JP2005038140A - Product sales number forecasting device, forecasting method and program therefor - Google Patents

Abstract

An object of the present invention is to predict the number of new products sold with high accuracy using a small amount of data, and to realize an efficient sales number prediction technique as a whole.
A transition pattern selection unit 52 determines whether a second predetermined period of time has elapsed for a new product from the model formula stored in the model formula storage unit 78a and the actual sales number stored in the actual sales number storage unit 72a. The cumulative predicted sales number until the first predetermined period is calculated by dividing the actual sales number for each unit period from the sales start time to the time when the first predetermined period has elapsed, and stored in the transition storage means 76a. The transition pattern that is most similar to the percentage transition is selected from the standard transitions that have been made. The predicted sales number calculation means 54 calculates from the standard transition of the selected transition pattern and the cumulative predicted sales number of the new product until the second predetermined period elapses, from the first predetermined period elapses to the second predetermined period elapses. Calculate the estimated number of sales by unit period.
[Selection] Figure 2

Description

[0001]
BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a technique for predicting the number of new products sold. In particular, the present invention relates to a technology for predicting the number of sales for each unit period from when the first predetermined period has elapsed since the start of sales of a new product until when the second predetermined period has elapsed.
[0002]
The main terms used in this specification will be explained.
The unit period is a period that is a unit for predicting the number of new products sold. For example, for a product with a short fluctuation cycle of the number of sales, one hour, one day, one week, etc. correspond to the unit period. Alternatively, if the product has a long fluctuation cycle of sales, January, March, half year, one year, etc. correspond to the unit period.
The first predetermined period is a period in which data indicating the transition of the actual sales number of new products sufficient to be used for predicting the sales number of new products for which sales have started is accumulated. The first predetermined period includes a plurality of unit periods. For example, if there is transitional data on the number of actual sales for a two-unit period from the start of sales of a new product, it is sufficient to predict the future number of sales of the new product. Become. Alternatively, if there is data on the actual number of sales for a four-unit period from the start of sales of a new product, it is sufficient to predict the future number of sales of the new product, and the first predetermined period will be a four-unit period. .
The second predetermined period is a period for which the number of sales is predicted after the first predetermined period elapses. For example, when the unit period is one week and the first predetermined period is four weeks, the second predetermined period is 48. May be a week. In this case, the time when the first predetermined period and the second predetermined period have elapsed since the sales start time is the time when 52 weeks have elapsed since the sales start time, and is the time when almost one year has passed since the sales start time. In this specification, this time point may be a time point when the second predetermined period has elapsed. The unit period may be 1 month, the first predetermined period may be 3 months, and the second predetermined period may be 9 months.
Further, not only the number of sales but also the number of orders can be predicted in the same way when the product is a car or the like, for example. In that case, it is also possible to use data before the start of sales, such as reservation order data. Therefore, the first predetermined period may be, for example, from 4 weeks before the sales start time to the sales start time.
[0003]
2. Description of the Related Art Accurately predicting the number of new products sold is extremely important in determining sales strategies and planning production plans. A technique for predicting the number of new products sold is disclosed in Patent Document 1. With this technology, the number of new products that have not yet been sold is predicted with reference to the sales performance of existing products that are actually sold. In this prediction technology, information on the characteristics of the product (vehicle) (vehicle type, specifications, specifications, color, etc.) and information obtained from the purchaser of the product (buyers such as gender, age, occupation, hobby, etc., purchase Motivation, evaluation of products, etc.), the products and purchasers are aggregated from their similarities, and the strength of the preference between the sets is defined as the purchase realization rate. The number of sales is forecasted. In this case, factors other than merchandise such as advertising and sales force are also taken into consideration.
Patent Document 2 also discloses a technique for predicting the number of new products sold by using information about the characteristics of a product and information obtained from a purchaser of the product.
[0004]
[Patent Document 1]
JP 2002-99706 A
[Patent Document 2]
Japanese Patent Laid-Open No. 9-120395
[0005]
Recently, in response to the diversification of consumer needs, companies often produce a wide variety of products. In the above technique, when the number of types of products is large, the amount of information related to the characteristics of the products to be used for the sales number prediction and the information obtained from the purchaser of the products becomes enormous. For this reason, the calculation amount and calculation time required for predicting the number of new products to be sold increase.
In particular, collecting information related to the purchaser takes a lot of labor, time, cost, etc., and is not efficient.
[0006]
The present invention predicts the number of subsequent sales of a new product from the transition of the actual sales number of the existing product and the transition of the actual sales number from when the new product starts to be sold until a predetermined period elapses. Use a small amount of data to predict the number of new products sold with high accuracy.
This reduces the amount of calculation and calculation time required to predict the number of new products sold, and reduces the labor, time, and cost required to collect information, and realizes overall efficient sales number prediction technology. To do.
[0007]
According to the present invention, when a first predetermined period elapses from the start of sales of a new product, the unit period by the second predetermined period elapses from that point. A device to predict the number of sales was created.
The apparatus includes a sales number storage means for storing a transition of a sales volume of a new product for each unit period divided into a plurality of first predetermined periods, a first predetermined period and a first sales period from the start of sales of existing products. 2. Standard transition storage means by transition pattern that stores standard transitions by transition pattern of values obtained by dividing actual sales by unit period by the cumulative actual sales until the predetermined period has passed; And a model formula storage unit that stores a model formula for calculating the cumulative actual sales number up to the second predetermined period elapses from the cumulative actual sales number up to the elapse of the predetermined period.
Further, from the model formula stored in the model formula storage means and the cumulative actual sales number stored until the first predetermined period elapsed stored in the actual sales number storage means, until the second predetermined period elapse of the new product. A cumulative predicted sales number calculating means for calculating the cumulative predicted sales number is provided.
Also, a transition calculation that calculates the transition of the value obtained by dividing the actual sales by unit period from the start of sales of the new product to the time when the first predetermined period has elapsed by the cumulative predicted sales calculated by the cumulative predicted sales calculation means. Means.
In addition, a transition pattern selection unit that selects a transition pattern most similar to the transition calculated by the transition calculation unit among the standard transitions by transition pattern stored in the standard transition storage unit by transition pattern is provided.
Further, from the standard transition of the transition pattern selected by the transition pattern selection means and the cumulative predicted sales number of the new product calculated by the cumulative predicted sales number calculation means until the second predetermined period has elapsed, To a predicted sales number calculating means for calculating a predicted sales number for each unit period from the time point until the second predetermined period elapses.
[0008]
The product sales number prediction device predicts the number of sales for each unit period from when the first predetermined period elapses from the start of sales of a new product until the second predetermined period elapses from that point. For example, when it is desired to predict the number of sales for one year (about 52 weeks) from the start of sales of a new product, when 4 weeks (first predetermined period) have elapsed from the start of sales of the new product, The number of sales by week (by unit period) for the remaining 48 weeks (second predetermined period) is predicted. Predict by using the transition of actual sales of new products and the transition of actual sales of existing products.
