a table in which rules for ranking for each evaluation index are determined in advance as shown in FIG.

FIG. 15 (b) shows an example in which the relationship between the average value of the evaluation points for each of the three evaluation indices for each airline and the price is shown in a scatter diagram. Here, the horizontal axis is the evaluation point average value by survey object and the vertical axis is the price, and the relationship between the average evaluation point value by survey object and the price ($) shown in FIG. 14 is shown. The average value by evaluation target for the evaluation score average value of each evaluation index that became significant can be interpreted as the total value by evaluation target of the evaluation index that became significant. This is considered to indicate a difference in value. Note that other statistical values such as a median value or a mode value may be used in place of the average value of the evaluation points (or the rank in which the evaluation points are ranked) and the average value of the evaluation points by survey object.

In general, when all the evaluation indicators that are significant are of the same type and the scale level is an interval scale or a proportional scale (for example, all are evaluation points), the average value and the average value of the evaluation indices for each survey target Is considered appropriate. The interval scale indicates, for example, a numerical value having no absolute zero point such as the year and the temperature in degrees Celsius, and the proportional scale indicates, for example, a numerical value having an absolute zero point, such as length or absolute temperature. In other cases, the median ranking and the median ranking of the evaluation index by survey object are appropriate. Therefore, the statistical calculation unit 13 determines whether or not the evaluation index that has become significant from the evaluation index numerical type is the same type by using a table as shown in FIG. 28 described later. May be determined whether or not is an interval or proportional, and a process of selecting a suitable statistical value may be performed.

Returning to FIG. 5, next, the value evaluation server 10 determines whether or not there are a plurality of significant evaluation indexes (S309). If there are no more than one, proceed to step S340. When there are a plurality of evaluation indexes that are significant, the statistical calculation unit 13 determines whether there is a statistically significant difference directly between the survey targets from the statistical values after the evaluation data conversion for the evaluation data of these evaluation indexes. A process for determining a significant difference between survey targets to be comprehensively determined is performed (S330).

16 and 17 are flowcharts showing an outline of an example of the flow of the significant difference determination process between the investigation targets in step S330 of FIG. First, the statistical calculation unit 13 calculates all of the survey target i (i is an integer from 1 to the number of survey targets k) and a significant evaluation index j (j is an integer from 1 to the number m of the significant evaluation indexes). It is determined whether or not the number k to be investigated is 2 for the combination (S3301). When the number of survey targets is 2, as will be described later, Wilcoxon signed rank test is performed. On the other hand, if the number of survey targets is not 2 (that is, the number of survey targets is 3 or more), Friedman test is performed.

[Friedman test]
First, the statistical value of the value after the evaluation data conversion in each survey target is ranked for each evaluation index (S3302). For example, when the evaluation index is an evaluation score, the average value of the scores after the evaluation data conversion is ranked. In addition, when the scale level of evaluation data is neither an interval scale nor a proportional scale, other statistical values such as a median value may be appropriately used instead of the average value. Next, it is determined whether or not there is the same rank in the evaluation index (S3303). Instead of the same rank determination process in step S3303, the same value determination process (step S3202 in FIG. 6 or S3219 in FIG. 7) performed in the Wilcoxon test or the Kruskal-Wallis test is performed before step S3302. Also good.

If there is no same rank in step S3303, it is determined whether or not the number k to be investigated is equal to or greater than a predetermined threshold (for example, 5) (S3304). When the number k to be investigated is less than 5, the FR amount is calculated by the following formula (S3305).

Here, i is an integer of 1 to survey the number k, R _i is the rank sum for each evaluation index in each study.

FIG. 18 is a diagram illustrating an example in which a significant difference between three survey targets is determined by statistical calculation from the value of a significant evaluation index. In the table of FIG. 18, three significant evaluation indices (“quality of AV equipment”, “attitude of CA”, “luxury of meal”) for the three airlines under investigation (airlines 1 to 3) The average value of the evaluation points after conversion of the evaluation data is shown. In addition, the rank calculated for the average value of the evaluation points after the evaluation data conversion for each survey target for each evaluation index, and information on the sum of ranks for each survey target (corresponding to R _i in Equation 6) It is written together. FIG. 18 shows an example where there are no survey targets of the same rank in the evaluation index. In this case, the number of survey targets k = 3 and the number m of significant evaluation indexes m = 3. From the formula 6, the FR amount = 2.

Next, the limit value fr (α) specified by the predetermined significance level α, the number k of survey targets, and the number m of significant evaluation indexes is specified from the Friedman test table held in advance by the statistical calculation unit 16 or the like. The obtained fr (α) is compared with the FR amount calculated in step S3305 (S3306). In the example of FIG. 18, the limit value fr (0.05) = 6 is obtained from the Friedman test table based on the significance level α = 0.05, the number of survey targets k = 3, and the number m of significant evaluation indexes. . Note that the Friedman test table includes the number of survey targets k, the number m of significant evaluation indexes, and the type that obtains a significant point (limit value) from the significance level α, the number of survey targets k, and the number of significant evaluation indexes. There is a type that obtains a significance probability (P value) from m and the FR amount, but the former test table is used for comparison here. The latter test table is used when calculating the P value in the selection of supplementary explanation items to be described later.

Next, based on the comparison result in step S3306, it is determined whether there is a significant difference between the survey targets (S3307), and the inter-survey target significant difference determination process is terminated. Here, it is determined that there is a significant difference when FR amount ≧ fr (α), and it is determined that there is no significant difference when FR amount <fr (α). In the example of FIG. 18, since (FR amount = 2) <{fr (0.05) = 6}, it is determined that there is no significant difference. That is, it is determined that there is no significant difference in value in the graph of FIG.

If there is no value to be compared in the Friedman test table because the number of survey targets k or the number m of significant evaluation indices is small, it cannot be said that there is a significant difference because there is not enough data. Make a decision. At this time, when the result is displayed by the processing described later, for example, a message such as “There is not enough data and it cannot be said that there is a significant difference as a result” may be added. Thereby, even when there is no value to be compared in the test table, a significant difference can be determined.

On the other hand, if the number k to be investigated is greater than or equal to a predetermined threshold (for example, 5) in step S3304 in FIG. 16, the FR amount is calculated by the same process as in step S3305 described above (S3308). FIG. 19 is a diagram illustrating an example in which a significant difference between five survey targets is determined by statistical calculation from the value of a significant evaluation index. In the table of FIG. 19, three significant evaluation indicators (“quality of AV equipment”, “attitude of CA”, “luxury of meal”) for the five airlines (airlines 1 to 5) to be surveyed The average value of the evaluation points after conversion of the evaluation data is shown. In addition, the rank calculated for the average value of the evaluation points after the evaluation data conversion in each survey target for each evaluation index, and information on the rank sum obtained by summing the ranks for each survey target are also shown.

FIG. 19 shows an example in which there are no survey targets of the same rank in the evaluation index. Since the number of survey targets k = 5 and the number of significant evaluation indexes m = 3, the above formula 6 Therefore, the FR amount = 9.6.

Next, a predetermined significance level α and a limit value χ ² (φ, α) given with a degree of freedom φ = k−1 are specified from the chi-square distribution table previously held by the statistical calculation unit 16 and the like. The obtained χ ² (φ, α) is compared with the FR amount calculated in step S3308 (S3309). In the example of FIG. 19, the limit value χ ² (4,0.05) = 9.49 is obtained from the chi-square distribution table from the significance level α = 0.05 and the degree of freedom φ = 5-1 = 4.

Next, based on the comparison result in step S3309, it is determined whether there is a significant difference between the survey targets (S3310), and the inter-survey target significant difference determination process is terminated. Here, it is determined that there is a significant difference when the FR amount ≧ χ ² (φ, α), and it is determined that there is no significant difference when the FR amount <χ ² (φ, α). . In the example of FIG. 19, since (FR amount = 9.6)> {χ ² (4,0.05) = 9.49}, it is determined that there is a significant difference. That is, it is determined that there is a significant difference in value in the graph of FIG.

In step S3303, when there is the same rank in the evaluation index, an average rank is assigned to the same rank by the same process as step S3214 in FIG. 6 (S3311). FIG. 20 is a diagram illustrating another example in which a significant difference between three survey targets is determined by statistical calculation from the value of a significant evaluation index. In the table of FIG. 20, three significant evaluation indicators (“quality of AV equipment”, “attitude of CA”, “luxury of meal”) for the three airlines under investigation (airlines 1 to 3) The average value of the evaluation points after conversion of the evaluation data is shown. In addition, rank information calculated for the average value of evaluation points after conversion of evaluation data in each survey target is written together for each evaluation index.

Here, in the table of FIG. 20, there are two survey targets with the ranking of “AV device quality” as the evaluation index (t ₁ = 2). When the average rank is calculated for this group using the formula used in step S3214 in FIG. 6, the average rank of this group is 1.5. The table in FIG. 20 also includes information on the average rank including the calculation result and rank sum obtained by summing up the average rank for each survey target.

Next, based on the average rank data calculated in step S3311, the FR * amount is calculated by the following equation. (S3312).

Here, FR is the FR value obtained by the equation (6), wherein the calculated information rank sum obtained by summing the average rank as R _i. Further, j is an integer from 1 to the number m of significant evaluation indexes, and i is an integer of the number e _j of different ranks in the average rank data of the 1st to j-th evaluation indexes. T _ij is the number of the i-th smallest rank among the different ranks in the average rank data of the j-th evaluation index. For example, when the three average ranks of the j-th evaluation index are 1.5, 1.5 and 3, e _j = 2, t _1j = 2 and t _2j = 1.

In the example of FIG. 20, FR * amount = 1.27 from Equation 7 above. Thereafter, the limit value χ ² (φ, α) is specified from the predetermined significance level α of the chi-square distribution table and the degree of freedom φ = k−1 by the same processing as in step S3309, and the obtained χ ² (Φ, α) is compared with the amount of FR * obtained in step S3312 (S3313).

Next, based on the comparison result in step S3313, it is determined whether there is a significant difference between the survey targets (S3314), and the inter-survey target significant difference determination process is terminated. Here, it is determined that there is a significant difference when FR * amount ≧ χ ² (φ, α), and it cannot be said that there is a significant difference when FR * amount <χ ² (φ, α). judge. In the example of FIG. 20, in the two-sided test with a significance level of 5%, (FR * amount = 1.27) <{χ ² (2,0.05) = 5.99} Judge that you can not say. That is, it is determined that there is no significant difference in value in the graph of FIG.

It is sufficient if the product of the number of survey targets k and the number m of significant evaluation indices in the Friedman test with the same rank is less than a predetermined threshold (for example, less than 30 which is generally considered to be inferior in accuracy of statistical calculations). It is not necessary to perform statistical calculation because it is impossible to secure a high accuracy. In this case, it may be determined that there is no significant difference because there is not enough data. At this time, when the result is displayed by the processing described later, for example, a message such as “There is not enough data and it cannot be said that there is a significant difference as a result” may be added. Further, after executing the statistical calculation, for example, a message such as “the accuracy of the result of the statistical calculation is not high” may be added.

In addition, when the survey target becomes significant by the Friedman test, a statistical significance difference between two survey targets (for example, airlines A and B) among the survey targets that become significant may be determined. Such a determination is also performed on the other two combinations in the survey target that become significant, and for example, a significant difference between all combinations of the two airlines is determined. The data used at this time is the data used in the Friedman test, and the Wilcoxon signed rank test described later is performed. Based on this result, it is possible to determine a significant difference between two specific survey targets among the survey targets determined to have a significant difference by the Friedman test. Can be provided.

[Wilcoxon signed rank test]
If the number of survey targets is less than 3 (ie, the number of survey targets is 2) in step S3301 in FIG. 16, the process proceeds to FIG. 17 to perform Wilcoxon signed rank test. For each combination, a difference between statistical values such as an average value after conversion of evaluation data and an absolute value thereof are calculated (S3315). Next, the rank is calculated in ascending order of the absolute value of the difference calculated in step S3315 (S3316). If zero is included in the difference value calculated in step S3315, no ranking is assigned to the data (data with zero difference). Next, it is determined whether or not the calculated ranks have the same rank (S3317). If there is no order, N is calculated for each combination to be investigated (S3318). Here, the N number is the number of non-zero values among the absolute values of the difference between the statistical values after the evaluation data conversion between the survey targets calculated for each evaluation index in step S3315.

FIG. 21 is a diagram showing an example in which a significant difference between two survey targets is determined by statistical calculation from the value of a significant evaluation index. In the table of FIG. 21, the average value of the evaluation points after conversion of the evaluation data of the seven evaluation indexes for the two airlines (airlines 1 and 2) to be investigated is shown. In addition, the absolute value of the difference between the average values of the evaluation points after the evaluation data conversion between the two airlines calculated in step S3315 and the positive / negative information of the difference are also shown. In addition, information on the rank calculated in step S3316 for the absolute value of the difference is also shown. FIG. 21 shows an example where there is no evaluation index of the same rank for the absolute value of the difference. Since the absolute value of the difference is zero in one of the seven evaluation indices, N number = 6. Become.

Next, it is determined whether or not the N number is less than a predetermined threshold (for example, 25) (S3319). In the example of FIG. 21, since the N number = 6 is less than the predetermined threshold (25), the WS amount is calculated (S3320). Here, first, the WS + amount and the WS− amount are respectively calculated based on the rank order of the difference values and the positive / negative information of the differences obtained in steps S3315 and S3316 of FIG. Here, the WS + amount is a sum of ranks obtained by summing the ranks of the evaluation indexes calculated in steps S3315 and S3316, and the difference between the average values of the evaluation points is positive. The WS− amount is the average value of the evaluation points. This is the sum of ranks obtained by summing up the ranks of evaluation indexes having negative differences. In the example of FIG. 21, WS + amount = 13 and WS−amount = 8. Next, the WS + amount and the WS− amount are compared, and the smaller one is set as the WS amount. If the values are equal, either one is taken as the WS amount.

Next, the significance point t _L (α / 2) given by a predetermined significance level α and N number was specified from the Wilcoxon signed rank test table held in advance by the statistical calculation unit 16 or the like, and obtained. The value of t _U (α / 2) calculated from t _L (α / 2) and t _L (α / 2) is compared with the WS amount calculated in step S3320 (S3321). Here, t _U (P) is a value obtained by the equation N (N + 1) / 2−t _L (P), P is a significance level assigned to the limit value, and in the example of FIG. Equivalent to. Since α / 2 = 0.025 is used when a significant difference is determined by a two-sided test with a significance level of 5%, t _L (0.025) = 0 is calculated from the Wilcoxon signed rank test table in the example of FIG. As a result, t _U (0.025) = 21 from the above formula.

