JP2002163440A - Correlated item detection method and device - Google Patents

Abstract

(57) [Summary] [Problem] To speed up the detection of correlated items by extracting only correlations between items having different item attributes (genres) using an access history. SOLUTION: For each item identifier of an analysis source item list which is a detection source item group, a customer identifier list of a customer accessing the item is extracted from an access history to be a related customer list. For each customer identifier in the related customer list, the item identifier of the item being accessed by the customer is extracted from the access history and set as a related item list. Then, the item attribute of each item identifier of the analysis source item list is compared with the item attribute of each item identifier of the related item list, and an item identifier having a different item attribute is extracted to be a correlated item list.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

TECHNICAL FIELD The present invention relates to the Internet,
In a system for browsing and purchasing items via a network in a digital interactive television, an information home appliance, and the like, and a database system for managing sales and the like of customers-products (items), browsing and selling histories of items, and the like. And an apparatus for detecting a correlated item from an access history of a computer.

[0002]

2. Description of the Related Art As a conventional correlation detection method, there is a data mining method. It is well known that the result of analyzing the sales data of products shows that the correlation between "diaper" and "beer" is strong, and that selling both in the same space has resulted in an increase in both sales. Conventionally, correlation detection using a statistical method based on a correlation coefficient is also generally known.

[0003]

In these conventional methods, when handling a large amount of data, it is necessary to apply an extraction, pre-processing, and data reduction method by an expert. It often took about a month. In addition, as for the computing capacity, a supercomputer is used for data mining in some cases, and a strong computing capacity is required.

[0004] It is an object of the present invention to provide a correlation item detection method and apparatus which eliminates the need for pre-examination, pre-processing, etc. by a dedicated house as in the conventional method, simplifies correlation analysis processing, and enables high-speed calculation. It is in.

[0005]

According to the present invention, a correlated item is detected from an access history such as an item browse and a sales history. At this time, it is noted that heuristic correlation detection, which has not conventionally been noticed in correlation detection, is particularly important, and only correlations between items having different genres (item attributes) are extracted.

FIG. 1 shows a processing image of a correlation item detection method according to the present invention. Now, it is assumed that an analysis source item list serving as a source of a correlation item list to be detected is given. First, for each item (item identifier) in the analysis source item list, a list of customers (customer identifiers) accessing the item is extracted and set as a related customer list. Next, for each customer in the related customer list, a list of items being accessed by the customer is extracted and used as a related item list. And the item attribute of each item in the analysis source item list (analysis source item attribute)
And the item attributes of each item in the related item list are compared, and a list of items having different item attributes is extracted to be a correlated item list. In addition, it is also possible to add an item evaluation value as a detection condition of a correlated item, and extract an item having an item attribute different from the analysis source item attribute and having an item evaluation value equal to or more than a predetermined value from the related item list. I do.

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below in detail with reference to the drawings. FIG. 2 is a system configuration diagram of one embodiment of the present invention. In the figure, 100 is a client terminal, 200 is a network, 300 is an item provider server, and 400 is a correlated item detection device according to the present invention. The connection form between the item provider server 300 and the correlation item detection device 400 is arbitrary,
For example, they may be connected by a dedicated line, or may be connected via the network 200 similarly to the client terminal 100.

The client terminal 100 is a Web-accessible device such as a personal computer, a portable terminal, a two-way digital TV, an information home appliance, and the like. Customer (user) is client terminal 1
00, the item provider server 300 is accessed via the network 200,
Make purchases. The item is a generic term for advertisements, information, titles and outlines of CDs, books, movies, and the like, and songs (music information).

[0009] The network 200 is not limited to the Internet. It may be a digital interactive TV or an information home appliance network.

[0010] The item provider server 300 is generally operated by a company (provider) that implements a business of providing items for browsing and purchasing to a plurality of customers. This type of company integrates, for example, an EC site that introduces and sells items such as CDs and books, a search and directory site that introduces and searches for items such as Web contents and information, and a plurality of EC sites. There are companies that operate electronic mall sites that introduce and sell items such as products. Not only an electronic company via a network but also a company such as a mail-order company or a department store that electronically stores a customer's product purchase history. Here, all servers operated by such a company are referred to as item provider servers.