The actual sales number storage means stores the transition of the actual sales number for each unit period from when the new product starts to be sold until the first predetermined period elapses. For example, a transition is stored in which the actual sales number in the first week is 10, the actual sales number in the second week is 9, the actual sales number in the third week is 11, and the actual sales number in the fourth week is 10.
Further, the transition pattern-specific standard transition storage means stores standard transitions obtained from the actual sales numbers of existing products for each transition pattern. Here, the transition is a value obtained by dividing the number of actual sales by unit period by the cumulative number of sales from the start of sales of existing products to the time when the first predetermined period and the second predetermined period have elapsed, for example, 52 weeks The transition of the ratio of the number of actual sales per unit period when the cumulative number of actual sales is assumed to be 100%. For example, the cumulative actual sales number of a certain existing product for 52 weeks is 100, the actual sales number in the first week is 10, the actual sales number in the second week is 5, the actual sales number in the third week is 3, and so on. Then, the ratio of the first week is 10%, the ratio of the second week is 5%, the ratio of the third week is 3%, and so on. By using existing methods such as cluster analysis, the transition of each existing product can be classified into patterns. For example, a “stable pattern” in which a certain ratio continues from the beginning to the end, an “attenuation pattern” in which the ratio is high at the beginning and then decays, or the ratio is low at the beginning but then increases. It can be classified into “increase patterns”. The standard transition is a standard transition of each transition pattern, and can be obtained, for example, by taking an average value of transitions of existing products belonging to the same transition pattern.
The model formula storage means stores a model formula generated based on the actual number of sales of existing products. The model formula is an expression for calculating the cumulative actual sales number of new products until the second predetermined period elapses from the cumulative actual sales number until the first predetermined period elapses. For example, as shown in FIG. 9 (1), the cumulative number of sales from the start of sales of existing products to the elapse of the first predetermined period (four weeks) and the second predetermined period (52 from the start of sales of existing products). It is possible to create a correlation graph by plotting the cumulative number of sold sales up to the point of elapse of the week in two-dimensional coordinates, and estimate the approximate line based on the primary expression or the secondary expression from the correlation graph to generate the model expression. In this way, the model formula can be generated by a simple existing method.
[0009]
The cumulative predicted sales number calculation means of the present product sales number prediction device is configured such that a cumulative number of actual sales until the first predetermined period of time stored in the actual sales number storage means is added to the model formula stored in the model formula storage means. Is substituted, and the cumulative number of predicted sales until the second predetermined period has elapsed for the new product is calculated.
The transition calculation means is a value obtained by dividing the actual sales number for each unit period from the sales start time of the new product to the time when the first predetermined period has elapsed by the cumulative prediction sales number calculated by the cumulative prediction sales number calculation means, that is, the first (1) The transition of the ratio of the number of actual sales in each unit period up to the point of the predetermined period is calculated.
The transition pattern selection unit selects the most similar “transition pattern” from the standard transitions by transition pattern stored in the transition pattern standard transition storage unit. For example, the transition of the number of actual sales by unit period calculated by the transition calculation means is 2% for the first week, 1.8% for the second week, 2.2% for the third week, If the ratio at the 4th week is 2%, the most similar “stable pattern” is selected.
The predicted sales number calculating means calculates the ratio of the actual sales number of each unit period in the standard transition of the transition pattern selected by the transition pattern selecting means to the second predetermined period of the new product calculated by the cumulative predicted sales number calculating means. Is multiplied by the cumulative predicted sales number until the time point elapses to calculate the predicted sales number for each unit period from the time point when the first predetermined period has elapsed to the time point when the second predetermined period has elapsed.
[0010]
According to the product sales number prediction apparatus, the sales number of new products is accurately predicted by using the transition of the actual sales number of the new product in the first predetermined period and the transition of the actual sales number of the existing product. be able to. Thereby, the burden of a prediction process can be reduced and it can contribute to efficiency improvement.
[0011]
In addition, another type of product sales number prediction device has been created by the present invention.
The apparatus includes a sales number storage means for storing a transition of a sales volume of a new product for each unit period divided into a plurality of first predetermined periods, a first predetermined period and a first sales period from the start of sales of existing products. 2. Standard transition storage means by transition pattern that stores standard transitions by transition pattern of values obtained by dividing actual sales by unit period by the cumulative actual sales until the predetermined period has passed; A transition pattern model expression storage unit that stores, for each transition pattern, a model expression for calculating an accumulated actual sales number from the cumulative actual sales number until the second predetermined period elapses.
In addition, the transition pattern assumption means for assuming the transition pattern of the new product, the model expression of the transition pattern assumed by the transition pattern assumption means stored in the model expression storage means for each transition pattern, and the actual sales number storage means Cumulative predicted sales number calculation means for calculating the cumulative predicted sales number of the new product until the second predetermined period elapses from the stored cumulative sales number until the first predetermined period elapses is stored.
Also, a transition calculation that calculates the transition of the value obtained by dividing the actual sales by unit period from the start of sales of the new product to the time when the first predetermined period has elapsed by the cumulative predicted sales calculated by the cumulative predicted sales calculation means. Means.
Also, a transition similarity calculation means for calculating the similarity of the transition pattern calculated by the transition calculation means, and the standard transition of the transition pattern assumed by the transition pattern assumption means, stored in the standard transition storage means for each transition pattern, Transition pattern selecting means for selecting a transition pattern having the highest similarity calculated by the transition similarity calculating means.
Further, from the standard transition of the transition pattern selected by the transition pattern selection means and the cumulative predicted sales number of the new product calculated by the cumulative predicted sales number calculation means until the second predetermined period has elapsed, To a predicted sales number calculating means for calculating a predicted sales number for each unit period from the time point until the second predetermined period elapses.
[0012]
The model expression storage means for each transition pattern of the product sales number prediction device calculates a model formula for calculating the cumulative actual sales number from the cumulative actual sales number until the second predetermined period elapses until the first predetermined period elapses of the existing product. Stored by transition pattern. For example, if the transition of the sales ratio of each unit period of the existing product is accumulated until the second predetermined period has elapsed, a graph as illustrated in FIG. 22 (1) can be obtained. From this graph, an approximate line based on a linear expression or a quadratic expression can be estimated, and a model expression for each estimated pattern can be generated. This storage means stores, for example, model expressions for each estimated pattern generated by this method. In the case of FIG. 22, a model formula for calculating the cumulative actual number of sales until the second predetermined period elapses from the cumulative actual number of sales until the first predetermined period elapses of the product having the transition pattern α, and the transition pattern β. And the product model formula that is the transition pattern γ are stored.