There are two types of Wilcoxon signed rank test tables: a type for obtaining a significant point (limit value) from N number and significance level α, and a type for obtaining a significance probability (P value) from N number and WS amount. For comparison, the former test table is used. The latter test table is used when calculating the P value in the selection of supplementary explanation items to be described later.

Next, based on the comparison result in step S3321, it is determined whether there is a significant difference between the survey targets (S3322), and the statistical processing is terminated. Here, when t _L (α / 2) <WS amount <t _U (α / 2), it is determined that there is no significant difference, and WS amount ≦ t _L (α / 2) or t _U When (α / 2) ≦ WS amount, it is determined that there is a significant difference. In the example of FIG. 21, since {t _L (0.025) = 0} <(WS amount = 8) <{t _U (0.025) = 21}, it is determined that there is no significant difference. To do. That is, it is determined that there is no significant difference in value in the graph of FIG.

If there is no value to be compared in Wilcoxon's signed rank test table for reasons such as the N number being small, it is determined that there is no significant difference because there is not enough data. At this time, when the result is displayed by the processing described later, for example, a message such as “There is not enough data and it cannot be said that there is a significant difference as a result” may be added. Thereby, even when there is no value to be compared in the test table, a significant difference can be determined.

In step S3319 in FIG. 17, when the number N of absolute values of the differences is not zero, the WS amount is calculated (S3323). Here, unlike step S3320, the WS + amount is the WS amount. Further, the u ₀ amount is calculated by the following equation (S3324).

You may use the following formula | equation which performed the continuous correction which improves a statistical precision.

FIG. 22 is a diagram illustrating another example in which a significant difference between two survey targets is determined by statistical calculation from the value of a significant evaluation index. In the table of FIG. 22, the average value of the evaluation points after the evaluation data conversion of 25 evaluation indexes is shown for the two airlines (airlines 1 and 2) to be surveyed, as in FIG. Yes. In addition, the absolute value of the difference between the average values of the evaluation points after the evaluation data conversion between the two airlines calculated in step S3315 and the positive / negative information of the difference are also shown. In addition, information on the rank calculated in step S3316 for the absolute value of the difference is also shown.

In FIG. 22, there is no evaluation index of the same rank with respect to the absolute value of the difference, and the number N of evaluation indexes whose absolute value of the difference is not zero is 25, which is equal to or greater than the threshold (25). . At this time, since WS + amount = 237, WS amount = 237, and u ₀ amount = 2.00 from the above equation (8).

After that, in FIG. 17, the limit value u (α) is specified from the predetermined significance level α of the normal distribution table held in advance by the statistical calculation unit 16 or the like, and the obtained u (α) is obtained in step S3324. The absolute value of the obtained u ₀ quantity is compared (S3325). When a significant difference is determined by a two-sided test with a significance level of 5%, a limit value u (0.05) = 1.96 is obtained from the normal distribution table.

Next, based on the comparison result in step S3325, it is determined whether there is a significant difference between the survey targets (S3326), and the inter-survey target significant difference determination process is terminated. Here, it is determined that there is a significant difference when the absolute value of the u ₀ quantity is not less than the limit value u (α), and it is determined that there is no significant difference when the absolute value is less than the limit value u (α). To do. In the example of FIG. 22, {| u ₀ | = 2.00}> {u (0.05) = 1.96}, so it is determined that there is a significant difference. That is, it is determined that there is a significant difference in value in the graph of FIG.

In step S3317 in FIG. 17, if there is an evaluation index having the same rank with respect to the absolute value of the difference, an average rank is assigned to the same rank in the same manner as in step S3214 in FIG. 6 (S3327). FIG. 23 is a diagram illustrating another example in which a significant difference between two survey targets is determined by statistical calculation from the value of a significant evaluation index. In the table of FIG. 23, as in FIG. 21, the average value of the evaluation points after the evaluation data conversion of the seven evaluation indexes is shown for the two airlines (airlines 1 and 2) to be investigated. . In addition, the absolute value of the difference between the average values of the evaluation points after the evaluation data conversion between the two airlines calculated in step S3315 and the positive / negative information of the difference are also shown. In addition, information on the rank calculated in step S2316 for the absolute value of the difference is also shown.

Here, in the table of FIG. 23, there are two evaluation indexes with the second rank, and there are two evaluation indexes with the fifth rank (t ₁ = 2 and t ₂ = 2). When the average rank is calculated for these two groups by the formula used in step S3214 in FIG. 6, the average rank of the group ranked second is 2.5, and the average rank of the group ranked fifth is 5 .5. In the table of FIG. 23, information on the average rank including these calculation results is also shown.

Next, based on the data of the average rank of the absolute value of the difference calculated in step S3327 and the positive / negative information of the difference, the WS + amount is calculated as the WS amount by the same method as in step S3323. For distinction, these are described as WS + * amount and WS * amount, respectively (S3328). In the example of FIG. 23, since WS + * amount = 20.5, WS * amount = 20.5. Further, the u ₀ * amount is calculated by the following equation. (S3329).

Here, j is an integer from 1 to the number g of evaluation index groups having the same rank.

In the example of FIG. 23, WS * amount = 20.5, the total number N of evaluation indexes whose difference is not zero is 7, and u ₀ * amount = 1.10 from Equation 10. Thereafter, the limit value u (α) is specified from the predetermined significance level α of the normal distribution table by the same processing as in step S3325 described above, and u (α) obtained is obtained and u ₀ * obtained in step S3329. The absolute value of the quantity is compared (S3330).

Next, based on the comparison result in step S3330, it is determined whether or not there is a significant difference between survey targets (S3331), and the inter-survey target significant difference determination process is terminated. Here, it is determined that there is a significant difference when the absolute value of the u ₀ * quantity is equal to or greater than the limit value u (α), and it cannot be said that there is a significant difference when the absolute value is less than the limit value u (α). judge. In the example of FIG. 23, in a two-sided test with a significance level of 5%, {| u ₀ * | = 1.10} <{u (0.05) = 1.96}, it can be said that there is a significant difference. Judge that there is no. That is, it is determined that there is no significant difference in value in the graph of FIG.

When the value of the total number N of evaluation indexes whose difference is not zero in Wilcoxon signed rank test with the same rank is less than a predetermined threshold (for example, less than 25, which is generally considered to be inferior in statistical calculation accuracy). Since sufficient accuracy cannot be secured, statistical calculation may not be performed. In this case, it may be determined that there is no significant difference because there is not enough data. At this time, when the result is displayed by the processing described later, for example, a message such as “There is not enough data and it cannot be said that there is a significant difference as a result” may be added. Further, after executing the statistical calculation, for example, a message such as “the accuracy of the result of the statistical calculation is not high” may be added.

With the above processing, the inter-survey target significant difference determination processing in step S330 of FIG. 5 is completed. In step S330, statistical calculation for comprehensively determining whether or not there is a significant difference between survey targets using the statistical value after the evaluation data conversion of the evaluation index determined to have a significant difference in step S320. The statistical calculation is performed based on the data including the statistical value after the evaluation data conversion of the evaluation index that cannot be said to have a significant difference, and the overall significant difference between the survey targets is determined. It may be. Alternatively, the difference in value between the survey targets may be evaluated based only on information about the evaluation index determined to have a significant difference without performing the processing.

5, next, the evaluation result output unit 14 of the value evaluation server 10 specifies the statistical result including the graph to be provided to the user and the comment information attached thereto, and the product obtained in step S206 of FIG. A result output process is performed by attaching to the content information (S340). FIG. 24 is a flowchart showing an outline of an example of the flow of the result output process in step S340 of FIG. The evaluation result output unit 14 first determines the statistical results for each evaluation index in step S320 of FIG. 5 and the table or graph created in step S308, the result of the significant difference determination between the survey targets in step S330, and the sufficient Since there is no significant number of data, information on the result of the significant difference determination without statistical calculation is specified (S3401).

Next, based on the information specified in step S3401, it is determined whether there is a significant evaluation index among the evaluation indices (S3402). Here, the evaluation index that becomes significant is, for example, an evaluation index that is determined to have a significant difference in the statistical processing in step S320 of FIG. There is a significant difference in the number of evaluation data, etc., such as an evaluation index (for example, “the accuracy of the result of this statistical calculation is not high”) It is also possible to use a non-added one).

If there is no significant evaluation index in step S3402, a comment that recommends selecting a product or the like only by price (that is, selecting a product or the like with a low price) is selected (S3409). As a comment, for example, “We recommend that you select a cheap product that is not statistically different in value. (If the number of products and services to be compared is 1, "We recommend the product.)"

In step S3402, if there is a significant evaluation index, it is determined whether the number is 1 (S3403). If there is one significant evaluation index, it is determined whether the user attaches importance to the significant evaluation index (S3404). For this, for example, a screen for inquiring the user via the user terminal 30 may be output and an answer may be input, or in advance when selecting or inputting a product or the like in step S203 of FIG. You may make it designate the evaluation index to attach importance.

In step S3404, if the user attaches importance to one evaluation index that is significant, the user selects a comment that recommends selecting a product with a low price and a high evaluation of the significant evaluation index (S3410). . Specifically, for example, in the graph created in step S308 of FIG. 5 and showing the relationship between the price and the value of the product as shown in FIG. 15B, the product plotted in the lower right area It is recommended to select etc. As a comment, for example, “Since it can be said that there is a statistically significant difference between the target products and services, the evaluation index that the customer emphasizes is highly evaluated and the price in the lower right area of the graph is high. It is recommended to select a service. "

If it is determined in step S3404 that the user does not place importance on one evaluation index that has become significant, it is further determined whether or not an evaluation index other than the evaluation index is emphasized (S3405). Here, similarly to step S3404, for example, a screen for inquiring the user via the user terminal 30 may be output and an answer may be input, or a product or the like is selected / input in step S203 of FIG. It is also possible to specify an evaluation index to be prioritized when performing the process.

If it is determined in step S3405 that the user attaches importance to an evaluation index that is not significant, the process proceeds to step S3409 described above to select a comment that recommends a product with a low price. Alternatively, it may be recommended that the evaluation and price of an insignificant evaluation index emphasized by the user are shown as reference information and a product or the like is selected at the user's discretion. An example of a comment in this case is, for example, “If the evaluation index that the customer places importance on is statistically not significantly different between the target products / services, As long as it is used as a reference indicator in the selection of products and services, "

If it is determined in step S3405 that the user does not attach importance to any evaluation index that is not significant, the process proceeds to step S3410 described above, and a product with a low price and a high evaluation of a significant evaluation index (see FIG. 15B). In the graph shown, select a comment that recommends that you select a product or the like plotted in the lower right area. As a comment in this case, for example, “Since an evaluation index having a statistically significant difference was found between the target products / services, a product / service in the lower right area of the graph that has a high evaluation and a low price is selected. It is recommended that the message "

If there are a plurality of significant evaluation indexes in step S3403, it is determined whether the user places importance on any of the evaluation indexes that become significant as in step S3404 (S3406). Here, similarly to step S3404, for example, a screen for inquiring the user via the user terminal 30 may be output and an answer may be input, or a product or the like is selected / input in step S203 of FIG. It is also possible to specify an evaluation index to be prioritized when performing the process.

In step S3406, if any of the evaluation indexes that the user has made significant is valued, a product with a low price and a high evaluation (evaluation point average value is high) for the evaluation index emphasized by the user is selected. A comment recommended to be selected is selected (S3411). Specifically, for example, in the radar chart centered on the price and each evaluation index as shown in FIG. 15A created in step S308 of FIG. It is recommended to select products with high average values and low prices. As a comment, for example, “Since it can be said that there is a statistically significant difference in the evaluation index that the customer emphasizes, select a product / service with a high average evaluation score and a low price on the radar chart. It ’s recommended that you do this. ”

If it is determined in step S3406 that neither of the evaluation indexes that have become significant is emphasized by the user, it is further determined whether or not the evaluation indexes other than the evaluation index that has become significant are emphasized (S3407). Here, similarly to step S3404, for example, a screen for inquiring the user via the user terminal 30 may be output and an answer may be input, or a product or the like is selected / input in step S203 of FIG. It is also possible to specify an evaluation index to be prioritized when performing the process.

If it is determined in step S3407 that the user attaches importance to an evaluation index that is not significant, the process proceeds to step S3409 described above, and a comment that recommends that a product or the like with a low price is selected. Alternatively, it may be recommended that the evaluation and price of an insignificant evaluation index emphasized by the user are shown as reference information and a product or the like is selected at the user's discretion. An example of a comment in this case is, for example, “If the evaluation index that the customer places importance on is statistically not significantly different between the target products / services, As long as it is used as a reference indicator in the selection of products and services, "

If it is determined in step S3407 that the user does not place importance on any evaluation index that is not significant, the result of the comprehensive significant difference determination between the survey targets in step S330 in FIG. It is determined whether or not there is a significant difference (S3408). If there is a significant difference overall, the process proceeds to step S3410 described above, and a comment that recommends that a product or the like with a low price and a high evaluation of a significant evaluation index is selected. That is, for example, it is recommended to select a product or the like plotted in the lower right area in the graph shown in FIG. As a comment in this case, for example, “Since an evaluation index having a statistically significant difference was found between the target products / services, a product / service in the lower right area of the graph that has a high evaluation and a low price is selected. It is recommended that the message "

If it is determined in step S3408 that there is no comprehensive difference between the survey targets, the process proceeds to step S3409, and a comment that recommends a product with a low price is selected. When a comment to be presented to the user is selected by the series of processes described above, the statistical result information including the graph as shown in FIG. 15 and the selected comment are collected in a predetermined format, and the result Information is transmitted and output to the user terminal 30 together with information such as the product acquired in step S206 of FIG. 3 (S3412). By outputting the result information output from the value evaluation server 10 by screen display or voice by the user terminal 30, the user can refer to and check the result of the value evaluation of the product etc. (step of FIG. 5) S310).

In the result output processing flow of FIG. 24 described above, the evaluation result output unit 14 of the value evaluation server 10 inquires about the evaluation index information to be emphasized to the user via the user terminal 30 and comments as necessary. Although selected, a program for executing a result output process and selecting a comment may be transmitted as a client program to the user terminal 30 and processed locally on the user terminal 30 side.