The correlation item detection device 400 includes an input / output unit 410, a correlation item analysis processing unit 420, an access history processing unit 430, an access history map management unit 440, an item attribute storage unit 450, an analysis list storage unit 460, and the like. You. In addition, a control unit for controlling the overall operation of the correlation item detection device 400, a communication interface for transmitting / receiving information to / from the item provider server, and the like are provided, but are omitted in FIG.

The input / output unit 410 is a general term for a keyboard, a mouse, a display, a media driver, and the like, and is used by an operator of the correlation item detection device 400.

The correlation item analysis processing unit 420 is a main processing unit for performing a process according to the present invention for detecting a correlated item from an access history such as an item browsing / sales history. The correlation item analysis processing unit 420
Read data from the access history map management unit 440 and the item attribute storage unit 450,
It is possible to write and read data to and from 60.

The access history processing section 430 processes the access history transmitted from the item provider server 300 to generate access history map data which can be easily used by the correlation item analysis processing section 420. The access history is data that records the history of actions taken by a customer to browse or purchase an item. One data is composed of, for example, “customer identifier, item identifier, purchase price” and the like.

The access history map management section 440 holds the access history map data 442 generated by the access history processing section 430. FIG. 3 shows an embodiment of the access history map data 442. As shown in FIG. 3, the access history map data 442 includes map data of an item identifier for each customer identifier (specifically, an item identifier,
Item evaluation value, item attribute map) 442-1,
And map data 4 of the customer identifier for each item identifier
42-2.

The item attribute storage section 450 holds item attribute data 452 transmitted from the item provider server 300. FIG. 4 shows item attribute data 452.
An example will be described. The item attribute data 452 is shown in FIG.
As shown in (1), it is composed of attribute information (item evaluation value, item attribute) for each item identifier.

The analysis list storage unit 460 holds and manages an analysis source item attribute 462, an analysis source item list 464, a correlated customer list 466, and a correlated item list 468. The analysis source item attribute 462 records an item attribute serving as a source of the correlation item analysis. The analysis source item list 464 records an item group as a source of the correlation item analysis. The related customer list 466 records a customer group (related customer group) accessing the analysis source item list. The correlation (related) item list 468 records an item group accessed by a related customer group, each item attribute, and an item evaluation value.

Here, customer identifiers, item identifiers, item attributes, and item evaluation values shown in FIG. 3 and FIG. 4 are collectively defined.

The customer identifier is an identifier capable of uniquely identifying a customer in both the item provider server 300 and the correlated item detection device 400. For example, a customer number notified by each company, a customer number recorded on a card or the like, a member number manually entered by the customer, a browser C
There is a means for adding a time stamp or the like recorded using the cookie function.

The item identifier is an identifier capable of uniquely identifying an item in both the item provider server 300 and the correlated item detection device 400. For example, when each page of the Web page is an item, the URL is the item identifier.

Item attributes (item attributes) are, for example, product types and the like. Large items are classified into "clothing", "food", "books", and the like, and medium items are "blouses" and "trousers". Etc., and the small items are attribute values such as "long sleeve blouse" and "business slacks". The attribute expression format is, for example, 6
Each attribute is represented by four bits.

An item evaluation value (item evaluation value) is a value serving as a measure for determining the value of an item, and is represented by a total purchase price, a total number of accesses, or a composite value thereof during the observation period of the item. is there. In the present embodiment, it is used to select significant items to be extracted as an item group as a source of a correlation item analysis and as an item group as a correlation result.

Next, the operation of the correlation item detecting device 400 according to the present invention in the system of FIG. 2 will be described.
The operation is roughly classified into recording of item attributes, generation of access history map data, and detection of correlated items.

[Recording of Item Attributes] FIG. 5 shows a processing flow of item attribute recording. The item provider server 300 sends the item attribute and item evaluation value for each item identifier to the correlation item analysis device 400 together with a list of item identifiers to be recorded together with the item attribute storage request when the item provider periodically or when analysis is desired. Send. Correlation item analysis device 400 transmits an item attribute storage request, a list of item identifiers, an item attribute,
An item evaluation value is received (step 11).