[0013]
The transition pattern assumption means assumes a transition pattern of a new product. For example, if the transition pattern of each existing product is classified into three transition patterns of α, β, γ, it is assumed that the transition pattern of the new product is α, then β, and finally Assume that γ.
The cumulative predicted sales number calculation means calculates the new product from the model expression of the transition pattern assumed by the transition pattern assumption means and the cumulative actual sales number stored until the first predetermined period stored in the actual sales number storage means. The cumulative number of predicted sales until the second predetermined period elapses is calculated. As described above, when it is assumed that the transition pattern of the new product is α, the accumulated actual number of sales until the first predetermined period of time is substituted into the model formula of the pattern α, and the second Calculate the cumulative predicted sales number Σα up to the point of the predetermined period. If it is assumed that the pattern is β, the cumulative estimated sales number until the second predetermined period elapses until the first predetermined period elapses by substituting the accumulated actual sales number until the first predetermined period elapses into the model formula of the pattern β. Σβ is calculated. If it is assumed that the pattern γ, the cumulative actual sales number until the first predetermined period elapses is substituted into the model formula of the pattern γ, and the cumulative predicted sales number until the second predetermined period elapses of the new product Calculate Σγ.
The transition calculation means calculates a transition Tα of a value obtained by dividing the actual sales number for each unit period from the start of sales of the new product to the elapse of the first predetermined period by the cumulative predicted sales number Σα, and the cumulative predicted sales number Σβ To calculate the transition Tβ of the value obtained by dividing the number of actual sales by unit period from the start of sales of the new product to the elapse of the first predetermined period, and the cumulative forecast sales number Σγ 1 A transition Tγ of a value obtained by dividing the number of actual sales for each unit period until the elapse of a predetermined period is calculated.
[0014]
The transition similarity calculation means calculates the standard transition of the transition pattern assumed by the transition pattern assumption means and the similarity between the transitions calculated by the transition calculation means. In the above example, the standard transition of the assumed transition pattern α and the similarity of the transition Tα are calculated, the standard transition of the assumed transition pattern β and the similarity of the transition Tβ are calculated, and the assumed transition pattern The similarity between the standard transition of γ and the transition Tγ is calculated. For example, an absolute error between the standard transition and the transition calculated by the transition calculating means can be calculated to obtain the similarity. In this case, the smaller the absolute error, the higher the similarity, and the larger the absolute error, the lower the similarity. Then, the transition pattern selection unit selects the transition pattern having the highest similarity calculated by the transition similarity calculation unit.
The predicted sales number calculating means calculates the ratio of the actual sales number of each unit period in the standard transition of the transition pattern selected by the transition pattern selecting means to the second predetermined period of the new product calculated by the cumulative predicted sales number calculating means. Is multiplied by the cumulative predicted sales number until the time point elapses to calculate the predicted sales number for each unit period from the time point when the first predetermined period has elapsed to the time point when the second predetermined period has elapsed.
[0015]
This product sales number prediction device uses a model formula prepared for each transition pattern. The transition of the new product is calculated based on the cumulative predicted sales number calculated from the model formula for each transition pattern, and the transition pattern that is most similar is selected by comparing each transition with the standard transition of each transition pattern. By using a model formula prepared for each transition pattern, it is possible to obtain a cumulative forecast sales number with a smaller error, and therefore it is possible to perform a more accurate forecast.
[0016]
The present invention has created a method for predicting the number of sales by unit period from when the first predetermined period has elapsed since the start of sales of a new product until the second predetermined period elapses from that point. This method is performed by the computer in the following steps:
A process of reading the transition of the actual sales number of new products by unit period divided into a plurality of first predetermined periods;
Reading the standard transition by transition pattern of the value obtained by dividing the number of actual sales by unit period by the cumulative number of actual sales from the start of sales of existing products to the time when the first predetermined period and the second predetermined period have elapsed;
A step of reading a model formula for calculating an accumulated actual number of sales until the second predetermined period has elapsed from a cumulative actual number of sales of the existing product until the first predetermined period has elapsed;
Calculating a cumulative estimated sales number until the second predetermined period elapses of the new product from the read model formula and the accumulated actual sales number until the first predetermined period elapses;
A step of calculating a transition of a value obtained by dividing the number of actual sales by unit period from the start of sales of a new product to the time when the first predetermined period has elapsed by the calculated cumulative estimated sales number;
A process of selecting a transition pattern most similar to the calculated transition among the standard transitions by the read transition pattern;
・ From the standard transition of the selected transition pattern and the cumulative estimated sales number of the new product calculated until the second predetermined period has elapsed, by unit period from the first predetermined period to the second predetermined period And a step of calculating a predicted sales number.
According to the method for predicting the number of sales of new products, the number of sales of new products can be predicted from the transition of the actual sales number of new products during the first predetermined period and the transition of the actual sales number of existing products. This can reduce the burden and improve efficiency.
[0017]
In addition, another method for predicting the number of products sold has been created by the present invention. This method is performed by the computer in the following steps:
A process of reading the transition of the actual sales number of new products by unit period divided into a plurality of first predetermined periods;
Reading the standard transition by transition pattern of the value obtained by dividing the number of actual sales by unit period by the cumulative number of actual sales from the start of sales of existing products to the time when the first predetermined period and the second predetermined period have elapsed;
A step of reading a model formula for calculating the cumulative number of sales until the second predetermined period elapses from the cumulative actual number of sales of the existing product until the first predetermined period elapses by transition pattern;
・ Assuming transition pattern of new products,
A step of calculating a model expression of the assumed transition pattern and the cumulative number of sales sold until the second predetermined period elapses of the new product from the read cumulative sales number until the first predetermined period elapses;
A step of calculating a transition of a value obtained by dividing the number of actual sales by unit period from the start of sales of a new product to the time when the first predetermined period has elapsed by the calculated cumulative estimated sales number;
・ The standard transition of the assumed transition pattern and the process of calculating the similarity of transition,
The process of selecting the transition pattern with the highest calculated similarity,
・ From the standard transition of the selected transition pattern and the cumulative estimated sales number of the new product calculated until the second predetermined period has elapsed, by unit period from the first predetermined period to the second predetermined period And a step of calculating a predicted sales number.
According to the method for predicting the number of sales of products, it is possible to perform prediction with higher accuracy.
[0018]
According to the present invention, a program for predicting the number of sales for each unit period from when the first predetermined period has elapsed since the start of sales of a new product until when the second predetermined period has elapsed has been created. This program performs the following processing on the computer:
A process of reading the transition of the actual sales number of new products by unit period divided into a plurality of first predetermined periods;
A process of reading the standard transition by the transition pattern of the value obtained by dividing the actual sales by unit period by the cumulative actual sales from the start of sales of existing products to the time when the first predetermined period and the second predetermined period have elapsed;
A process of reading a model formula for calculating an accumulated actual sales number from the accumulated actual sales number until the second predetermined period elapses until the first predetermined period elapses of the existing product;
A process of calculating the read model formula and the cumulative predicted sales number until the second predetermined period elapses of the new product from the read cumulative sales number until the first predetermined period elapses;
A process of calculating a transition of a value obtained by dividing the actual sales number for each unit period from the sales start time of the new product to the time point of the first predetermined period by the calculated cumulative sales number;
A process of selecting a transition pattern that is most similar to the calculated transition among the standard transitions read by transition pattern;
・ From the standard transition of the selected transition pattern and the cumulative estimated sales number of the new product calculated until the second predetermined period has elapsed, by unit period from the first predetermined period to the second predetermined period And a process for calculating the predicted number of sales.