Also, instead of outputting the comment selected by the result output process or the program for executing the result output process and selecting a comment to the user terminal 30, for example, the determination contents of the result output process are expressed. The table may be transmitted to the user terminal 30 together with the statistical result. FIG. 25 is a diagram illustrating an example of a table expressing the determination contents of the result output process. Users can grasp the recommended content as selection criteria for products, etc., by referring to a combination of such judgment patterns in a tabular format and statistical results. It is possible to select a product or the like more accurately after understanding the overall image of the selection criteria.

Further, in either case of the result output process of FIG. 24 or the case of presenting the table of FIG. 25 to the user, a comment for recommending relative value judgment to the user may be added. An example of such a comment is, “Value evaluation data is inherently non-parametric data, and thus may not accurately represent a difference in values. It is recommended to make a reasonable decision ”.

As described above, when the user determines a product to be purchased or the like, the user refers to and confirms the result output by the value evaluation server 10 by the above-described series of processing, so that the statistical data existing between the survey targets can be obtained. The evaluation index having a significant difference, the average value of the evaluation points, and the price can be grasped at the same time, and the relationship between the value of the product etc. and the price can be recognized more accurately. At the same time, other evaluation indexes having no significant difference can be excluded from the value determination indexes of products and the like, and the value of products and the like can be grasped more clearly and simply.

For example, in the example of the radar chart shown in FIG. 15A, there are three significant evaluation indexes: the quality of AV equipment, the attitude of CA, and the luxury of meals. Can be understood as a value index that affects Therefore, for example, a user who places emphasis on the luxury of a meal selects a product that is worth considering by the user by selecting a product with a high evaluation score average value for the luxury of the meal if the price is about the same. it can. In addition, users who place emphasis on indicators other than these three value indicators (for example, seating comfort) are not significant, and therefore between airlines 1 to 3 Then, it can be considered that there is no big difference, and it becomes possible to select based on the price alone.

In addition, even for users who do not have a particular evaluation index to be emphasized, or users who place importance on both a significant evaluation index and a non-significant evaluation index, the average score of the three evaluation indices is calculated for each airline. From the graph as shown in FIG. 15B in which the numerical value obtained by simple averaging and the price are compared, the relationship between the value of the product and the price can be grasped. In the graph of FIG. 15B, it can be seen that the products and the like in the high-value and low-price area (the lower right area of the graph) are generally bargain products.

In the graph of FIG. 15 (b), the horizontal axis is a value obtained by simply averaging the evaluation score average values of the indicators that became significant for each survey target, and it cannot be said that the accuracy is high as a value indicator. Therefore, in the significant difference determination process between the survey targets in step S330 of FIG. 5, there is a statistically significant difference directly between the survey targets from the average value of the evaluation data converted of the plurality of significant evaluation indexes. Whether or not is comprehensively determined. If it cannot be said that there is a significant difference overall, it is recommended not to use the graph but to select products by price alone, and if there is a significant difference, it is recommended to select products by referring to the graph.

In addition, the significant difference determination process between investigation object of step S330 is performed when there are a plurality of evaluation indexes that are significant by the determination in step S309, and when there is one significant index, step S320 is performed. This is not done because it can be substituted with the result of significant difference determination in the statistical processing. However, it may be carried out when the significant difference between the survey targets is comprehensively determined by performing the significant difference determination process between the survey targets in step S330 including the evaluation index that is not significant in step S320.

[Processing flow (evaluation history information recording process)]
FIG. 26 is a flowchart showing an outline of an example of the flow of the evaluation history information recording process (step S04 in FIG. 2) in the value evaluation server 10. In advance in step S310 of FIG. 5, the user outputs information indicating the relationship between the price difference between the value of the product such as the graph and the price, the statistical result, the product information, and the like, The recommended method for selecting a product or the like is referred to or confirmed via the user terminal 30 to determine a product or the like to be purchased.

After that, the user uses the user terminal 30 to make a payment for the product, etc., for which purchase has been decided by a predetermined method such as Internet banking, credit card, electronic money, etc. The user terminal 30 inputs necessary information such as information related to the payment such as the option made, and transmits the purchase payment information related to the determined product to the value evaluation server 10 (S401).

Upon receiving the purchase settlement information, the value evaluation server 10 transfers a copy of the purchase settlement information to the product providing system 20 of the sales destination of the target product or the like (S402). The product etc. providing system 20 that has received the purchase settlement information performs sales and settlement processing on the target product etc. based on the contents of the received purchase settlement information, and transmits the processing result including the sales settlement information to the value evaluation server 10. (S403). The sales settlement information includes, for example, the contents of the target product, seller information, sales amount, purchased user information, and the like.

Upon receiving the sales settlement information, the value evaluation server 10 confirms that the target product etc., price, and other sales conditions are the same for the sales settlement information and the purchase settlement information received in step S402. Further, the questionnaire processing unit 15 specifies and extracts information on a questionnaire for inputting an evaluation for the target product or the like from the questionnaire DB 105 (S404).

It should be noted that the contents of the questionnaire may include, for example, a plurality of uniform question items corresponding to any product. However, in this case, all respondents may not be able to fill in an appropriate answer to a special question focused on individual products, or no answer may be entered. If these are used for statistical calculations, accuracy will be reduced. Result. Further, in the determination process of the lower limit value in step S303 in FIG. 5 (and step S6201 in FIG. 32 in the second embodiment to be described later), value index data less than the lower limit value is increased, which increases the calculation load of the computer and the calculation speed. To slow down. Therefore, it is desirable that the contents of the questionnaire include a plurality of suitable question items in association with the type of product or the like.

Thereafter, the questionnaire processing unit 15 transmits the extracted questionnaire information and the sales settlement information received from the product providing system 20 to the user terminal 30 (S405). The information to be transmitted may include help information related to a questionnaire input method and the like. The user terminal 30 receives the information and outputs it to the user, so that the user inputs, uses, uses, or consumes the purchased product, etc., and then inputs an evaluation of the product, etc. into the questionnaire can do. The input time of the questionnaire may be before or during use, use, consumption, etc., as long as it is a timing at which the purchased products can be evaluated appropriately.

The inputted questionnaire information is transmitted from the user terminal 30 to the value evaluation server 10 (S406). Upon receiving the questionnaire input result, the value evaluation server 10 extracts the contents of the questionnaire input by the questionnaire processing unit 15 and records it in the evaluation history DB 104 in association with the product etc. as evaluation information for the target product etc. (S407). The questionnaire processing unit 15 transmits the questionnaire again to the user terminal 30 when the questionnaire input result cannot be received even after a predetermined period of time has elapsed since the transmission of the questionnaire in step S405. You may ask for a reminder.

It should be noted that various methods can be used for exchanging questionnaire information between the value evaluation server 10 and the user (user terminal 30). For example, as in the processing flow described above, a questionnaire file in which the contents of a question and an answer column for evaluating each evaluation index are described is transmitted from the value evaluation server 10 to the user terminal 30, and the user On the other hand, after inputting an evaluation, the file can be configured to be returned to the value evaluation server 10.

In addition, in response to a questionnaire input request from the user via the user terminal 30, an HTML that displays the contents of the question on a web browser (not shown) on the user terminal 30 and receives input answer data The questionnaire information may be configured as a file or the like. Further, a configuration may be adopted in which a questionnaire is conducted by telephone, FAX, mail, etc., and an answer content is input to the evaluation history DB 104 of the value evaluation server 10 by an operator or the like.

FIG. 27 is a diagram showing an example of the contents of a questionnaire presented to the user. Here, an example is shown in which a questionnaire input screen is displayed on the Web browser on the user terminal 30 and the response content input by the user is acquired. As a method for inputting evaluation data, etc., for example, a method of directly inputting numerical values such as evaluation points and satisfaction within a predetermined range (such as “0 to 100”), or a suitable expression (representing the degree and state) Method of selecting an evaluation by selecting a radio button corresponding to a phrase, numerical range, etc., a method of directly inputting a count value such as the number of defects or a probability, a method of directly inputting a measurement value such as time or length, etc. Is included.

For example, when selecting and evaluating a suitable expression or phrase using a radio button or the like, numerical values associated with the selected phrase or the like (in the example of FIG. 27, “very good” = 0, “good”). = 2.5) is used for the statistical calculation in step S320 of FIG. Here, even if the numerical values are the same, the meanings of the numerical values are different for each evaluation index, for example, 80 evaluation points (points) and 80 measurement values (minutes, etc.). Therefore, for example, when analyzing statistical calculation or the like in the processing of steps S320 and S330 in FIG. 5, evaluation data conversion is performed to convert the evaluation index into a value having a unified magnitude relationship with respect to value. Further, when calculating the rank, a unified rank is calculated for the value after the evaluation data conversion.

In this embodiment, for example, ascending order is used in which the order is given in ascending order of values. For example, when displaying the evaluation data to the user or operator in the table or graph of FIG. 14 or FIG. 15, the original evaluation data (in the present embodiment, by performing inverse conversion of the evaluation data conversion, etc.) You may use the original evaluation score which does not perform evaluation data conversion. As a result, the user or operator can more clearly understand the difference in value.

FIG. 28 is a diagram showing an example of a table in which evaluation data conversion rules and ranking rules are defined for each type of numerical value of the evaluation index. Here, as the evaluation data conversion rule, a method of converting the evaluation data in association with each type of evaluation index is defined. For example, since the numerical value originally inputted to the questionnaire takes a value of 0 to 100, for example, if the conversion formula of “100− (original evaluation score value)” is used as the evaluation data conversion rule, the original evaluation score When the point is 80, the converted value is 20. When the evaluation data conversion rule is “10 times”, the original value is multiplied by 10. In the case of “no conversion”, the original values are ranked as they are without conversion. In the case of “sign inversion”, the original value is inverted by multiplying the sign of the converted value by (−1) or the like.

Also, the ranking rules are defined in association with each type of evaluation index. The ascending order is a rule in which the smallest value among the numerical values entered in the questionnaire is ranked first, and the descending order is a rule in which the largest value among the numerical values entered in the questionnaire is ranked first. The value evaluation server 10 holds the information of such a table in the form of a file or the like in advance, so that the statistical calculation unit 13 is based on, for example, a unified ranking rule (ascending order in the present embodiment). In addition, it is possible to identify the value expression method used for each type of original evaluation index that is not subjected to evaluation data conversion, and to determine the scale display of the graph of FIG. For example, in the radar chart of FIG. 15A, a scale display is selected in which the high value of the original evaluation score is on the outside and the low value is on the center side.

In the example of FIG. 28, a rule is defined for each evaluation index numerical type, but there are also cases where the same type has different rules. For example, some measured values are not necessarily in ascending order. For example, the waiting time (minutes) from the completion of boarding at the departure airport to the actual takeoff is higher in evaluation value as the value is smaller and is in ascending order. Even with the same count value, the pass rate is in descending order, but the defect rate is in ascending order. Therefore, rules are defined for each evaluation index.

<Embodiment 2>
In the value evaluation support system according to the second embodiment of the present invention, for example, buying and selling of a product is performed between a seller and a user via a purchase agent service using a computer system as an example. The retailer that sells the product that the user wants to purchase and the sales agent evaluate the difference in value, and based on this, the retailer and the sales agent that purchase the product to the user The criteria for selecting a person are recommended.

For example, a product sold at a retail store such as a supermarket is purchased and settled by a user at home via the Internet, and the product is actually received at the store or delivered to the user by a sales agent. Case. In this case, the value of goods and services is influenced by retail stores and sales agents. Therefore, in the present embodiment, the business operators involved in the sales of these products are targeted for investigation. In addition, as an evaluation index indicating value, satisfaction with retailers and sales agents, i.e. satisfaction with all products sold by retailers and sales agents or all services performed, is an example. Instead, a more specific evaluation index such as the freshness of the product may be used. Further, instead of a retail store and a sales agent, the present invention can also be applied to other combinations such as a manufacturer and a retail store, or a retail store and its sales products.

In the present embodiment, two groups of a survey target group A (for example,

retail store

1, 2,...) And a survey target group B (for example,

sales agent

1, 2,...) Each having a plurality of survey targets. In an environment in which there is a survey, one evaluation index is compared for each combination of survey targets between survey target groups, and it is determined whether there is a statistically significant difference between the survey targets. In an environment in which three or more survey target groups exist, for example, two survey target groups selected based on conditions specified by the user can be targeted. In order to determine a statistically significant difference, the Friedman test described above is used.

In the first embodiment described above, statistical calculation is performed in two steps. That is, as a first step, in S320 of FIG. 5, a significant difference between the survey targets is tested for each evaluation index, and as a second step, evaluation of the evaluation index that becomes significant in step S330 of FIG. Based on the point average value, the overall significant difference between the subjects was determined. On the other hand, in this Embodiment, the 1st step is not performed but the comprehensive significant difference between investigation object is determined in a 2nd step.

[System configuration]
FIG. 29 is a diagram showing an outline of a configuration example of the value evaluation support system according to the second embodiment of the present invention. The value evaluation support system 1 basically has the same configuration as the system configuration of the first embodiment shown in FIG.

In the present embodiment, the seller having the product providing system 20 (20a, b in the example of FIG. 29) and the product content DB 201 (201a, b in the example of FIG. 29) is a retail store such as a supermarket. Apart from these, the sales agent has a plurality of sales agent systems 21 (21a, b in the example of FIG. 29), which is an information processing system for providing a product sales agent service on behalf of the supermarket. Yes. Each sales agent system 21 stores information for performing sales agent in the agent content DB 211 (211a, b in the example of FIG. 29).

The value evaluation server 10 basically has the same functional blocks as the functional blocks of the first embodiment shown in FIG. The seller DB 101 is a table that holds information related to the seller (supermarket) and the merchandise providing system 20 of the seller. For example, the seller DB, the name of the person in charge, the address, the telephone number, the FAX number, This includes personal information such as an e-mail address, user ID, and password, and sales information such as payment terms.

The user DB 102 is a table that holds information related to users. For example, name, age, date of birth, address, gender, telephone number, FAX number, e-mail address, ID used in the value evaluation support system 1 And personal information such as account information such as passwords. In addition, a list of retailers and sales agents that the user wants to request for quotations, a list of priorities for requesting sellers / representatives with a priority order, a request for quotations schedule that indicates the user's plan to request a quotation, and usage Purchased product history that shows the actual purchase history of the user, list of requested products that the user wants to request a quote for, quote request service list that lists the agency services that the user wants to request for a quote, and value for the user A comparative value evaluation index used for comparison, purchase information such as a payment method, and the like are included.