Usually, the item provider server 300
Sends an item attribute recording request for all items managed by itself or all items to be analyzed. When transmission via a network or the like is difficult due to a large amount of information to be recorded, the information may be artificially transferred using a recording medium or the like.

As shown in FIG. 4, the correlated item analyzer 400 that has received the item attribute recording request stores the item attribute and the item evaluation value for each item identifier in the received item identifier list.
It is recorded and managed in 50 item attribute data 452 (step 12).

[Generation of Access History Map] FIG. 6 shows a processing flow of generating an access history map. The access history in which the customer accesses the item provider server 300, browses the posted information, purchases a product, and the like is stored in the item provider server 300. In addition, the same process may be performed on merchandise sales data in a normal mail-order company or the like, or merchandise sales data in a normal store.

The item provider server 300 transmits the stored access history to the correlated item detection device periodically or collectively when analysis is desired. The access history is composed of, for example, a list of “customer identifier, item identifier, purchase fee”. The correlation item detection device 400 receives the access history transmitted from the item provider server 300, and
30 (step 21).

The access history processing unit 430 converts the access history into a map format that can be easily used in the correlation item detection processing by the correlation item analysis processing unit 420 (step 22). That is, an access history map is generated.
As shown in FIG. 3, the access history map includes, for example, a list 442-1 of item identifiers, item attributes, and item evaluation values accessed by each customer for each customer identifier,
And a list 442-2 of customer identifiers accessing the item for each item identifier. When generating an access history map, the item attribute and the item evaluation value of the item attribute data 452 read from the item attribute storage unit 450 for each item identifier are used as the item attribute and the item evaluation value. The item evaluation value of the list 442-1 indicates the total purchase amount, the total number of accesses, or a composite value thereof during the observation period of the item.

Next, the access history processing unit 430 sends the access history map together with the access history map storage request to the access history map management unit 440 to record the generated access history map. It is stored and managed in the access history map data 442 (step 23).

[Detection of Correlated Item] FIGS. 7 and 8 show the processing flow of correlated item detection. The item provider server 300 wants to detect the item attribute that is the source of the correlation item analysis together with the correlation item detection request, the lower limit value of the item evaluation value of the analysis source item list to be extracted, and the correlation item detection request when the analysis is desired periodically or. The lower limit value of the item evaluation value of the correlation item list is transmitted to the correlation item detection device 400.

The correlated item detection device 400 includes a correlated item detection request, an analysis source item attribute, a lower limit of the item evaluation value of the analysis source item list, and a lower limit of the item evaluation value of the correlation item list transmitted from the item provider server 300. And the correlation item analysis processing unit 4
20 (step 31).

The correlation item analysis processing unit 420 that has received the correlation item detection request first has the same item attribute as the analysis source item attribute from the item provider server 300 and has the item evaluation value of the analysis source item list to be extracted. A list of item identifiers having values equal to or greater than the lower limit is obtained from the item attribute data 452 of the item attribute storage unit 450 (step 32). Then, the correlation item analysis processing unit 420 records the item attribute received from the item provider server 300 in the analysis source item attribute 462 of the analysis list storage unit 460, and stores the attribute data 4 in the item item attribute storage unit 450.
The list of the item identifiers acquired from 52 is recorded in the analysis source item list 464 of the analysis list recording unit 460 (step 33).

When the list of item identifiers to be processed is explicitly given, the processing of steps 32 and 33 can be omitted.

Next, the correlation item analysis processing unit 420
For each item identifier in the analysis source item list 464 recorded in the analysis list storage unit 460, a list of customer identifiers accessing the item identifier is stored in the access history map data 442 of the access history map management unit 440.
(Step 34). Specifically, extraction is performed using the customer identifier map data 442-2 (FIG. 3) for each item identifier in the access history map data 442. The extracted list of customer identifiers is referred to as a related customer group. The correlation item analysis processing unit 420 records this related customer group in the related customer list 466 of the analysis list recording unit 460 (Step 35).