According to this program, the number of new products sold can be accurately calculated from the transition of the actual sales number of new products during the first predetermined period and the actual sales number of existing products, reducing the burden of prediction processing. Can contribute to efficiency improvement.
[0019]
Further, according to the present invention, another program for predicting the number of products sold has been created. This program performs the following processing on the computer:
A process of reading the transition of the actual sales number of new products by unit period divided into a plurality of first predetermined periods;
A process of reading the standard transition by transition pattern of the value obtained by dividing the actual sales number for each unit period by the cumulative actual sales number from the start of sales of existing products to the time when the first predetermined period and the second predetermined period have elapsed;
A process of reading a model formula for calculating the cumulative number of sales until the second predetermined period elapses from the cumulative actual number of sales of the existing product until the first predetermined period elapses by transition pattern;
・ Process that assumes the transition pattern of new products,
A process of calculating a model expression of the assumed transition pattern and the cumulative actual sales number until the second predetermined period elapses from the read cumulative actual sales number until the first predetermined period elapses;
A process of calculating a transition of a value obtained by dividing the actual sales number for each unit period from the sales start time of the new product to the time point of the first predetermined period by the calculated cumulative sales number;
・ The standard transition of the assumed transition pattern and the process of calculating the similarity of transition,
The process of selecting the transition pattern with the highest calculated similarity,
・ From the standard transition of the selected transition pattern and the cumulative estimated sales number of the new product calculated until the second predetermined period has elapsed, by unit period from the first predetermined period to the second predetermined period And a process for calculating the predicted number of sales.
According to the program for predicting the number of products sold, it is possible to perform more accurate prediction.
[0020]
DETAILED DESCRIPTION OF THE INVENTION The main features of the embodiments described below are listed first.
(Embodiment 1) The sales number prediction device includes:
A seasonal index calculation means for calculating a seasonal index based on the number of actual sales by unit period of existing products that have passed one year from the start of sales;
Add seasonality to the forecasted sales quantity by multiplying the seasonal index calculated by the seasonal index calculation means by the forecasted sales quantity for the corresponding unit period of the new product calculated by the forecast sales quantity calculation means Means,
Is added.
(Mode 2) The seasonality adding means further adds seasonality to the predicted number of sales in consideration of sales series characteristics.
(Mode 3) The actual sales number storage means stores the sales areas of the products in association with each other, and the transition pattern-specific standard transition storage means includes the new product among the existing products that have passed 52 weeks or more from the sales start time. I remember the standard transition according to the transition pattern of the same sales area,
The model formula storage means stores a model formula generated based on the actual number of sales in the same sales area as the new product among the existing products that have passed 52 weeks or more from the sales start time.
[0021]
First, the main terms used in this embodiment will be described.
In this embodiment, the vehicle type refers to the type of automobile that is a product. A category refers to a classification to which a plurality of vehicle types having similar sizes and shapes belong. For example, compact cars, sedans, wagons and the like are applicable.
The sales line refers to the line to which the automobile dealer belongs. Car models that can be sold in one sales line are often determined.
The ratio transition refers to the transition of the percentage (%) of the actual sales number for each week, where the cumulative actual sales number for 52 weeks is 100%.
[0022]
With reference to FIG. 1, the relationship between the sales period of the prediction target vehicle model of the sales number of this embodiment and the prediction target period, and the relationship between the sales period of the vehicle type that provides data used for sales number prediction will be described.
In this embodiment, the sales target vehicle type for sales is X. The vehicle type X has a current sales period (first predetermined period) of 4 weeks, which is a new vehicle type immediately after the start of sales. The forecast target period (second predetermined period) is 48 weeks (shaded grant week) from the present four weeks after the start of sales, and the period from the start of sales to the end of the forecast target period (52 weeks) ) Corresponds to the period (one year in this embodiment) that is the target of the sales plan or production plan.
Vehicle types that provide data used to predict the number of sales of vehicle type X are A to J. Vehicle types A to J are existing vehicle types with a current sales period of 52 weeks or more. For example, vehicle type A has 52 weeks since the start of sales, vehicle type B has passed 104 weeks, vehicle type C has passed 53 weeks, ..., vehicle type J is 156 A week has passed. For any vehicle type, the data used to create the rate transition is for 52 weeks (hatch grant week) after the start of sales.
[0023]
(First Embodiment) FIG. 2 is a block diagram showing a main configuration of a sales number prediction apparatus according to a first embodiment embodying the present invention.
The sales number prediction device 20 is a computer that inputs data, stores, calculates, outputs a calculation result, and makes a peripheral device function based on the calculation result, and a description of pure hardware is omitted. The sales number prediction device 20 may be configured by a single computer, or may have input, storage, calculation, and output functions distributed among a plurality of computers as necessary, and the computers are connected via a network. However, a plurality of computers may be configured to operate as one sales number prediction device 20.
The sales number prediction apparatus 20 is connected to the management computer 10 in the sales department and the management computer 60 in the production department through a computer network so that they can communicate with each other.
[0024]
The sales number prediction apparatus 20 shown in FIG. 2 associates the number of sales by week from the sales department management computer 10 to the vehicle type, the category to which the vehicle type belongs, the sales series, and the sales region via the network. Will be sent. The actual sales number storage means 72a stores the actual sales number that has been transmitted.
FIG. 4 shows a data configuration example stored in the actual sales number storage means 72a. For example, vehicle type A has category 1, sales series (1), sales regions a and b, and the actual number of sales is 3520 in the first week from the start of sales, 1575 in the second week, and the third week 990, the fourth week is 675, and the actual number of sales up to the 52nd week is stored (as shown in FIG. 1, 52 weeks have passed since the sales start time for the vehicle type A). Car type B has category 1, sales series (1), sales regions c and d, and the actual number of sales is 1200 in the first week from the start of sales, 1150 in the second week, and 3 in the third week. 1230, 1170 is the 4th week, and the actual number of sales up to the 104th week is stored (as shown in FIG. 1, 104 weeks have passed since the sales start time for the vehicle type B). In the same manner, the category, sales series, sales area, and actual sales by week are shown for each model.