The product etc. DB 103 is a table that holds information related to the product etc. extracted from each product etc. providing system 20, for example, product name, grade, number of pieces, contents, production area, product image, description, etc. Information etc. are included. Regarding the contents of the evaluation history DB 104 and the questionnaire DB 105, information related to the sales agent is added to the contents of the evaluation history DB 104 and the questionnaire DB 105 in the first embodiment.

In the present embodiment, the value evaluation server 10 further includes an agent DB 106. The agent DB is a table that holds information related to the sales agent and the sales agent system 21 of the sales agent. For example, the sales agent name, address, telephone number, person in charge name, telephone number, FAX number, Personal information such as e-mail address, account information such as ID and password used in the value evaluation support system 1, contents and conditions of agency services, service information such as range, time zone, handling amount, sales information such as payment conditions, etc. Is included.

[Processing flow (main processing)]
FIG. 30 is a flowchart showing an overview of an example of the flow of main processing in the value evaluation server 10. First, an initial registration process for registering information of a seller, a sales agent, and a user is performed (S05). Here, for example, an administrator, an operator, or the like may initially register these pieces of information in each DB of the value evaluation server 10, or a seller who does not have a user ID used in the value evaluation support system 1 or a sales agent. Initial registration may be performed when there is an access from a user or a user. At this time, for example, from the seller, the sales agent or the user who performs registration, the seller information including personal information other than the user ID, the registration information input to the DB of the sales agent information or the user information is received, After these information is formally registered in the DB, a user ID is issued and a password is registered.

Next, in accordance with the quotation request schedule registered by the user in the user DB 102, the quotation request seller / agent priority order list, and the quotation request commodity list, information including the price related to the target product for the quotation request, It is obtained by inquiring the product providing system 20 or each sales agent system 21 and obtaining statistical information by extracting evaluation history information about the target investigation object (seller and sales agent) from the evaluation history DB 104. A price / value survey process is performed to survey (S06). In addition, we conduct a survey on the value of purchased products, etc., sellers, and sales agents for users who select and purchase products, sellers, and sales agents, and record information related to the results The evaluation history information recording process is performed (S07), and the series of processes is terminated.

[Processing flow (price / value survey processing)]
FIG. 31 is a flowchart showing an overview of an example of the flow of price / value survey processing (step S06 in FIG. 30) in the value evaluation server 10. First, the value evaluation server 10 uses the product etc. information acquisition unit 11 according to the estimate request schedule for each user by periodically referring to the information of the estimate request schedule registered in the user DB 102 by each user. The estimate request process is automatically started (S601). Instead of automatic activation based on the quotation request schedule, the user may manually request a quotation request from the value evaluation server 10 via the user terminal 30, and may be activated as a trigger.

The product etc. information acquiring unit 11 specifies the seller and the sales agent registered in the quotation request seller / representative priority list registered in the user DB 102 by the target user as the survey target (S602). ). Next, the evaluation history information acquisition unit 12 extracts evaluation data related to the evaluation index associated with the investigation target specified in step S602 from the evaluation history DB 104 (S603). Here, when there are a plurality of questions in the questionnaire for the user, which will be described later, since there are a plurality of evaluation indexes, it is necessary to set in advance which evaluation index to extract the evaluation data. This is performed, for example, by the user registering information of the comparative value evaluation index used for value comparison in the user DB 102 in the initial registration process in step S05 of FIG. In the present embodiment, it is assumed that the user has registered satisfaction (0 to 100 points) in the user DB 102 as a comparative value evaluation index.

Next, the statistical calculation unit 13 performs matrix creation processing for creating a matrix table of the evaluation data extracted in step S603 for the two types of survey targets (seller and sales agent) identified in step S602 (S620). FIG. 32 is a flowchart showing an overview of an example of the flow of matrix creation processing in step S620 of FIG. When the matrix creation process is started, the statistical calculation unit 13 first starts a loop process that repeats the process for each of all combinations of the two types of survey targets (seller and sales agent) identified in step S602. .

In each loop process, first, the number of evaluation data (satisfaction with the combination of the seller and the sales agent) to be processed in the loop is counted, and a predetermined lower limit is set. It is determined whether or not the value is greater than or equal to the value (S6201). If it is less than the lower limit value, the process proceeds to the process for the next combination to be investigated in the loop process. Here, as the lower limit value, a minimum number (for example, 2) that can be randomly extracted, which will be described later, is set.

If it is determined in step S6201 that the number of data is greater than or equal to the lower limit, it is next determined whether the number of data is less than or equal to a preset upper limit (S6202). Here, the upper limit value is, for example, a value obtained by subtracting 1 from the minimum number that is generally recognized as the number of populations (the number of populations), as in step S304 of FIG. 5 of the first embodiment. (For example, 600,000) is set. If the number of data is less than or equal to the upper limit value, the process advances to step S6204 for all data as processing targets. On the other hand, if the number of data exceeds the upper limit, the upper limit several pieces of data are extracted from the latest evaluation data in order from the latest, and are processed (S6203), and the process proceeds to step S6204.

In step S6204, a predetermined number is randomly extracted from the data to be processed using a random number table or the like previously stored in the value evaluation server 10 (S6204). That is, a predetermined number of evaluation data of 1 or more and (the number of evaluation data to be processed −1) or less is randomly extracted. Note that if the number of data is less than the lower limit in step S6201, random extraction cannot be performed in this step, and as a result, the number of data extraction is zero (ie, missing data) for the survey target combination. Become. Thereafter, the process proceeds to the process for the next combination to be investigated in a loop process.

When the loop processing for all combinations of survey targets is completed, a matrix table of evaluation data for two types of survey targets is created (S6205). As in the first embodiment, the value converted by the evaluation data conversion rule shown in FIG. 28 is used as the evaluation data used in the matrix table. At this time, when there are a plurality of values after the evaluation data conversion randomly extracted by the loop processing described above for each combination to be investigated, the average value of these evaluation data is set at a corresponding position on the matrix table. Other statistical values such as a median value and a mode value may be used instead of the average value. If there is one randomly extracted value after evaluation data conversion, the value of the evaluation data is set at a corresponding position on the matrix table. If random data cannot be extracted and the data is missing, the missing data (for example, a NULL value) is set at a corresponding position on the matrix table.

Next, the number of missing data is counted for each row and column of the matrix table created in step S6205 (S6206). Next, it is determined whether or not there is missing data in the result counted in step S6206 (S6207). If there is missing data, the row or column with the most missing data in the matrix table (sales agent or seller). Is deleted (S6208). At this time, if there are a plurality of corresponding rows or columns, the row or column corresponding to the seller or sales agent having a low priority is deleted. The priority order for the seller and the sales agent is determined based on the quotation request seller / agent priority list registered in the user DB 102 in the initial registration process in step S05 of FIG.

33 and 34 are diagrams showing an example of deleting a row or column with missing data from the matrix table of the combination of the seller and the sales agent created in step S6205 of FIG. Here, an example of the order in which rows or columns with missing data are deleted in a matrix table composed of combinations of seven sellers (sellers A to G) and seven sales agents (agents a to g) Shows about.

The upper table of FIG. 33 shows an example of the matrix table created in step S6205 of FIG. 32. The representatives ag are each row, the sellers Ag are each column, and the rows and columns are shown. Each combination (combination of each seller and sales agent) has an average value of satisfaction after conversion of evaluation data for these. The priorities in the table are priorities for sellers and sales agents set in the quotation request seller / agent priority list registered in the user DB 102. In the example of FIGS. 33 and 34, the priority is set in order from the first place for the sales agent and the entire seller (row and column). The matrix table also includes information on the number of missing data in each row and each column counted in step S6206 in FIG.

33. In the upper table of FIG. 33, missing data exists, and the most missing data is the row of the agent g and the column of the seller F (number of missing data = 3). Here, since the row with the lowest priority among these rows and columns is the row of the substitute g (priority = 14), this row is deleted by the processing of step S6208 in FIG. A table obtained by counting the number of missing data in step S6206 of FIG. 32 with respect to the deleted matrix table is shown in the lower table of FIG. Here, there is still missing data, and the rows with the most missing data are the row of the agent f and the column of the seller F (number of missing data = 2). Here, the row having the lowest priority among these rows and columns is the row of the substitute f (priority = 12), so this row is deleted. A table obtained by counting the number of missing data in the deleted matrix table is shown in the upper table of FIG.

In the upper table of FIG. 34, missing data still exists, and the most missing data are the rows of agents c and e and the columns of sellers F and G (number of missing data = 1). Here, the column of the seller G (priority = 13) has the lowest priority among these rows and columns, so this column is deleted. A table in the middle of FIG. 34 shows the number of missing data counted in the deleted matrix table. Here, the missing data still exists, and the rows with the most missing data are the row of the agent e and the column of the seller F (number of missing data = 1). Here, among these rows and columns, the column with the lowest priority is the column of seller F (priority = 11), so this column is deleted. The matrix table of the deleted result is shown in the lower part of FIG. Since there is no missing data in this matrix table, this is the final matrix table.

In this way, by determining the rows or columns to be deleted based on the information on the number of missing data and the priority information, the missing data is displayed in a matrix table while leaving as many survey targets as possible to be evaluated. It becomes possible to remove from. In the examples of FIGS. 33 and 34, the order of priority is set in order from the first place for the sales agent and the seller (row and column) as a whole, but each sales agent and each seller (each row and each column). ) May be set individually in order from the first. That is, the priority order may be set in order from the first place for each sales agent, and the priority order may be set in order from the first place for each seller.

In this case, if you try to delete a row or column with missing data using the same procedure as above, the sales agent (row) and the seller (column) may have the same value with the lowest priority. Come. Therefore, in this case, one row or column is selected according to a predetermined rule set in advance. In the present embodiment, for example, a rule such as deleting sellers (columns) preferentially (treating sellers with lower priority than agents) is set, and in the initial registration process of step S05 in FIG. The information is registered in advance in the user DB 102, for example, by adding the information to the estimate request seller / agent priority list.

Returning to FIG. 32, in step S6207, if there is no missing data (has been lost), that is, if all the rows and columns with missing data have been deleted, the deleted columns and rows (seller or sales agent). The deletion rate is calculated, and it is determined whether or not the deletion rate is less than a predetermined threshold (S6209). The deletion rate is a percentage of the number of columns or rows deleted in the matrix table divided by the number of columns or rows before deletion. Alternatively, a percentage obtained by dividing the total of deleted columns and rows by the total of columns and rows before deletion may be used. The predetermined threshold is a percentage at which the minimum number of survey targets to be evaluated can be secured. For example, half of the number of sellers or sales agents registered in the quotation request seller / agent priority list of the user DB 102 A value indicating 50% is used. If the deletion rate is less than the threshold value in step S6209, the matrix creation process ends. In this case, Friedman's test or Wilcoxon's signed rank test is performed in the statistical processing in step S630 in FIG.

On the other hand, if the deletion rate is greater than or equal to the threshold value in step S6209, the Kruskal-Wallis test or Wilcoxon test is performed in place of the Friedman test or Wilcoxon signed rank test in the statistical processing in step S630 of FIG. For this reason, first, the statistical calculation unit 13 is equal to or higher than the lower limit value of the number of data used for the Kruskal-Wallis test or the Wilcoxon test (for example, the lower limit value similar to that set in step S303 in FIG. 5 of the first embodiment). The number S of survey targets having the number of data is counted for each survey target group (S6210). Specifically, among the survey targets extracted in step S603 of FIG. 31, the number of items whose total number of evaluation index data possessed by the survey target is equal to or greater than the above lower limit value is counted.

Next, it is determined whether or not the number of survey targets in the corresponding survey target group in the matrix table from which all missing values are deleted is less than the number S of survey targets calculated in step S6210 (S6211). If it is less than S, the matrix creation process ends. In this case, for each survey target that is counted in step S6210, for example, a predetermined number of samples are extracted by the process shown in steps S304 to S306 in FIG. The Kruskal-Wallis test or Wilcoxon test is performed in the statistical processing of S630.

On the other hand, if the number of survey targets in the corresponding survey target group in the matrix table from which all missing values are deleted in step S6211 is greater than or equal to the number S of survey targets calculated in step S6210, the table is further displayed on the matrix table. It is determined whether there is data, that is, whether there are any remaining rows or columns (S6212). If there is data on the matrix table, that is, if there are remaining rows or columns, the matrix processing is terminated. In this case, Friedman test or Wilcoxon signed rank test is performed in the statistical processing in step S630 in FIG.

If there is no data on the matrix table in step S6212, that is, if all the rows and columns of the matrix table have been deleted by the series of processing in steps S6206 to S6208, a message to stop the processing is displayed. The data is output to the user terminal 30, and the entire process is terminated (S6213). Examples of such messages include a message such as "Cannot be calculated due to lack of recorded data. Please review the contents of the quotation request seller / substitute priority list and request a quotation again." can do. The user terminal 30 notifies the user by outputting the message by screen display or voice.

Alternatively, instead of ending the entire process, the matrix creation process may be ended, and the process may proceed to an estimate request in step S604 in FIG. In this case, since the statistical processing in step S630 after the request for quotation in step S604 in FIG. 31 is not performed, the statistical result regarding the value evaluation is not output to the user.

Returning to FIG. 31, when the matrix creation processing in step S620 is completed and there is data in the matrix table, that is, when there are remaining rows or columns, each investigation object corresponding to the rows and columns of the matrix table is next. An estimate request is transmitted to (S604). The quote request to be transmitted is, for example, the product etc. information acquisition unit 11 based on the quote request product list or the quote request service list registered in advance in the user DB 102 in the initial registration process in step S05 of FIG. Can be created. In addition, when the product etc. providing system 20 or the sales agent system 21 to which the quotation request is transmitted have an interface for accepting the quotation request online, a file including the contents of the quotation request is transmitted. Instead, an estimate request may be input automatically or manually using the interface.