Next, the correlation item analysis processing section 420
Step 7 if necessary, such as when the number of related customers is large
The process of narrowing down the related customer group extracted in step 04 is performed (step 36). As an example for narrowing down, the related customer groups are sorted in the order of the number of accessing each item stored in the analysis source item list 464 of the analysis list recording unit 460 (the order of the overlapping). , Top NN
g2 is a new related customer group. As conditions for narrowing down, in addition to the number of accesses, an access time, an access interval, a combination thereof, and the like can be considered. The narrowed down related customer group is recorded again in the related customer list 466 of the analysis list recording unit 460.

Next, the correlation item analysis processing section 420
For each customer identifier recorded in the related customer list 466 of the analysis list recording unit 460, the access history map data of the access history map management unit 440 includes a list of item identifiers accessed by the customer along with item attributes and item evaluation values. 442 (Step 3
7). Specifically, the map data 442 of the item identifier for each customer identifier in the access history map data 442
-1 (FIG. 3). The list of the extracted item identifiers is called a related item group. The correlation item analysis processing unit 420 records this related item group, its item attributes, and item evaluation values in the correlation item list 468 of the analysis list storage unit 460 (step 38).

Next, the correlation item analysis processing section 420
A narrowing-down process is performed based on the attribute value and the evaluation value of the related item group (step 39). Specifically, the correlation item analysis processing unit 420 determines the analysis source item attribute 462 of the analysis list storage unit 460 and the item identifier of each related item group recorded in the correlation item list 468 of the analysis list storage unit 460. By comparing the item attributes, a list of item identifiers having different item attributes and having an item evaluation value equal to or more than the lower limit of the item evaluation value of the correlated item list received from the item provider server 300 is extracted, and a new association is obtained. Item group. The correlation item analysis processing unit 420 newly records the related item group, the item attribute of each item, and the item evaluation value in the correlation item list 468 of the analysis list storage unit 460.

In general, since an item matching the item attribute of the analysis source item is easily extracted as a correlated item, a considerable amount of items are automatically deleted by the processing of step 39, and the subsequent processing is performed at high speed. It becomes possible. Note that the narrowing-down condition in step 39 may be only the item attribute and may not consider the item evaluation value.

Next, the correlation item analysis processing unit 420
Further narrowing processing of the related item group extracted in step 39 is performed (step 40). As an example for narrowing down, the related item group is rearranged in the order of the number of accesses by the customers in the related customer list 466 of the analysis list storage unit 460 (the order of the overlap), and the upper NN
Let g3 be a new related item group. Also in this case, as the narrowing-down condition, in addition to the number of accesses, an access time, an access interval, a combination thereof, and the like can be considered. The correlation item analysis processing unit 420 re-uses the narrowed-down related item group, the item attributes thereof, and the item evaluation value as a final correlation item detection result, and re-examines the correlation item list 468 of the analysis list storage unit 460.
To record.

Finally, the correlation item detection device 400
Both the correlation item detection response and the analysis list storage unit 460
The list of item identifiers, which are the correlation item detection results recorded in the correlation item list 468, and the item attributes and item evaluation values of each item are sent to the item provider server 300 that is the source of the correlation item detection (step 41). .

The item provider server 300 holds the correlation item result sent from the correlation item detection device 400. For example, a customer accesses the item provider server 300 and browses information or purchases a product. In this case, relevant information and products are provided to the customer.

When the analysis source item list is enormous, for example, the analysis source item list 464 of the analysis list storage unit 460 is set by dividing the analysis source item list into 100 items. It may be repeated. The result at that time is to merge each detection result and return it to the item provider server 300 as a correlated item detection response.

Although the embodiment of the present invention has been described above, the present invention is not limited to the Internet via a provider server, but also manages, for example, customer-product sales in the existing product sales. The present invention can be implemented for a database to be executed, and can also be implemented for sale of a product via a network in a digital interactive television, an information home appliance, or the like.

[0045]

As described above, according to the present invention,
Since only the access history is used as the base information for the correlation item analysis, no pre-examination and pre-processing for extracting the base information is required. In addition, by making the group of items accessed by the related customers who are accessing the group of detection source items the group of correlation items, pre-examination and pre-processing such as extraction of statistical parameters for each applied business or product handled can be performed. Not only is it unnecessary, but it can greatly simplify the analysis process and perform high-speed calculations. further,
Normally, it is difficult to speed up the processing because many of the items detected to be correlated belong to the same genre, but in the present invention, in the present invention, attributes having different attributes (different genres) from the original item are used. By extracting correlation items, heuristic correlation item detection can be performed at high speed.