[0025]
The data preparation means 30 shown in FIG. 2 prepares data used for predicting the number of sales in advance. The data preparation means 30 prepares seasonal index data, seasonally adjusted (hereinafter referred to as “seasonally adjusted”) actual sales data, ratio transition (including standard transition) data, and model formula It is data of.
The seasonal index data and the seasonally adjusted actual sales data are not essential to the present invention. However, since it is necessary for performing the seasonality addition process described later, it is prepared in this embodiment. Accordingly, the seasonal index calculating means 32 and the seasonal adjusting means 34 described below are not essential components of the present invention, but are provided in the present embodiment.
[0026]
FIG. 3 shows a block diagram of the main configuration of the data preparation means 30.
Seasonal index data is prepared by the seasonal index calculation means 32. The seasonal index calculating means 32 (1) totals the weekly actual sales numbers of all the sales unit periods stored in the sales number storage means 72a by month, and (2) calculates the seasonal index from the total results.
The calculation processes (1) to (2) will be described in detail later.
In this case, in (1), (a) by vehicle type, (b) by category, (c) by unit of all vehicle types, and (d) by sales unit by unit, and in (2) these four Calculate the seasonal index for each unit.
The seasonal index storage means 74 stores (a) the vehicle type, (b) the category, (c) all the vehicle types, and (d) the sales series-specific seasonal index calculated by the seasonal index calculation means 32. FIG. 5 (1) shows an example of the data structure of (b) category-specific seasonal indices stored in the seasonal index storage means 74. For example, the seasonal index of category 1 (car types A and B) is 103.5 in January, 105.5 in February, 88.1 in March, and so on, and the minimum is 64.0 in November. The highest is 165.3 in June. Similarly, the seasonal index from January to December is shown for each category.
[0027]
The data on the number of seasonally adjusted sales is prepared by the seasonal adjustment means 34 of FIG. The seasonal adjustment means 34 multiplies all weekly actual sales numbers stored in the actual sales number storage means 72a by the seasonal index stored in the seasonal index storage means 74 to calculate the seasonally adjusted actual sales numbers. . The seasonal index is calculated based on (a) vehicle type, (b) category, (c) all vehicle types, (a) vehicle type, (b) category, and (c) all vehicle types. Calculate the number of seasonally adjusted sales in units. This calculation process will be described in detail later.
The seasonally adjusted actual sales number calculated by the seasonal adjustment means 34 is stored in the seasonally adjusted actual sales number storage means 72b.
[0028]
2 and 3, the solid line indicates the flow of data that is not seasonally adjusted, and the alternate long and short dash line indicates the flow of data that is seasonally adjusted (including seasonal index data itself). Further, the storage means to which the symbol “a” is attached stores data that is not seasonally adjusted, and the storage means to which the symbol “b” is attached stores data that has been seasonally adjusted.
[0029]
The ratio transition data (including standard transition data) is prepared by the ratio transition creation means 36 of FIG. The ratio transition creating means 36 (1) creates a transition of the ratio of the actual sales numbers for each week over 52 weeks based on the weekly actual sales numbers stored in the actual sales number storage means 72a.
Alternatively, (1) based on the weekly seasonally adjusted actual sales number stored in the seasonally adjusted actual sales number storage unit 72b, a ratio transition of the actual sales number over 52 weeks is created. In this case, the actual number of seasonally adjusted sales is (a) by vehicle type, (b) by category, (c) calculated using all vehicle types, (a) by vehicle type, (b) by category, ( c) Proportional transitions are created for each vehicle type.
Then, (2) classify similar ones of the created ratio transitions into patterns, and (3) create a standard transition of the transition pattern.
The creation processes (1) to (3) will be described in detail later.
The ratio transition of each vehicle type created based on the actual sales number and the standard transition of each transition pattern are stored in the transition storage means 76a. Further, the ratio transition of each vehicle type and the standard transition of each transition pattern created on the basis of the number of seasonally adjusted sales are stored in the transition storage means 76b.
[0030]
The model formula data is prepared by the model formula generation means 38 of FIG. The model formula generation means 38 creates (1) an actual sales number correlation graph from the weekly actual sales numbers stored in the actual sales number storage means 72a. The correlation graph here is a graph representing the correlation between the cumulative actual sales number from the sales start time of the vehicle types A to j to 4 weeks and the cumulative actual sales number from the sales start time to 52 weeks.
Alternatively, (1) a seasonally adjusted actual sales number correlation graph is created from the weekly seasonally adjusted actual sales numbers stored in the actual sales number storage means 72b. In this case, the actual number of seasonally adjusted sales is (a) by vehicle type, (b) by category, (c) calculated using all vehicle types, (a) by vehicle type, (b) by category, ( c) A correlation graph is created for each vehicle type.
(2) A model formula is generated from the created correlation graph.
The generation processing (1) to (2) will be described in detail later.
The model formula generated based on the actual sales number is stored in the model formula storage means 78a. In addition, the model formula created based on the number of seasonally adjusted sales is stored in the model formula storage unit 78b.
[0031]
The data prepared as described above is used by the sales number prediction means 50 shown in FIG.
The sales number predicting means 50 calculates the predicted sales number, which is the target number of the sales number prediction process, using the various data described above. For this purpose, the sales number prediction unit 50 includes a transition pattern selection unit 52, a predicted sales number calculation unit 54, and a seasonality addition unit 56.
The seasonality adding means 56 is not essential for the present invention, but is effective in improving the prediction accuracy, and is therefore provided in this embodiment.
[0032]
The sales number prediction unit 50 executes the sales number prediction process according to the information input from the required information input unit 40. Various information necessary for the sales number prediction process is input from the required information input means 40. This information will be described below.
[0033]
The transition pattern selection means 52 of the sales number prediction means 50 selects the transition pattern most similar to the ratio transition of the prediction target vehicle type among the standard transitions stored in the transition storage means 76a. When an instruction to perform seasonality addition processing is input from the required information input means 40, a transition pattern that is most similar to the ratio transition of the prediction target vehicle type is selected from the standard transitions stored in the transition storage means 76b. .
Specifically, when the instruction is not input from the required information input means 40,
(1) Using the prediction model stored in the model formula storage unit 78a, calculate the cumulative predicted sales number over 52 weeks from the actual sales number stored in the actual sales number storage unit 72a.
(2) Divide the actual number of sales by week for 4 weeks from the start of sales of the new product by the cumulative predicted number of sales, calculate the percentage transition for the first 4 weeks, and (3) be stored in the transition storage means 76a The transition pattern that is most similar to the percentage transition is selected from the standard transitions.
On the other hand, when the instruction is input from the required information input means 40, (1) using the prediction model stored in the model formula storage means 78b, the season adjustment stored in the season adjusted actual sales number storage means 72b. Calculate the cumulative forecast sales number over 52 weeks from the actual sales count, and (2) divide the actual sales count by week for the first four weeks from the start of sales of the new product by the cumulative forecast sales count. A ratio transition is calculated, and (3) a transition pattern most similar to the ratio transition is selected from the standard transitions stored in the transition storage unit 76b. The instructions further specify whether to use a seasonal index calculated in units of (a) vehicle type, (b) by category, or (c) all vehicle types, so in (1) to (3) Uses seasonally adjusted data in specified units.