FIG. 35 is a diagram showing an example of an estimate request transmitted to the seller's product etc. providing system 20. In the example of FIG. 35, in addition to identification information such as the IDs of sellers and users to whom an estimate request is transmitted, each item for which an estimate request is made is set in the estimate request item list registered in the user DB 102 in advance. Based on the contents, attribute information of the target product such as the product name, production area, grade, and size, and purchase information such as the planned number of purchases and the planned purchase date and time are set. In addition, since the estimate request is automatically performed, the scheduled purchase date and time may be specified by a relative date and time such as “after the estimated date and time”, for example, instead of the absolute date and time.

FIG. 36 is a diagram showing an example of an estimate request transmitted to the sales agent system 21. In the example of FIG. 36, in addition to identification information such as IDs of sales agents and users to which a quote request is transmitted, based on contents set in a quote request service list registered in advance in the user DB 102. Information such as the purchase date / time, delivery date / time, and delivery destination, and a list of sellers to be subjected to the agency service are set. The IDs and names of the sellers and sales agents are set based on the contents registered in the seller DB 101 and the agent DB 106. Since the request for quotation is made automatically, the purchase date / time and delivery date / time are not specified by the absolute date / time, for example, relative to each other such as “after XX hours after the estimated date / time” or “after XX hours after the purchase date / time”. It may be specified by date and time. Further, the list of sellers may be limited to the seller who has transmitted the request for quotation in step S604 of FIG.

The product providing system 20 or the sales agent system 21 that has received the request for quotation specifies the target product from the content of the request for quotation, and includes the price of the target product or agent service from the product content DB 201 or the agent content DB 211. Is extracted and transmitted to the value evaluation server 10 as an estimate (S605). The processing may be manually performed by a seller or a sales agent.

Specifically, the merchandise provision system 20 of the seller who has received the quote request identifies the target product etc. from the contents of the quote request, and extracts price and inventory information of the merchandise etc. from the merchandise content DB 201. . Thereafter, based on the planned number of purchases specified in the estimate request and the unit price extracted from the product etc. content DB 201, a subtotal, an estimated total amount, a consumption tax, and the like are calculated. In addition, information such as sales products that are not subject to estimation, advertisements / sale / feature products, products selected based on the purchase history of the target user, and the like are attached and transmitted to the value evaluation server 10 as an estimate. The selection of the product based on the purchase history may be performed using the purchase information of the user in the past recorded in the product etc. content DB 201, or the value evaluation server 10 uses the product information acquisition unit 11 to provide the product etc. providing system. When transmitting the quote request to 20, the product selected using the purchased product history information in the user DB 102 may be added to the quote request in advance.

FIG. 37 is a diagram showing an example of an estimate created by the seller's product etc. providing system 20. In the example of FIG. 37, the information of the estimate request in FIG. 35 received from the value evaluation server 10 and information such as the number of items for sale, unit price, subtotal, estimated total amount (bold frame in the figure) Item) is added for estimation. It should be noted that the product providing system 20 does not add information to the column of the planned purchase number and purchase date and time of unquoted products.

Also, the sales agent system 21 of the sales agent who has received the quotation request specifies the content of the sales agent service of the agent or the like acting on the basis of the content of the quotation request, and extracts the corresponding service fee from the agent content DB 211. Based on the extracted information, the content of the portion related to the charge (the item in the thick frame in the figure) is supplemented to the content of the request for quotation in FIG. 36 received from the value evaluation server 10, and the value evaluation server 10 is estimated. Send to.

Returning to FIG. 31, the value evaluation server 10 transmits an estimate request to the product etc. providing system 20 or the sales agent system 21 in step S604 to obtain an estimate, while the statistical calculation unit 13 creates the matrix table created in step S620. Statistical processing is performed to test whether there is a significant difference for each of the above survey targets (seller and sales agent) (S630). The statistical calculation method is the same as that in FIGS. 16 and 17 of the first embodiment when the Friedman test or Wilcoxon signed rank test is performed. That is, when comparing a plurality of evaluation indexes for three or more survey targets, the same as the Friedman test performed in step S3302 and subsequent steps in FIG. 16, and for the two survey targets, Wilcoxon performed in step S3315 and subsequent steps in FIG. This is the same as the signed rank test.

FIG. 38 is a diagram showing an example in which a significant difference between two types of survey target groups each having five survey targets is determined by statistical calculation from the value of the evaluation index. In the upper table of FIG. 38, the evaluation data is converted into a matrix table of combinations of five sellers (sellers A to E) and five sales agents (agents a to e) to be investigated. Average values of satisfaction (0 to 100 points) are shown. In the middle left table of FIG. 38, the sales agent group is calculated, and in the middle right table, the seller group is calculated with respect to the ranking calculated for the average value of satisfaction after conversion of the evaluation data (same sales agent). The ranking of each seller and the ranking of each sales agent in the same seller), and information on the sum of ranks obtained by summing the rankings for each survey target. FIG. 38 shows an example in which the Friedman test is performed on both the seller and the sales agent.

FIG. 38 shows an example in the case where there is no same rank in the ranks calculated for the survey target. In this case, the FR amount is calculated by the same process as step S3308 in FIG. In the example of FIG. 38, the number of survey targets (for example, the number of sellers) k = 5 and the number of evaluation data in each survey target (for example, the number of sales agents for the seller) for both the seller and the sales agent. ) Since m = 5, when the significant difference between the sellers is determined by the above equation (6), the FR amount is 10.72 in the case of determining the significant difference between the agents in the left table. In the right table, the FR amount = 3.20.

Next, the limit value χ given by the predetermined significance level α and the degree of freedom φ = k−1 from the chi-square distribution table previously held by the statistical calculation unit 16 and the like by the same processing as step S3309 in FIG. ² (φ, α) is specified, and the obtained χ ² (φ, α) is compared with the FR amount calculated above. In the example of FIG. 38, in the case of any significant difference determination between sellers and sales agents, the limit value is obtained from the chi-square distribution table based on the significance level α = 0.05 and the degree of freedom φ = 5-1 = 4. χ ² (4,0.05) = 9.49 is obtained.

Next, by the same process as step S3310 of FIG. 16, it is determined whether there is a significant difference between the investigation targets based on the comparison result, and the statistical process is terminated. Here, it is determined that there is a significant difference when the FR amount ≧ χ ² (φ, α), and it is determined that there is no significant difference when the FR amount <χ ² (φ, α). . In the example of FIG. 38, in the determination of the significant difference between the sellers (the figure on the left side), (FR amount = 10.7)> {χ ² (4,0.05) = 9.49}. Judge that there is a difference. In the determination of a significant difference between sales agents (right figure), there is a significant difference because (FR amount = 3.20) <{χ ² (4,0.05) = 9.49}. It is determined that it cannot be said.

Note that, as described above, when the number of survey targets k to be significant is 2, the Wilcoxon signed rank test is used and the test is performed in the same manner as the method shown in FIG. At this time, each survey target in a survey target group different from the survey target group subjected to the significant difference test corresponds to each evaluation index in the first embodiment. When the number k of the investigation objects to be tested for significance is 1, the statistical calculation is not performed, and it is determined that there is no significant difference because there is no investigation object to be compared. At this time, when the result is displayed by the processing described later, for example, a message such as “There is no investigation target to be compared, and it cannot be said that there is a significant difference as a result” may be added. If there is a significant difference in the Friedman test, the Wilcoxon signed rank test is used as a significant difference test between any two survey subjects in the survey target group that becomes significant as in the first embodiment. You may go further.

On the other hand, when the Kruskal-Wallis test or the Wilcoxon test is performed, the significance test is performed by the same processing as the statistical processing shown in FIGS. 6 and 7 of the first embodiment. That is, when comparing three or more survey targets, a significant difference test is performed by the same process as the Kruskal-Wallis test performed in step S3219 and subsequent steps in FIG. For the two survey targets, a significant difference test is performed by the same process as the Wilcoxon signed rank test performed in step S3202 and subsequent steps in FIG. In addition, when performing the same statistical calculation for the two types of survey target groups, it is performed for each group.

Also, if the number of survey targets k is 1, statistical calculation is not performed, and it is determined that there is no significant difference because there is no survey target to be compared. At this time, when displaying the result by the processing described later, for example, a message such as “There is no investigation target to be compared and it cannot be said that there is a significant difference as a result” may be added. If the survey target becomes significant by the Kruskal-Wallis test, between the two businesses in the survey target group that became more significant (for example, between sellers A and B, or between users a and b) It is also possible to determine a statistically significant difference by using the Wilcoxon test. As a result, information corresponding to the result obtained by Wilcoxon signed rank test after Friedman test as described in the first embodiment is obtained.

FIG. 39 is a diagram showing an example of combinations of test means when the deletion rate exceeds the threshold and the number S of survey targets is larger than the number of survey targets in the matrix table after all missing values are deleted. Here, in the matrix table to create a series of processing shown in FIG. 32, the row and column in the initial matrix table created in step S6205 the number (the number of surveyed) _{S p} and a series of steps S6206 ~ S6208 Survey with the number of survey targets S _a in the matrix table after deleting rows and columns with missing data by processing, the deletion rate calculated in step S6209, and the number of data greater than or equal to the predetermined lower limit calculated in step S6210 The method of the test corresponding to the combination of the number S of objects is shown.

In the example of FIG. 39, when the threshold value of the deletion rate in step S6209 in FIG. 32 is 50%, for the combination of the number of survey targets before and after deletion of rows and columns with missing data in the matrix table, The corresponding Kruskal-Wallis test or Wilcoxon test is shown. Using this combination information as definition information, the deletion rate is greater than or equal to the threshold in step S6209 in FIG. 32, and the number of survey targets in the matrix table after deletion of rows and columns with missing data is smaller than S in step S6211. In some cases, it can be determined whether to perform Kruskal-Wallis test or Wilcoxon test.

Note that the determination processing in steps S6209 to S6211 of FIG. 32 is not provided, and it is also possible to perform dedicated processing for the Friedman test and Wilcoxon signed rank test, or the Kruskal-Wallis test and Wilcoxon test, respectively. is there.

Returning to FIG. 31, next, the evaluation result output unit 14 of the value evaluation server 10 receives the product etc. from the merchandise providing system 20 of the seller or the sales agent sales agent system 21 that transmitted the request for quotation in step S <b> 604. Estimate information including information on the price, inventory, etc., and based on this information and the result of statistical calculation in step S630, the price and value for each survey target (evaluated based on the target evaluation index) A table or graph in which the values can be compared is created (S606). As for the value, as in the first embodiment, the rank of the statistical value of the value after the evaluation data conversion, the statistical value of the value before the conversion of the evaluation data, or the statistical value of the value after the conversion of the evaluation data can be used. .

If there is insufficient information in the received estimate information, or if the estimate information is not received within a predetermined time, the row or column including the sender (seller or sales agent) of the estimate is changed to the step of FIG. Delete from the matrix table created in S620, and re-execute the statistical processing in step S630. Then, the graph creation of step S606 is performed. In order to avoid re-execution of such statistical processing, the value evaluation server 10 targets the sender of the estimated information included in the matrix table created in step S620 after receiving the estimated information in step S605. As described above, the statistical processing in step S630 may be performed.

FIG. 40 is a diagram showing an example of the relationship between the price (estimated price) and the value (value evaluated based on the target evaluation index) for each survey target in a tabular format and a graph. In FIG. 40 (a), for each of the two types of survey subjects (sellers A to E) and sales agents (agents a to e), estimated price information and evaluation indices (satisfaction after conversion of evaluation data) are obtained. This shows an example in which information on the sum of ranks based on (degree) (for example, the rank sum shown in the middle table of FIG. 38) is represented in a table format. The table also includes information on the significant difference between the survey targets determined by the statistical processing in step S630 in FIG. 31 and information on the presence or absence of graph display. Note that the graph may be created only when it is determined that there is a significant difference between the survey targets.

In FIG. 40B, since it is determined that there is a significant difference in evaluation index between sellers in FIG. 40A, this is based on price (estimated price) and value (target evaluation index (satisfaction)). The example shows the relationship of (evaluated value) in a scatter diagram. Here, the estimated price is on the vertical axis, and the value of the sum of ranks based on the satisfaction after the evaluation data conversion is on the horizontal axis. When there is a significant difference between the two survey targets, a scatter diagram for the two survey targets may be shown. In addition, a scatter diagram of one three-dimensional (for example, the estimated price is Z-axis, the seller's rank sum value is the X-axis, the sales agent rank sum value is the Y-axis) for the two survey target groups is shown. Also good.

Returning to FIG. 31, next, the evaluation result output unit 14 of the value evaluation server 10 performs a result output process in which the statistical result provided to the user and the information including the comment attached thereto are collected and output (S640). . FIG. 41 is a flowchart showing an overview of an example of the flow of the result output process in step S640 of FIG. The evaluation result output unit 14 first identifies the result of the statistical processing in step S630 or the information of the table or graph created in step S606, and based on these, there is a significant difference in evaluation index between sellers. It is determined whether or not (S6401).

In step S6401, if there is a significant difference in evaluation index between sellers, the price is low and the value of the rank sum is small, that is, the seller with the highest degree of satisfaction and the highest evaluation (FIG. 41 (b The comment that recommends the selection of sellers plotted in the lower left area in the graph shown in FIG. 6) is selected from those defined in advance (S6402). As a comment in this case, for example, a message such as “It is recommended to select a seller in the lower left area of the graph that is highly evaluated and cheap because there is a significant difference between sellers” be able to.

In step S6401, if there is no significant difference in evaluation index between sellers, a comment that recommends that a seller be selected only by price (that is, a seller with a low estimated price) is selected (S6403). The comment can be, for example, a message such as “There is no statistical difference in value and it is recommended to select an inexpensive seller”.

After that, further, a comment is selected for the sales agent on the basis of the information on the significant difference in the evaluation index by the same processing as the above steps S6401 to S6403 (S6404 to S6406). When a comment to be presented to the user is selected by the above-described series of processing, the result of processing without performing statistical calculation due to statistical information including a table or graph as shown in FIG. , The selected comment, and the received estimate information are collected in a predetermined format, and transmitted and output as result information to the user terminal 30 (S6407), and the result output process is terminated.

In the example of the result output processing flow of FIG. 41 described above, unlike the first embodiment, information emphasized by the user (evaluation index information emphasized in the first embodiment) is not included in the comment selection process. In addition, a step of determining based on information emphasized by the user can be added between steps S6401 and S6402 or between S6404 and S6405. For example, after step S6401, a step of determining whether or not the user attaches importance to a significant difference between sellers can be added. If it is important, the process proceeds to step S6402. If it is not important, the process proceeds to step S6403. As in the first embodiment, the information that is emphasized by the user may be stored in the user DB 102 at the time of user information registration in the initial registration process of S05 in FIG. The requested program may be attached and transmitted, and the information may be obtained by local processing in the user terminal 30.