[Brief description of the drawings]

FIG. 1 is a diagram showing a processing image of correlated item detection of the present invention.

FIG. 2 is a diagram showing an embodiment of an application system of the correlation item detection device according to the present invention.

FIG. 3 is a diagram showing one embodiment of access history map data.

FIG. 4 is a diagram showing an embodiment of item attribute data.

FIG. 5 is a processing flowchart of item attribute recording.

FIG. 6 is a processing flowchart of generating an access history map.

FIG. 7 is a processing flowchart of an embodiment of detecting a correlated item.

FIG. 8 is a processing flowchart continued from FIG. 7;

[Explanation of symbols]

 Reference Signs List 100 client terminal 200 network 300 item provider server 400 correlated item detecting device 410 input / output unit 420 correlated item analysis processing unit 430 access history processing unit 440 access history map management unit 442 access history map data 450 item attribute storage unit 452 item attribute data 460 Analysis list storage unit 462 Analysis source item attribute 464 Analysis source item list 466 Related customer list 468 Correlation item list

Claims (7)

[Claims] 1. An item (hereinafter referred to as an analysis source item) using an access history of a customer accessing an item.
Is a method of detecting an item (hereinafter, a correlation item) having a correlation with a customer, extracting a customer (hereinafter, a related customer group) accessing an analysis source item, and an item group being accessed by the related customer group. (Hereinafter referred to as a related item group), and extracting an item having an item attribute different from the item attribute of the analysis source item from the related item group to obtain a correlated item. 2. The correlated item detection method according to claim 1, wherein an item having an item attribute different from the item attribute of the analysis source item and having an item evaluation value equal to or greater than a predetermined value is extracted from the related item group. A correlated item detection method. 3. The correlation item detection method according to claim 1, wherein the number of customers in the related customer group is narrowed down according to the access status of the customer. 4. The correlation item detection method according to claim 1, wherein the extracted correlation items are narrowed down according to the access status of each customer of the related customer group. 5. The correlated item detection method according to claim 1, wherein an item group having an item attribute serving as a source of correlated item analysis and having an item evaluation value equal to or greater than a predetermined value is defined as an analysis source item list. A correlation item for each item in the analysis source item list to obtain a correlation item list. 6. An item (hereinafter referred to as “analysis source item”) using an access history of a customer accessing the item.
A device that detects an item having a correlation with the item (hereinafter, a correlation item), an item attribute storage unit that stores an item attribute and an item evaluation value for each item, an access history management unit that manages an access history, and a correlated item. The item attribute that is the source of the analysis (hereinafter, analysis source item attribute), the lower limit of the item evaluation value of the analysis source item list that you want to extract (hereinafter, the first lower limit), and the lower limit of the item evaluation value of the correlation item list that you want to detect Means for inputting (hereinafter, a second lower limit) a list of items (hereinafter, an analysis source item list) which are item attributes of the analysis source item attribute and whose item evaluation value is equal to or greater than the first lower limit. Means for extracting from the item attribute storage means, and for each item in the analysis source item list, accessing the item For a list of the customer (hereinafter referred to as related customer group)
From the access history of the access history management means, and for each customer of the related customer group, a list of items accessed by the customer (hereinafter, related item group) is stored in the access history of the access history management means. Means for extracting from the analysis source item attribute, and comparing the item attribute of each item of the related item group with different item attributes,
And a means for extracting a list of items whose item evaluation values are equal to or greater than the second lower limit as a correlated item list. 7. The correlated item detecting device according to claim 6, wherein the item attribute and the item evaluation value of each item are received from the item provider server together with the item list and held in the item attribute storage means, and accessed from the item provider server. Receiving the history, generating the access history of the item for each customer and the access history of the customer for each item and storing them in the access history management means, extracting the attribute of the analysis source item together with the correlation item detection request from the item provider server Receiving the lower limit of the item evaluation value of the source item list to be analyzed, the lower limit of the item evaluation value of the correlation item list to be detected, and returning the detected correlation item list to the item provider server as a response to the correlation item detection request. Features Correlation item detection device.