The selection process (1) to (3) will be described in detail later.
[0034]
The predicted sales number calculation means 54 of the sales number prediction means 50 uses the standard transition of the transition pattern selected by the transition pattern selection means 52 and uses the cumulative predicted sales number calculated by the transition pattern selection means 53 to predict the target vehicle model. The forecasted number of sales by week from 5 to 52 weeks is calculated.
Also in this case, whether or not to use the standard transition or the cumulative sales number based on the seasonally adjusted data is determined depending on whether or not an instruction to perform the seasonality addition process is input from the required information input means 40.
The predicted sales number calculated based on the actual sales number is stored in the predicted sales number storage unit 80a. The predicted sales number calculated based on the seasonally adjusted data is stored in the predicted sales number storage unit 80b.
[0035]
When an instruction to perform seasonality addition processing is input from the required information input means 40, the seasonality addition means 56 operates. The seasonality adding unit 56 adds the seasonality to the predicted sales number by multiplying the predicted sales number stored in the predicted sales number storage unit 80 b by the seasonal index stored in the seasonal index 74. As described above, this instruction specifies whether to use the seasonal index calculated in (a) vehicle type, (b) category, or (c) all vehicle types. Use the seasonal index calculated in.
Furthermore, when an instruction that takes into account the sales line characteristics is input, (d) seasonality is added using the seasonal index calculated for each sales line.
The predicted sales number with seasonality added by the seasonality adding means 56 is stored in the seasonality additional predicted sales number storage means 82c.
[0036]
The data stored in the predicted sales number storage means 80a, 80b or the seasonality-added predicted sales number storage means 82c is transmitted to the sales department management computer 10 and the production department management computer 60 via a computer network.
The sales department management computer 10 can determine a sales strategy, make a sales plan, and the like using the transmitted data. The production department management computer 60 makes it possible to predict orders based on the transmitted data and to make production plans.
[0037]
Next, with reference to the flowcharts of FIGS. 11 to 17, the product sales number prediction process performed by the sales number prediction apparatus 20 will be described.
FIG. 11 is a flowchart of the main process of the merchandise sales number prediction process. The main process is roughly divided into four processes: a data preparation process (step S2), a transition pattern selection process (step S4), a predicted sales number calculation process (step S6), and a seasonality addition process (step S10). The Hereinafter, each of these processes will be described in detail.
[0038]
In the data preparation process in step S2 of FIG. 11, four processes are executed: a seasonal index calculation process, an actual sales number seasonal adjustment process, a ratio transition creation process, and a model formula generation process.
Of these, the seasonal index calculation process and the actual sales number seasonal adjustment process are not essential processes of the present invention, but are effective in improving the prediction accuracy, and are therefore executed in this embodiment.
[0039]
First, the seasonal index calculation process executed by the seasonal index calculation means 32 of the data preparation means 30 will be described with reference to the flowchart of FIG.
First, the actual sales number of the vehicle type for which 52 weeks or more have elapsed from the sales start time is read from the actual sales number storage means 72a (step S202). The weekly actual sales numbers of all the sales unit periods of the vehicle types A to J shown in FIG. 4 are read.
Next, the read actual sales numbers are summed up by month for each unit of seasonal index calculation (step S204). That is, the actual sales numbers shown in FIG. 4 are as follows: (a) By vehicle type (A to J), (b) By category (1 to 5), (c) By sales series ((1) to (3)), ( d) Aggregate by month from January to December in each vehicle type (A to J). In this case, since the sales unit period of each vehicle type is different, for example, for vehicle type A, the number of weekly actual sales for one cycle from January to December is totaled, and for vehicle type B, the number of weeks for two cycles is counted. Aggregate the number of actual sales.
[0040]
In the subsequent step S206 in FIG. 12, the seasonal index is calculated based on the result of aggregation in step S204 in four units: (a) vehicle type, (b) category, (c) sales series, and (d) all vehicle types. Calculate. As a method for calculating the seasonal index, various existing methods such as a monthly average method, a linkage ratio method, a moving average method, and a census station method can be employed. The formula shown in FIG. 5 (2) is based on the most common monthly average method. In the monthly average method, a seasonal index is obtained by dividing the actual sales number of each month by the average of the actual sales number for one year and multiplying it by 100.
The calculated seasonal index is stored in the seasonal index storage means 74 (step S208). FIG. 5 (1) shows an example of the data configuration of (b) category-specific seasonal indices stored in the seasonal index storage means 74.
[0041]
Next, the actual sales number seasonal adjustment process executed by the seasonal adjustment means 34 will be described.
In this process, all the weekly actual sales numbers of the vehicle types that have been stored for 52 weeks or more from the sales start time stored in the actual sales number storage means 72a are multiplied by the seasonal index stored in the seasonal index storage means 74. Calculate the number of seasonally adjusted sales. In this case, the seasonal index is calculated using (a) vehicle type, (b) category, (c) all vehicle types, (a) vehicle type, (b) category, (c) all vehicle types. Calculate the number of seasonally adjusted sales in each unit.
For example, if the 1st to 4th weeks of the sales unit period for the category 1 car model A in FIG. 4 is April, the April season index (78.0) in category 1 in FIG. When multiplied, the number of seasonally adjusted sales by car category for model A is 3520 × 0.78≈2746 in the first week, 1575 × 0.78≈1229 in the second week, and 990 × 0.78 in the third week. ≒ 772, the 4th week is calculated as 675 x 0.78 ≒ 527 (rounded to the first decimal place).
[0042]
Next, the ratio transition creation processing executed by the ratio transition creation means 36 will be described with reference to the flowchart of FIG.
The ratio transition is created based on the actual sales number stored in the actual sales number storage unit 72a and the seasonally adjusted actual sales number stored in the seasonally adjusted actual sales number storage unit 72b. There is something. Below, the ratio transition created based on the actual sales number memorize | stored in the actual sales number memory | storage means 72a is demonstrated as a representative.
First, the actual number of sales for 52 weeks is read from the sales number storage means 72a for a vehicle type for which 52 weeks or more have passed since the sales start time (step S212). The weekly actual sales numbers for 52 weeks of the vehicle types A to J shown in FIG. 4 are read.
[0043]
Next, a ratio transition is created for each vehicle type from the actual sales number read in step S212 (step S214). The ratio transition is created by accumulating the actual sales numbers for 52 weeks, and determining the ratio of the actual sales numbers for each week from 1 week to 52 weeks with the accumulated actual sales numbers being 100%.