In the example of FIG. 41, the evaluation result output unit 14 of the value evaluation server 10 selects a comment. However, as in the first embodiment, a program for executing a result output process and selecting a comment is It may be transmitted to the user terminal 30 as a client program and processed locally on the user terminal 30 side.

Further, as in the first embodiment, instead of outputting the comment selected by the result output process or the program for selecting the comment by executing the result output process to the user terminal 30, for example, You may make it transmit the table | surface expressing the determination content of the result output process to the user terminal 30 with a statistical result. FIG. 42 is a diagram illustrating an example of a table expressing the determination contents of the result output process. Users can grasp the recommended content as a selection criterion for sellers or sales agents by referring to a combination of such judgment patterns in a tabular format and statistical results. It is possible to select a seller or a sales agent more accurately after understanding the overall picture of recommended selection criteria.

As in the first embodiment, relative value judgment is recommended to the user regardless of whether the result output process of FIG. 41 is used or the table of FIG. 42 is presented to the user. Comments may be added. An example of such a comment is, “Value evaluation data is inherently non-parametric data, and thus may not accurately represent a difference in values. It is recommended to make a reasonable decision ”.

Returning to FIG. 31, the user terminal 30 outputs the result information including the statistical result and the comment output from the value evaluation server 10 by screen display or voice, so that the user can receive the result of the value evaluation between the survey targets. Reference / confirmation is made (S607). At this time, a message prompting reconfirmation of a product or service to be estimated and determination of a purchased product may be displayed. Thereafter, it is determined whether or not the content of the estimation target by the user has been modified by the user (S608). Here, for example, if the user has made corrections such as adding a product other than the target product of the request for quotation as a purchased product, etc., recalculate the estimate contents and correct the estimated total amount, etc. (S609).

Specifically, for example, in the example of the quotation created by the seller's product provision system as shown in FIG. 37, the “planned purchase number” and “planned purchase date and time” fields in the “unquoted product” field on the right side. The user terminal 30 recalculates the fields such as “subtotal” and “total charge”. Accordingly, the contents of the estimated price in the statistical result table or graph presented to the user shown in FIG. 40 may be updated. The recalculation process may be performed by the evaluation result output unit 14 of the value evaluation server 10 that has received a recalculation request from the user terminal 30, or the user terminal 30 may be executed by a client program or spreadsheet software. Local processing may be performed on the side.

For example, the estimated purchase date and time of the unquoted product accepts the date and time after the estimated purchase date and time of the estimate request product, and if it is not the same as the estimated purchase date and time of the estimate request product, it is temporarily stored in the estimate request schedule of the user DB 102. Then, a temporary automatic quotation request process may be activated a predetermined time before the scheduled purchase date and time. At this time, target quotation request commodity information is also temporarily stored in the quotation request commodity list of the user DB 102 and associated with the date and time. After the temporary automatic quotation request process is activated, the temporary storage information is deleted from the user DB 102.

Returning to FIG. 31, in step S <b> 608, if there is no correction to the content of the estimation target by the user, the user can determine the final estimation content or the content of the comment recommending the statistical result and selection criteria by the value evaluation server 10. The sales agent who purchases the merchandise or the like or requests the sales agent service is determined (S610).

[Processing flow (evaluation history information recording process)]
The processing flow of the evaluation history information recording process in step S07 of FIG. 30 is substantially the same as the processing flow shown in FIG. 26 of the first embodiment. That is, in step S610 of FIG. 31, the user selects the relationship between the value and price of the seller and the sales agent and the seller and the sales agent that are output by the value evaluation server 10 in advance. Information on recommended methods and the like are referred to and confirmed via the user terminal 30, and a combination of a seller and a sales agent is selected and a product to be purchased is determined.

After that, the user uses the user terminal 30 to make a payment for the product, etc., for which purchase has been decided by a predetermined method such as Internet banking, credit card, electronic money, etc. The user terminal 30 inputs necessary information such as information relating to payment, such as the option that has been selected, and transmits the purchase payment information related to the selected seller, sales agent, and the determined product to the value evaluation server 10. (S401).

Upon receiving the purchase settlement information, the value evaluation server 10 transfers a copy of the purchase settlement information to the merchandise provision system 20 of the seller selected by the user and the sales agency system 21 of the selected sales agent. (S402). The product provision system 20 and the sales agent system 21 that have received the purchase settlement information perform sales and settlement processing on the target product based on the contents of the received purchase settlement information, and value evaluation is performed on the processing result including the sales settlement information. It transmits to the server 10 (S403). The sales settlement information includes, for example, the contents of the target product, information on the seller and the sales agent, the sales amount, and information on the user who has purchased.

The value evaluation server 10 that has received the sales settlement information from both the merchandise providing system 20 of the target seller and the sales agent system 21 of the target sales agent, the sales settlement information and the purchase settlement information received in step S402. Confirm that the target product, price, and other sales conditions are the same. Further, the questionnaire processing unit 15 specifies and extracts information on a questionnaire for inputting evaluations on the target seller and sales agent (or products sold by them) from the questionnaire DB 105 (S404). This questionnaire can be obtained by adding a question item related to a sales agent to the one shown in the example of FIG. 27 of the first embodiment, for example. The questionnaire preferably includes a plurality of questions associated with the type of seller, sales agent, product, or the like.

Thereafter, the questionnaire processing unit 15 transmits the extracted questionnaire information and the sales settlement information received from the product providing system 20 and the sales agent system 21 to the user terminal 30 (S405). The information to be transmitted may include help information related to a questionnaire input method and the like. The user terminal 30 receives the information and outputs it to the user, so that the user can receive the product or the like purchased from the sales agent according to the information, Evaluations for agents, purchased products, etc. can be entered into the questionnaire.

The inputted questionnaire information is transmitted from the user terminal 30 to the value evaluation server 10 (S406). Upon receiving the questionnaire input result, the value evaluation server 10 extracts the contents of the questionnaire input by the questionnaire processing unit 15 and evaluates it as evaluation information for the target seller, sales agent, purchased product, etc. Record in the history DB 104 (S407). The questionnaire processing unit 15 transmits the questionnaire again to the user terminal 30 when the questionnaire input result cannot be received even after a predetermined period of time has elapsed since the transmission of the questionnaire in step S405. You may ask for a reminder.

As in the first embodiment, various methods can be used for exchanging questionnaire information between the value evaluation server 10 and the user (user terminal 30). For example, as in the processing flow described above, a questionnaire file in which the contents of a question and an answer column for evaluating each evaluation index are described is transmitted from the value evaluation server 10 to the user terminal 30, and the user On the other hand, after inputting an evaluation, the file can be configured to be returned to the value evaluation server 10.

In this embodiment, significant differences are determined using the Friedman test or Wilcoxon signed rank test mainly in cases where two types of survey target groups (seller group and sales agent group) are related to value evaluation. However, in cases where each surveyed group is solely involved in value assessment, Kruskal-Wallis test is used instead of Friedman test, and Wilcoxon test is used instead of Wilcoxon signed rank test. Significant difference can be determined.

For example, in an Internet auction, there are a plurality of sellers and a plurality of users, and each of them evaluates the other party with whom the transactions were conducted (in this case, two types of seller groups and user groups are subject to investigation) In a group, where both parties are actually involved in value evaluation), there will be situations where most users will only deal with a small number of sellers. In this case, in the Friedman test, the missing data in the matrix table created in step S6205 in FIG. 32 increases for the two types of survey target groups, and the rows and columns containing the missing data are obtained by the processing in steps S6206 to S6208 in FIG. In the matrix table as a result of deleting, the number of investigation objects is greatly reduced, and the intended comparison may not be possible.

On the other hand, in the Kruskal-Wallis test, the number of survey target groups related to the evaluation of value is set to 1, so there is no need for a matrix table for two types of survey target groups, and therefore significant without reducing the number of survey targets. The difference can be tested. For example, whether or not there is a statistically significant difference between sellers by randomly extracting and ranking a predetermined number of information evaluated by each user for each of a plurality of sellers and performing statistical calculations Can be determined. Further, by performing the same calculation for each user, a significant difference between users can be determined. The conditions such as the number of data necessary for the calculation are the same as those described in the first embodiment.

Note that the Kruskal-Wallis test or the Wilcoxon test is used in a situation where the two companies (the seller and the sales agent) mutually affect the value of the evaluation index as in the example of the present embodiment described above. And the evaluation data obtained for each combination of survey targets (between two operators) in the matrix table cannot be statistically calculated. Therefore, the statistical accuracy is compared to the Friedman test or Wilcoxon signed rank test. It will be inferior.

<Embodiment 3>
In general statistical calculations including the statistical calculation used in the first embodiment and the second embodiment described above, it is assumed that the sample used for the statistical calculation accurately represents the population. . However, there may be a case where the sample of evaluation data extracted from the evaluation history information accumulated in the evaluation history DB 104 of the value evaluation server 10 does not necessarily accurately represent the population.

Therefore, the value evaluation support system according to the third embodiment of the present invention has the same configuration and function as the value evaluation support system 1 according to the first or second embodiment, and further uses evaluation for statistical calculation. It has a function of selecting an appropriate supplementary explanation for the result of the statistical calculation from the combination of the property of the data and the information related to the statistical calculation, and outputting and notifying it via the user terminal 30.

This supplementary explanation can be output when the evaluation result output unit 14 of the value evaluation server 10 outputs the processing result to the user terminal 30, for example. Alternatively, a client program that performs such processing may be transmitted to the user terminal 30 and an appropriate supplementary explanation may be selected and displayed by local processing on the user terminal 30 side. Thereby, even when there is no knowledge about statistical calculation, the user can understand easily and in detail about the degree of reliability of the statistical result output by the value evaluation server 10.

43 to 44 are diagrams showing examples of supplementary explanation setting tables that hold supplementary explanation contents corresponding to combinations of properties of evaluation data used for statistical calculation and statistical calculation information. FIG. 43 is a diagram showing an example of a part in the supplementary explanation setting table that defines the property of the evaluation data and the combination of statistical calculation information for each of a plurality of investigation objects (survey objects A, B,...). In the table, the nature of the evaluation data in each survey target is defined by items such as the number of samples of evaluation data, the number of abnormal values, the acquisition period, the number of evaluators, as shown in the figure.

Here, the number of abnormal values indicates, for example, the number of evaluation data outside the range of a predetermined threshold (for example, average value of evaluation data ± 4σ, etc.). The number of evaluators indicates, for example, the number of persons who have evaluated the evaluation index (provided evaluation data). In the first embodiment and the second embodiment described above, this corresponds to the number of writers who have filled in an answer to the questionnaire that is the basis for obtaining the evaluation data. The term “acquisition period” indicates the period of the entry date entered by the entrant in the questionnaire.

43. As shown in FIG. 43, the nature of evaluation data used for statistical calculation is defined for each survey target. Here, for example, the evaluation data included in one survey target (for example, survey target A) may be a value determined by a combination of a plurality of evaluation indexes or other survey targets. In this case, for the target evaluation data, for example, a value that compares the values for each combination with another evaluation index or another survey target and that most deteriorates the accuracy of statistical calculation is used.

Specifically, for example, the number of samples is the number of data in the combination with the smallest number of data among the plurality of combinations of the survey target A. Similarly, as the number of abnormal values, the number of abnormal values in the combination having the largest number of abnormal values among a plurality of combinations of the investigation target A is used. Similarly, for the acquisition period, the acquisition period in the combination having the shortest data acquisition period among the plurality of combinations of the survey target A is used. Similarly, the number of evaluators uses the number of evaluators in the combination having the smallest number of evaluators among the plurality of combinations of the survey target A.

Statistic calculation information is information on values and conditions used for statistical calculation, and is defined by items such as significance level, test method, P value range, average rank, approximate distribution, as shown in the figure. Here, as described above in the first embodiment, the P value is not a type for obtaining a significant point (limit value) when using various test tables or distribution tables in statistical calculation, but P value (significant). It is a numerical value specified using a test table or a distribution table of a type for obtaining a probability. For example, the P value is specified from the FR amount, the number k of survey targets, and the number m of significant evaluation indexes using a Friedman test table of a type for obtaining the P value. Alternatively, the P value is specified from the FR amount and the degree of freedom φ using a chi-square distribution table of a type for obtaining the P value. In the example of FIG. 43, the range of the P value is shown as a percentage.

Further, the average rank indicates information on whether or not the average rank used when the evaluation data has the same value or the same rank in step S3214 in FIG. In addition, the approximate distribution indicates information on whether or not the approximate distribution used when the number of data is large or the average ranking is used, such as a chi-square distribution or a normal distribution. In the table of FIG. 43, No. 1 is sequentially applied from 1 for each combination pattern of the evaluation data property and statistical calculation information for each survey target. Can be identified.

FIG. 44 is a diagram showing an example of a part in the supplementary explanation setting table that defines the contents of the supplementary explanation corresponding to the combination of the evaluation data property and the statistical calculation information for each survey target. In the table of FIG. 44, each No. identifying the nature of the evaluation data and the combination pattern of the statistical calculation information for each survey object in FIG. In contrast, the contents of supplementary explanations such as comments, conclusions, and advice to be presented to the user are defined. As shown in the table of FIG. 44, the contents of each supplementary explanation are expressed by symbols (C1, D1, A1, etc.) instead of the explanatory text, and the specific explanatory text may be held as another table or data. Good. FIG. 45 is a diagram showing an example of a table defining a pattern of explanatory text for supplementary explanation.

In the following, an example of the flow of processing when selecting and outputting a supplementary explanation will be described. First, in the value evaluation server 10 having the same function and configuration as those of the first embodiment or the second embodiment described above, the evaluation result output unit 14 performs step S340 in FIG. 5 or step S640 in FIG. The statistical calculation information in the statistical calculation in steps S320 and S330 and step S630 in FIG. 31 and the nature of the evaluation data used for the test are specified. Next, based on the specified information, the contents of each item of the supplementary explanation are extracted from the supplementary explanation setting table shown in FIGS. 43 and 44 and the table defining the explanatory text pattern of the supplementary explanation of FIG. . The content of the extracted supplementary explanation is transmitted to the user terminal 30 together with the result of the statistical processing or the like in step S340 of FIG. 5 or step S640 of FIG.