FIG. 6 illustrates the ratio transition of the created vehicle type A. (1) shows the ratio of the number of actual sales in each week out of 52 weeks as a number (%), and (2) shows a graph with the horizontal axis as week and the vertical axis as%. It can be seen that the number of actual sales immediately after the start of sales is the highest, and after that, it is a rate transition that repeats a slight increase and decrease.
[0044]
In the subsequent step S216 in FIG. 13, similar ratio transitions are classified into patterns. As an example of the classification method, cluster analysis can be used. In the cluster analysis, the distance is defined by regarding the data as a point in a multidimensional space, and those having a short distance are classified into the same pattern (cluster) as being similar.
FIG. 7 briefly explains pattern classification using cluster analysis. FIG. 7 is a graph in which the horizontal axis is plotted in two-dimensional coordinates with the horizontal axis representing the percentage of actual sales in the fourth week and the vertical axis representing the percentage of actual sales in the 12th week (%). . Each pattern (cluster) surrounded by an ellipse includes a change in the proportion of vehicle types in which the distance between plot points is short. Here, the ratio transition of the vehicle types A to J is classified into three transition patterns of α, β, and γ. In this example, a two-dimensional graph is used for simplicity, but a three-dimensional graph or more can also be used. In addition to the ratio (%) of the actual sales number, the actual sales number (units) and other values may be used for plotting.
[0045]
In the subsequent step S218 of FIG. 13, a standard transition of the transition pattern classified in step S216 is created. A standard transition is created by taking an average value of the ratio transition of each vehicle type in the transition pattern. For example, when creating a standard transition of the pattern α, the ratio of the actual sales number for the first week of the vehicle type A belonging to the pattern α (see FIG. 6 (1)) and the actual sales number for the first week of the vehicle type D The average value of the ratio and the ratio of the actual sales volume of the first week of the vehicle type G is taken as the ratio of the first week of the standard transition, and the ratio of the actual sales volume of the vehicle model A in the second week (Fig. 6 (1)) The average value of the ratio of the actual sales number for the second week of the vehicle type D and the ratio of the actual sales number for the second week of the vehicle type G is taken as the ratio of the second week of the standard transition. This calculation is performed over 52 weeks. Similarly, this calculation is executed in creating standard transitions of the patterns β and γ.
FIG. 8 shows the created standard transition. (1) shows the ratio of the number of actual sales for each week in each transition pattern as a number (%), and (2) shows a graph with the horizontal axis as week and the vertical axis as%. Pattern α has the highest number of actual sales immediately after the start of sales, and thereafter repeats a slight increase / decrease. Pattern β is a pattern that has a slight constant increase / decrease, but has a substantially constant actual sales number, and pattern γ It can be seen that the number of actual sales is small immediately after the start of sales, but gradually increases and then gradually increases and decreases.
The created ratio transition is associated with the vehicle type, and the created standard transition is associated with the transition pattern and stored in the transition storage unit 76a (step S220). The ratio transition and the standard transition created based on the number of seasonally adjusted sales are stored in the transition storage unit 76b (also in step S220).
[0046]
Next, a model formula generation process executed by the model formula generation unit 38 will be described with reference to the flowchart of FIG.
The model formula is generated based on the actual sales number stored in the actual sales number storage unit 72a and the seasonally adjusted actual sales number stored in the seasonally adjusted actual sales number storage unit 72b. There is something. Below, the model formula produced | generated based on the results sales number memorize | stored in the results sales number memory | storage means 72a as a representative is demonstrated.
First, the actual sales number of the vehicle type for which 52 weeks or more have elapsed from the sales start time is read from the actual sales number storage means 72a (step S222). The weekly actual sales numbers of all the sales unit periods of the vehicle types A to J shown in FIG. 4 are read.
[0047]
Next, a correlation graph of the actual sales number read in step S222 is created (step S224). The correlation graph here is a graph representing the correlation between the cumulative actual sales number up to w1 week (first predetermined period) and the cumulative actual sales number up to w2 week (first predetermined period + second predetermined period). FIG. 9 (1) shows an example of a correlation graph in which the horizontal axis is w1 week = 4 weeks and the vertical axis is w2 week = 52 weeks.
[0048]
In the subsequent step S226 in FIG. 14, an approximate line based on a primary expression or a secondary expression is estimated from the correlation graph created in step S224, and a model expression for calculating the 52-week cumulative predicted sales number of the prediction target vehicle type is generated. In the case of the example of FIG. 9 (1), since the approximate line f1 by a quadratic equation is estimated, the model equation shown in (2) is generated.
The generated model formula is stored in the model formula storage unit 78a (step S228). Note that the model formula created based on the number of seasonally adjusted sales is stored in the model formula storage means 78b (also step S228).
[0049]
As described above, when the preparation of each data of the seasonal index, the ratio transition (including the standard transition), and the model formula is completed by the data preparation process, the transition pattern selection process is executed in step S4 of FIG.
With reference to the flowchart of FIG. 15, the transition pattern selection process which the transition pattern selection means 52 of the sales number prediction means 50 performs is demonstrated.
First, various information necessary for the sales number prediction process is input from the required information input means 40 (step S402). As the information essential for the sales number prediction process, the actual sales number for four weeks of the prediction target vehicle type is input. In the present embodiment, the actual number of sales for four weeks for the vehicle type X shown in FIG. At this time, as optional information, it is also possible to input instruction information for seasonality addition processing described below (details of the instruction information will be described later).
[0050]
Next, the model formula is read from the model formula storage unit 78a (step S404). Alternatively, when a seasonality addition processing instruction is input in step S402, the model formula is read from the model formula storage unit 78b (also step S404).
In step S406, using the model formula read in step S404, the cumulative predicted sales number over 52 weeks is calculated from the four-week cumulative actual sales number of the prediction target vehicle type input in step S402.
As shown in FIG. 10 (1), the cumulative actual number of sales for the four weeks of the prediction target vehicle type X is calculated as 6300 + 2240 + 1820 + 1450 = 111810. By substituting this into the model formula of FIG. 9 (2), it is possible to calculate the cumulative forecast sales number over 52 weeks. The calculation result is illustrated in FIG.
[0051]
In the subsequent step S408 in FIG. 15, the weekly percentage change for four weeks is calculated by dividing the weekly actual sales number for four weeks by the cumulative predicted sales number. For example, when the weekly actual sales number for four weeks shown in (1) is divided by the cumulative predicted sales number 84335 for 52 weeks of the vehicle type X illustrated in FIG. 10 (2), 7.5% in the first week, The percentage change is calculated to be 2.7% for the second week, 2.2% for the third week, and 1.7% for the fourth week (rounded to the first decimal place). The calculation result is shown in FIG.
[0052]
In step S410 of FIG. 15, the standard transition of each transition pattern is read from the transition storage means 76a. Alternatively, when a seasonality addition processing instruction is input in step S402, the standard transition of each transition pattern is read from the transition storage unit 76b.