In step S401 in FIG. 26, the user refers to the explanatory text such as comments, conclusions, and advice in the supplementary explanation transmitted from the value evaluation server 10 via the user terminal 30 when determining the product to be purchased. To do. This makes it possible to understand the degree of reliability of the statistical calculation results and to make a purchase decision more accurately.

In the above-described second embodiment, since two types of tests, ie, a test for a seller and a test for a sales agent, are performed, the evaluation result output unit 14 selects the contents of supplementary explanation for each test, The user is notified of the result of the test and the corresponding supplementary explanation. Thereby, the user can grasp | ascertain the supplementary explanation with respect to which test, without being confused even when there are a plurality of supplementary explanations.

Further, the process of selecting supplementary explanation can be made to correspond to various tests in the first and second embodiments. For example, after Wilcoxon signed rank test, corresponding supplementary explanation items may be extracted from the statistical calculation information obtained by the test and the nature of the evaluation data used for the test. Similarly, it may be performed after the Friedman test, Kruskal-Wallis test, or Wilcoxon test. Furthermore, not only when the test is a two-sided test, but also when the test is a one-sided test, a supplementary explanation can be selected by the same method as described above after the statistical calculation.

Also, the evaluation result output unit 14 may perform processing such as directly selecting recommended products instead of selecting supplementary explanations. For example, instead of selecting the supplementary explanation of “D8” in the example of FIG. 45, the product having the highest value and the lowest price shown in FIG. 15B is specified, and this information is used as the recommended product. You may transmit to the terminal 30. Alternatively, both the supplementary explanation of “D8” and the information on the specified product may be transmitted to the user terminal 30. Thereby, the user can be notified of the statistically most valuable product or the like. Further, for example, in the result output process of step S340 in FIG. 5, by notifying the user of the product specified when performing the process of step S3410 or S3411 in FIG. 24, the optimal product etc. for the user is recommended. can do.

As described above, according to the value evaluation support system 1 described in the first to third embodiments of the present invention, for example, as illustrated in the first embodiment, various evaluation indexes indicating the value of a product or the like. By presenting an evaluation index having a statistically significant difference to the user, it is possible to exclude an evaluation index having no significant difference from the judgment materials at the time of purchasing the product. Thereby, it is possible to support the determination of the difference in value when the user purchases a product or the like, and to improve the value determination power.

At this time, instead of converting the nonparametric data into an index or index that is parametric data under a certain rule as in the prior art, the statistical calculation is performed directly on the nonparametric data. It is possible to improve the accuracy of statistical calculation by solving the above problem. In addition, the parametric data is statistically calculated as non-parametric data, whereby all evaluation indexes can be evaluated by a unified calculation method and criteria. Therefore, if a significant difference is determined for all necessary evaluation indexes, it is possible to determine a difference in value existing between products and the like.

In addition, a statistically significant difference between sellers is determined by using a Wilcoxon signed rank test or a Friedman test to determine a significant difference from a combination of a statistically significant evaluation index and multiple sellers. And provide information to support the seller's choice. In addition, the user can receive recommendation of an appropriate selection criterion such as a product from the combination with the calculated statistical result by showing the evaluation index that the user attaches importance to the value evaluation server 10.

Further, for example, as shown in the second embodiment, an evaluation index of the value of a product etc. affected by two types of businesses (for example, a seller and a sales agent) (the influence of the two businesses is entangled) Use the Friedman test as a two-way method using the random block method to separate the influence of each business operator and determine the statistically significant difference between each seller or sales agent To do. Thereby, it becomes possible to judge the influence by each operator more accurately, clearly and easily. If the evaluation index data by the two types of combinations do not exist in many combinations, the Kruskal-Wallis test is used to make a significant difference, but the accuracy is slightly inferior to the Friedman test, but the value for more survey subjects It can be evaluated and can be judged accurately, clearly and easily.

In addition, the significant difference existing between any two operators (for example, sellers A and B) within the operators that became significant by Friedman test or Kruskal Wallis test is determined by Wilcoxon signed rank test or Wilcoxon test. By doing so, the significant difference between the operators in all combinations can be further determined. In addition, the user can receive the recommendation of the operator's selection criteria from the combination with the calculated statistical result by indicating the type of the operator that the user attaches importance to the value evaluation server 10.

In addition, for example, as shown in the third embodiment, when recommending selection criteria for a product or a business to a user based on the result of statistical calculation, the properties and statistics of evaluation data used for statistical calculation An appropriate supplementary explanation for the result of the statistical calculation is selected and notified from the combination with the information related to the calculation. Thereby, even when the user has no knowledge of statistical calculation, the user can easily and in detail understand the degree of reliability of the statistical result. Furthermore, by specifying the recommended product etc. using the selected supplementary explanation information, it is possible to recommend a product that is statistically most valuable to the user or optimal for the user.

As mentioned above, the invention made by the present inventor has been specifically described based on the embodiment. However, the present invention is not limited to the embodiment, and various modifications can be made without departing from the scope of the invention. Needless to say.

For example, in Embodiments 1 to 3 described above, the case where merchandise is bought and sold by electronic commerce has been described as an example. However, the present invention is not limited to this. For example, the value of a survey object whose value is expressed by an evaluation index such as sensory evaluation data where the data itself is ambiguous (large variation) or an evaluation index where data is ambiguous due to low measurement accuracy. Can be applied to a system or the like in which a statistically significant difference is determined and a standard for determining value is recommended to the user. For example, surveys of products, services, brands, companies, political parties, talents, characters, mascots, etc., and surveys on the value evaluated by evaluation indexes with many ambiguous data such as popularity, support rate, and favorableness. The present invention can be applied to a system that performs prediction and the like.

More specifically, evaluation data accumulated for each survey target for a certain evaluation index is extracted, converted into rank data, and then statistically significant difference is determined. Here, the Wilcoxon test is used for comparison between two persons, and the Kruskal-Wallis test is used for three or more persons. In addition, when comparing the evaluation data (popularity, honesty, support rate, favorability, etc.) and the evaluation data accumulated for each combination of survey subjects, Wilcoxon's signed rank test is performed. Use Friedman's test for more than 3 persons. Using the obtained test result, a comment or the like set in advance is selected and a survey result is created. In addition, in the same way, as a prediction based on various evaluation indices, for example, even when predicting CD ranking or land value, etc., it is possible to test the significant difference between survey subjects and create a prediction result. it can.

Also, as other fields of use, for example, it is possible to collect and store information indicating human emotions, sensations, moods, etc., and convert them into rank data for use in analysis or control. For example, in the control of an interactive robot with a person, it may be applied to a system that converts a person's emotion into rank data, causes the robot CPU to statistically calculate the difference in emotion, and selects an action to be taken by the robot. it can. More specifically, for example, the loudness of a person's speaking voice is measured and recorded multiple times in advance, and compared with the loudness of the voice measured and recorded multiple times at the present time, the human emotion is valued. Can be implemented as a program. Emotion changes when there is a significant difference in the volume of the voice by converting it into rank data that ranks the voice loudness (for example, the evaluation index is volume (decibel)) in ascending order and performing statistical calculations. Can be determined. Using the determination result, an action set in advance is selected.

In this case, since there are two types of subjects, current and past speech of a person, it can be determined by using the Wilcoxon test. When multiple other evaluation indicators are provided in addition to volume as an evaluation indicator indicating emotion, Wilcoxon's signed rank test is used for the data for each combination of these multiple evaluation indicators and tense (current or past). Test. Moreover, it can determine similarly about a sensory difference. For example, by giving a person certain light, sound, vibration, current, etc., measuring and recording the reaction speed (reaction time until a certain part of the person moves), and comparing the current and past ranking data A significant difference is determined. If there is a significant difference, it can be determined that the sense has changed.

When using the Wilcoxon test or the Wilcoxon signed rank test, it is possible to determine not only whether there is a significant difference between the past and the present, but which can be said to be significantly stronger between the past and the present. Specifically, for example, it can be determined that the current voice is stronger (or weaker) than the past. This calculation is made possible by changing the determination formula for the significance level α and the significant difference in the calculation method described with reference to FIGS. 6 and 17 of the first embodiment.

For example, in the Wilcoxon test, when the number of evaluation data n _MIN <15, W amount ≦ W _L (α) when the W amount consists of a past sample, and W amount ≧ W when the W amount consists of the current sample. _{If U} (α), it is determined that the current voice is stronger at the significance level α. _{Further, n MIN} if the case of ≧ 15 or have the same value, the W content is _{u 0} weight ≦ when consisting of past samples _{{-u (2α)}, u} 0 when the W content consists current sample When the amount ≧ u (2α), it is determined that the current voice is stronger at the significance level α.

On the other hand, in the Wilcoxon signed rank test, assuming that the difference calculation method is (present-past), when N <25, and when WS amount = WS + amount, WS amount ≧ t _U (α), WS amount = In the case of WS-amount, if WS amount ≦ t _L (α), it is determined that the current voice is stronger at the significance level α. Further, when N ≧ 25 or when there is the same order, it is determined that the current voice is stronger at the significance level α when u ₀ amount ≧ u (2α). Note that whether or not the current voice is weaker can also be determined using the same determination method by exchanging the current and past information and performing the same statistical calculation. The description here is based on the premise that the calculation formula for the u ₀ quantity is performed using Formula 1 and Formula 8.

Furthermore, for example, the present invention can be applied to a program that analyzes an image or video of an indefinite person or an object and determines whether or not a predetermined condition is met. For example, for cells with normal shape and cancer cells with slightly distorted shape, the shape such as the diameter of the cell is measured and stored from multiple angles, and is expressed by the shape between the cancer cell and the normal cell. It can be determined whether there is a significant difference in value. Here, for example, a Wilcoxon signed rank test is used to determine whether or not there is a difference in shape of the cells at the significance level α (that is, whether or not cancer cells can be distinguished by shape) and set in advance. You can select the conclusions of the survey you have made.

Also, for example, for the horizontal length of a person's lips, the length measured and recorded several times in the past and the length measured and recorded several times in the past are converted into rank data, and Wilcoxon test is used. Compare. Thereby, it is possible to determine whether or not the current length of the lips in the horizontal direction is longer (whether or not a person is smiling). In addition, when expanding the number of survey targets (number of people) to multiple, record the length for each survey target (people) and tense (current or past), and rank the average length for each combination. Conversion to data may be performed using Wilcoxon signed rank test. This evaluates the difference in value represented by the horizontal length of the lips (the degree to which multiple people are smiling). For example, whether or not the audience was smiling at a performance such as a comedy or a laugh. It is also possible to determine whether or not you are pleased.

Parametric data such as the degree of distortion of the shape of the cancer cell and the horizontal length of the lips when a person is smiling, even if an image measurement device with high measurement accuracy is used, the measurement object itself varies greatly. If the statistical calculation is performed as it is, the determination accuracy may decrease. In such a case, by using the mechanism as shown in the first to third embodiments of the present invention, by converting these parametric data into rank data that is nonparametric data, statistical calculation is performed. Reasonable and appropriate statistical significance can be determined.

As described above, the evaluation index is not particularly limited as long as it can represent the value of the product or the like. For example, various indicators such as a nominal scale, an order scale, an interval scale, or a proportional scale can be used. . The nominal scale includes, for example, three colors of red, blue, and yellow, which are the three primary colors of light, but these are noun information and do not represent the magnitude relationship. However, for example, when the length of the wavelength is used as a reference, a magnitude relationship is created in the order of blue, yellow, and red. For example, in the example of the questionnaire input screen shown in FIG. 27, the radio button displays the color of the heater (stove) as blue (3), yellow (2), red (1), etc. The evaluation of the value of the heating appliance can be performed as an evaluation index of “the effect of feeling warmth”.

Each processing procedure described in each of the above-described embodiments, particularly processing related to statistical calculation (for example, statistical processing in step S320 in FIG. 5 or step S630 in FIG. 31 and between survey targets in step S330 in FIG. 5). The processing procedure in the (significant difference determination process) is merely an example. Needless to say, as long as the same processing result can be obtained, optimization or the like by appropriately changing the order of some processes is possible.

For example, steps S3204, S3208, and S3213 in FIG. 6 and steps S3224, S3228, and S3232 in FIG. 7 that are ranking processes are performed in the same manner. For example, before the processing branches, such as before step S3201 immediately after the start of statistical processing, the processing may be performed in common. In this case, in step S3202 of FIG. 6 and step S3219 of FIG. 7, as described above, determination of the same rank is performed instead of determination of the same value. Similarly, for example, steps S3305 and S3308 in FIG. 16 perform the same FR amount calculation processing. Therefore, these processings may be integrated and placed in common between steps S3303 and S3304. .

Further, in each of the above-described embodiments, the value evaluation support system 1 includes a price survey process (for example, step S02 in FIG. 2) in which the value evaluation server 10 acquires information on a product such as a price, and the value of the product etc. Although both of the value evaluation processes (for example, step S03 in FIG. 2) such as calculating the difference are performed, the present invention is not limited to this. For example, the value evaluation server 10 may be configured to cause a price survey process to be performed by another server (price survey process dedicated server not shown).

For example, the server dedicated to price survey processing receives a request from the user, performs price survey processing, and requests the value evaluation server 10 to calculate value evaluation information regarding the specified product. The value evaluation server 10 that has received the request identifies the target product or the like from the received request information, and performs value evaluation processing by the same processing as described above. The processing result is returned together with the questionnaire information including the questionnaire entry request to the price survey processing dedicated server that transmitted the request. After that, the value evaluation server 10 receives the information that the user wrote in the questionnaire and sent it, records it in the evaluation history DB 104, and ends the evaluation history information recording process in step S04 of FIG.

Further, as another configuration example, the product etc. providing system 20 or the sales agent system 21 may have the function of the above-described price survey dedicated server, or may further have the function of the value evaluation server 10. Further, the product providing system 20, the sales agent system 21, and the value evaluation server 10 may have the functions of the user terminal 30, respectively.

Further, in each of the embodiments described above, in the processing related to statistical calculation, a significance level is set in advance and the presence or absence of a significant difference is determined. The degree of significant difference may be indicated depending on the size. In this case, in the process of determining a significant difference (for example, steps S3206 and S3207, steps S3211 and S3212, and steps S3217 and S3218 in FIG. 6), the P value is obtained by the above-described method. You may transmit to the user terminal 30 with the message to the effect of determining with it being small. In the case where a significance level is used as the threshold value for determination, it is possible to determine a significant difference as in the first to third embodiments.

In each of the above embodiments, the seller is not particularly limited as long as it is an entity capable of selling products and the like. For example, in addition to an individual or a corporation, if possible, an entity other than a person ( Computer systems etc.) can also be widely applied. Similarly, the sales agent is not particularly limited as long as it is an entity that can receive the product from the seller on behalf of the user and deliver it to the user. In addition, subjects other than humans can be widely applicable when possible. Similarly, the user is not particularly limited as long as it is an entity capable of requesting a value evaluation survey. For example, in addition to an individual or a corporation, an entity other than a person may be widely applicable.

The present invention can be used in a value evaluation support system and a value evaluation support program that provide information related to a difference in value as a judgment material to a user when there are a plurality of compatible products and services.

1 ... Value evaluation support system,
DESCRIPTION OF SYMBOLS 10 ... Value evaluation server, 11 ... Product etc. information acquisition part, 12 ... Evaluation history information acquisition part, 13 ... Statistical calculation part, 14 ... Evaluation result output part, 15 ... Questionnaire processing part, 101 ... Seller DB (database), 102 ... User DB, 103 ... Product etc. DB, 104 ... Evaluation History DB, 105 ... Questionnaire DB, 106 ... Agent DB,
20 (20a-d): Product etc. providing system, 201 (201a-d) ... Product etc. content DB, 21 (21a, b) ... Sales agency system, 211 (211a, b) ... Agency content DB,
30 ... user terminal,
40 ... Network.

Claims

Statistical analysis of the difference in value between surveyed items for multiple products or services whose value is expressed by one or more evaluation indicators including ambiguous data, or for businesses that are involved in sales of these products or services. A value evaluation server that determines a significant difference and outputs a determination result;
An information processing system connected to the value evaluation server via a network for the business to perform the business related to the sale of the product or service;
A value evaluation support system having a user terminal connected to the value evaluation server via the network,
The value evaluation server
The information processing system of the business operator that receives the input of conditions for specifying the survey object for which the user wishes to evaluate the difference in value via the user terminal and handles the product or service related to the survey object A product etc. information acquisition unit for acquiring information including the price of the product or service that meets the conditions from,
From the contents of the questionnaire input via the user terminal, a questionnaire for inputting evaluation data that is an evaluation result of the value for each evaluation index related to the survey target is output to the user terminal. An evaluation information input unit that acquires the evaluation data for each evaluation index and accumulates it in an evaluation history database;
An evaluation history information acquisition unit that extracts the evaluation data from the evaluation history database for each evaluation index for the product or service related to the survey target acquired by the product etc. information acquisition unit, or the business operator, and
Based on the evaluation data extracted by the evaluation history information acquisition unit, rank data is calculated according to a predetermined procedure, and a statistically significant difference between the survey targets is determined by statistical calculation based on the rank data. A statistical calculator;
Along with information on the result of statistical calculation by the statistical calculation unit, based on information on statistically significant differences between the respective survey objects determined by the statistical calculation unit, selected from among predefined, A value evaluation support system, comprising: an evaluation result output unit that outputs recommended information related to a criterion for determination of a difference in value between the survey targets to the user terminal.
In the value evaluation support system according to claim 1,
The statistical calculation unit ranks the evaluation data for each evaluation index based on the evaluation data extracted by the evaluation history information acquisition unit, and ranks the evaluation data for the evaluation target. The first statistical processing for calculating the statistical data based on the first ranking data and determining a statistically significant difference between the respective survey targets for each of the evaluation indices can be executed.
The evaluation result output unit includes information on a statistical calculation result in the first statistical process in the statistical calculation unit, and a statistically significant difference between the respective survey targets determined by the first statistical process. A value evaluation support system, wherein the recommended information is selected based on information.
In the value evaluation support system according to claim 2,
In the first statistical processing executed by the statistical calculation unit of the value evaluation server, when the number of survey targets is 2, a statistically significant difference is determined by Wilcoxon test, and the number of survey targets is 3 In this case, the value evaluation support system is characterized in that a statistically significant difference is determined by Kruskal-Wallis test.
In the value evaluation support system according to claim 2,
The statistical calculation unit of the value evaluation server is further determined to have a significant difference when there are a plurality of the evaluation indexes determined to have a significant difference between the survey targets by the first statistical processing. For each of the evaluation indexes including the plurality of evaluation indexes, an evaluation data statistical value that is a statistical value obtained from the evaluation data for each survey target, or the evaluation data statistical value for each evaluation index between the survey targets It is possible to execute second statistical processing for ranking the differences, calculating second ranking data, and determining a statistically significant difference between the respective survey objects by statistical calculation based on the second ranking data. Yes,
The evaluation result output unit includes information on a result of statistical calculation in the first and second statistical processing in the statistical calculation unit, and each of the investigation objects determined by the first and second statistical processing. A value evaluation support system, wherein the recommended information is selected based on statistically significant difference information.
In the value evaluation support system according to claim 4,
In the second statistical processing executed by the statistical calculation unit of the value evaluation server, when the number of survey targets is 2, a statistically significant difference is determined by Wilcoxon signed rank test, and the survey targets When the number of is 3 or more, the value evaluation support system characterized by determining a statistically significant difference by Friedman test.
In the value evaluation support system according to claim 4,
The statistical calculation unit of the value evaluation server further performs three or more investigations by statistical calculation based on the second rank data obtained by ranking the evaluation data statistical values in the second statistical processing. When it is determined that there is a significant difference between the subjects, the difference between the evaluation data statistical values for each of the evaluation indexes between any two of the survey subjects determined to have a significant difference is ranked third. And the third statistical processing for determining a statistically significant difference between the respective survey objects can be executed by statistical calculation based on the third ranking data.
The evaluation result output unit includes information on a result of statistical calculation in the first to third statistical processes in the statistical calculation unit, and between each of the investigation objects determined by the first to third statistical processes. A value evaluation support system, wherein the recommended information is selected based on statistically significant information.
In the value evaluation support system according to claim 6,
In the third statistical processing executed by the statistical calculation unit of the value evaluation server, a statistically significant difference is determined by Wilcoxon signed rank test.
The value evaluation support system according to any one of claims 1 to 7,
The evaluation result output unit of the value evaluation server, when selecting the recommended information, inquires the user about information indicating the importance of each evaluation index via the user terminal, The recommended information is selected based on information indicating the importance of each evaluation index designated by the user and information on statistical significance between the survey targets determined by the statistical calculation unit. A value evaluation support system characterized by this.
In the value evaluation support system according to claim 1,
The evaluation history information acquisition unit of the value evaluation server includes two survey target groups each composed of one or more survey targets, which are selected based on conditions specified by the user via the user terminal. Extracting the evaluation data relating to the evaluation index designated by the user from the evaluation history database for the combination of each survey object during
The statistical calculation unit, for each of the survey targets of one of the two survey target groups, for each of the survey targets of the other survey target group based on the evaluation data extracted by the evaluation history information acquisition unit The fourth ranking data is calculated by ranking the evaluation data for each combination of the survey objects or the evaluation data statistical value which is a statistical value obtained from the evaluation data, and is based on the fourth ranking data. By statistical calculation, it is possible to execute a fourth statistical process for determining a statistically significant difference between the survey targets,
The evaluation result output unit includes information on a result of statistical calculation in the fourth statistical process in the statistical calculation unit, and a statistically significant difference between the respective survey targets determined by the fourth statistical process. Select the recommended information based on the information,
The evaluation information input unit receives the input of the evaluation data relating to the combination of the investigation targets, and accumulates the evaluation data in an evaluation history database.
In the value evaluation support system according to claim 9,
In the fourth statistical process executed by the statistical calculation unit of the value evaluation server, the number of the survey target groups related to the evaluation of the value is 2, and the number of each survey target in the one survey target group Is 2, statistically significant difference is determined by Wilcoxon signed rank test, the number of the survey target groups related to the evaluation of the value is 2, and each survey target in the one survey target group is Is 3 or more, the statistical significance is determined by Friedman's test, the number of the survey target groups related to the evaluation of the value is 1, and the number of each survey target in the survey target group is 2 In the case of, the statistically significant difference is determined by the Wilcoxon test, and the number of the survey target groups related to the evaluation of the value is 1, Value assessment support system, wherein the number of surveyed if 3 or more, is to determine the statistically significant difference by Kruskal-Wallis test.
In the value evaluation support system according to claim 9,
The statistical calculation unit of the value evaluation server further includes a statistic based on the fourth rank data obtained by ranking the evaluation data or a statistical value obtained from the evaluation data in the fourth statistical process. When it is determined by calculation that there is a significant difference in the survey target group having three or more survey targets, between any two of the survey targets of the survey target group determined to have a significant difference The fifth rank data is calculated by ranking the evaluation data for each evaluation index or the difference in statistical values obtained from the evaluation data, and the respective surveys are calculated by statistical calculation based on the fifth rank data. A fifth statistical process for determining a statistically significant difference between subjects can be performed;
The evaluation result output unit includes information on a result of statistical calculation in the fourth and fifth statistical processing in the statistical calculation unit, and each of the investigation objects determined by the fourth and fifth statistical processing. A value evaluation support system, wherein the recommended information is selected based on statistically significant difference information.
In the value evaluation support system according to claim 11,
In the fifth statistical process executed by the statistical calculation unit of the value evaluation server, when the number of the investigation target groups related to the value evaluation is 2, a statistically significant difference is determined by Wilcoxon signed rank test. A value evaluation support system characterized in that, when the number of the investigation target groups related to the evaluation of the value is 1, a statistically significant difference is determined by a Wilcoxon test.
The value evaluation support system according to any one of claims 9 to 12,
The evaluation result output unit of the value evaluation server, when selecting the recommended information, inquires the user for information indicating the importance of each survey target group via the user terminal, Based on the information indicating the importance of each survey target group specified by the user and the statistically significant difference information between the survey targets determined by the statistical calculation unit, the recommended information A value evaluation support system characterized by selection.
The value evaluation support system according to any one of claims 9 to 13,
The statistical calculation unit of the value evaluation server has the missing data for each survey target when there is missing data for which the evaluation data statistical value does not exist for each combination of the survey targets between the two survey target groups. Based on the number of data and the priority information for each of the survey targets in the two survey target groups specified by the user via the user terminal, the survey target having the missing data is When the deletion data is deleted from the combination between the survey targets and the missing data disappears in the combination between the survey targets, the fourth rank data for the evaluation data statistical value is calculated, and When all the survey targets are deleted in the combination of the above, the fourth and fifth statistical calculation processes are stopped. Value evaluation support system that.
The value evaluation support system according to any one of claims 1 to 14,
The statistical calculation unit of the value evaluation server, from each evaluation data extracted by the evaluation history information acquisition unit, from a predetermined number that is less than the number of each evaluation data and statistically recognized as a population A value evaluation support system characterized in that statistical calculation is performed by randomly extracting a number of samples that are less than an upper limit value and larger than a lower limit value including a predetermined number that is statistically inferior in accuracy.
The value evaluation support system according to any one of claims 1 to 14,
The statistical calculation unit of the value evaluation server extracts, from each of the evaluation data extracted by the evaluation history information acquisition unit, a number of first samples that is greater than a predetermined lower limit value and less than or equal to a predetermined upper limit value. Furthermore, the value evaluation support system characterized by randomly extracting from the first sample a number of second samples that are greater than or equal to the lower limit value and less than the number of the first samples.
The value evaluation support system according to any one of claims 1 to 16,
When the statistical calculation unit of the value evaluation server determines a statistically significant difference between the survey targets, if there is no numerical value corresponding to the test table to be referred to, the statistics between the survey targets A value evaluation support system characterized by determining that there is no significant difference.
The value evaluation support system according to any one of claims 1 to 17,
The evaluation result output unit of the value evaluation server is a combination of the property of the evaluation data used for the statistical calculation by the statistical calculation unit and the statistical calculation information which is the value and condition used for the statistical calculation, or statistical result information. A value evaluation support system comprising: selecting a supplementary explanation for a result of statistical calculation selected from predefined ones based on the result and outputting the selected supplementary explanation to the user terminal.
A value evaluation server that determines a statistically significant difference in a value difference between the survey objects and outputs a determination result for a plurality of survey objects whose values are represented by one or more evaluation indexes including ambiguous data And a value evaluation support system having a user terminal connected to the value evaluation server via a network,
The value evaluation server
An evaluation information input unit that receives input of evaluation data that is an evaluation result of value for each evaluation index related to each of the survey targets, and accumulates the evaluation data in an evaluation history database;
An evaluation history information acquisition unit that extracts the evaluation data from the evaluation history database for each of the plurality of investigation targets selected based on the conditions specified by the user via the user terminal. When,
Based on the evaluation data extracted by the evaluation history information acquisition unit, rank data is calculated according to a predetermined procedure, and a statistically significant difference between the survey targets is determined by statistical calculation based on the rank data. A statistical calculator;
Along with information on the result of statistical calculation by the statistical calculation unit, based on information on statistically significant differences between the respective survey objects determined by the statistical calculation unit, selected from among predefined, A value evaluation support system, comprising: an evaluation result output unit that outputs recommended information related to a criterion for determination of a difference in value between the survey targets to the user terminal.
For a plurality of survey targets whose values are represented by one or more evaluation indexes including ambiguous data, a statistically significant difference in the difference in value between the survey targets is determined, and the determination result is transmitted to the user terminal. A value evaluation support program that allows a computer to function as a value evaluation server
An evaluation information input process that receives input of evaluation data that is an evaluation result of the value for each evaluation index related to each of the survey targets, and accumulates the evaluation data in an evaluation history database;
An evaluation history information acquisition process for extracting the evaluation data from the evaluation history database for each of the evaluation indexes for the plurality of survey targets selected based on conditions specified by the user via the user terminal; ,
Based on the evaluation data extracted by the evaluation history information acquisition process, rank data is calculated according to a predetermined procedure, and a statistically significant difference between the survey targets is determined by statistical calculation based on the rank data. Statistical calculation processing,
Along with the information on the result of statistical calculation by the statistical calculation process, based on the information on the statistically significant difference between the respective survey objects determined by the statistical calculation process, selected from the predefined, A value evaluation support program, which executes an evaluation result output process for outputting recommended information relating to a criterion for determination of a difference in value between each survey target to the user terminal.