In step S412, the transition pattern most similar to the ratio transition is selected from the standard transitions read in step S410. The absolute error between the standard transition shown in FIG. 8 and the weekly ratio transition of 4 weeks shown in FIG. 10 (3) is that pattern α is 2.8, pattern β is 7.0, and pattern γ is 12.1. Since the pattern α is most similar to the pattern α, the pattern α is selected.
[0053]
When the transition pattern is selected as described above, the predicted sales number calculation process is executed in step S6 of FIG.
With reference to the flowchart of FIG. 16, the predicted sales number calculation process executed by the predicted sales number calculation unit 54 of the sales number prediction unit 50 will be described.
First, the cumulative predicted sales number for 52 weeks of the prediction target vehicle type calculated by the transition pattern selection means 52 is acquired (step S602). The cumulative estimated sales number of the vehicle type X shown in FIG.
Next, the standard transition of the transition pattern selected by the transition pattern selection unit 52 is acquired (step S604). In the above example, the standard transition of the pattern α is acquired.
In step S606, using the standard transition of the selected transition pattern, the predicted number of sales for each week from 5 weeks to 52 weeks is calculated from the cumulative predicted number of sales over 52 weeks. For example, in the standard transition of the pattern α shown in FIG. 4, the ratio at the 50th week is 1.6%, the ratio at the 51st week is 1.9%, and the ratio at the 52nd week is 2.1%. Multiplying the cumulative forecast sales numbers illustrated in FIG. 10 (3), 1.6% × 84335≈1349 for the 50th week, 1.9% × 84335≈1602 for the 51st week, and 2.1% for the 50th week A predicted sales number of x84335≈1771 (rounded to one decimal place) is calculated. Similar calculations are performed for all weeks from 5 to 52 weeks.
The cumulative predicted sales number for 52 weeks acquired in step S602 and the weekly predicted sales number for weeks 5 to 52 calculated in step S606 are stored in the predicted sales number storage unit 80a (step S608). Alternatively, when a seasonality addition processing instruction is input from the required information input means 40, it is stored in the predicted sales number storage means 80b (same step S608).
[0054]
When the predicted sales number is calculated as described above, it is determined in step S8 in FIG. 11 whether or not seasonality is added to the predicted sales number.
When the seasonality addition processing instruction is input from the required information input means 40 in step S402 in FIG. 15 (YES in step S8 in FIG. 11), the seasonality addition means 56 of the sales number prediction means 50 operates, In step S10 of FIG. 11, a process of adding seasonality to the predicted sales number is executed. For any product, the number of sales varies depending on the season. Therefore, by adding seasonality, a more accurate predicted sales number can be obtained.
[0055]
With reference to the flowchart of FIG. 17, the seasonality addition process which the seasonality addition means 56 performs is demonstrated.
First, the sales start date of the prediction target vehicle type and the unit of the seasonal index used ((a) vehicle type, (b) by category, (c) input together with the seasonality addition processing instruction from the required information input means 40 Any of all vehicle types) is acquired (step S102).
Next, in step S104, the seasonal index is read from the seasonal index storage means 74. At this time, the seasonal index to which the prediction target vehicle type belongs among the units input in step S102 (for example, (b) the category index seasonal index to which the vehicle type X belongs when the unit for each category is input) is read. Also, the month to which the input sales start date belongs is read as the first month (for example, if the sales start date of vehicle type X is April 3, then April is the first month). At the same time, the predicted sales number of the prediction target vehicle type is read from the predicted sales number storage unit 80b.
[0056]
In step S106, it is determined whether or not an instruction for adding sales series characteristics to the seasonal index has been input from the required information input means 40 together with the seasonality addition processing instruction. If NO, the process moves to step S110, and if YES, the process moves to step S108.
When the process proceeds to step S108, the sales series characteristic index is added to the seasonal index read in step S104. The sales series characteristic index is the difference between (d) the seasonal index for all vehicle types and (d) the seasonal index calculated for each sales series. For example, if the April seasonal index for all models is 77.0 and the April seasonal index for sales series (1) is 80.0, the characteristic index for sales series (1) is 77.0- 80.0 = −3.0. If the seasonal index read in step S104 is (b) classified by category, the seasonal index for April in category 1 is 78.0 in the example of FIG. The characteristic index -3.0 of 1 ▼ is added to obtain a seasonal index that takes into account the sales line characteristic of 75.0.
In this way, taking into account the characteristics of the sales line, for example, the number of sales that do not depend on the season, such as the month in which the sales volume temporarily increased as a result of a sales promotion campaign in the sales line or the sales power of each sales line Can also be considered.
[0057]
In step S110, the seasonal sales number is multiplied by the weekly predicted sales number for the corresponding week to calculate the seasonality additional predicted sales number. If the sales series characteristics are taken into account in step S108, the calculation is made using the seasonal index with the sales series characteristics taken into account.
The calculated seasonal additional predicted sales number is stored in the seasonal additional predicted sales number storage unit 82c.
[0058]
Thus, the product sales number prediction process of the first embodiment is completed.
In the product sales number prediction process of the present embodiment, only the data created from the actual sales number of the existing vehicle type and the actual sales number for the four weeks of the prediction target vehicle type are used. By combining simple calculation methods that have been conventionally used for forecasting the sales volume, it is possible to obtain the predicted sales volume with the same high accuracy as in the past.
Therefore, the calculation amount and calculation time can be saved, and the efficiency of the prediction process can be improved.
[0059]
(Second Embodiment) In the second embodiment, a model formula is generated for each transition pattern, and the cumulative predicted sales number of the prediction target product is calculated for each transition pattern. As a result, it is possible to obtain an accumulated predicted sales number with a small prediction error, and to perform a more accurate sales number prediction process.
Since other processes and configurations are the same as those of the first embodiment, only processes and configurations different from the first embodiment will be described below.
[0060]
First, the sales number prediction device of the second embodiment will be described. FIG. 18 is a block diagram of the main configuration of the sales number prediction device 120 of the second embodiment. Since the main configuration of the sales number prediction device 120 is the same as the main configuration of the sales number prediction device 20 of the first embodiment, a duplicate description is omitted.
FIG. 19 shows a block diagram of the data preparation means 130 of the sales number prediction apparatus of the second embodiment. In this data preparation means 130, the data used by the model formula generation means 138 is different from that of the first embodiment. The data used by the model formula generation unit 138 is a standard transition stored in the transition storage unit 176a or a standard transition stored in the transition storage unit 176b.
[0061]
A model formula generation process executed by the model formula generation unit 138 will be described with reference to the flowchart of FIG.
First, the standard transition of each transition pattern is read from the transition storage unit 176a or the transition storage unit 176b (step S232). In the following, as a representative, the description will be made based on the standard transition read from the transition storage unit 176a. Standard transitions of transition patterns α, β, and γ shown in FIG. 8 are read.
Next, a cumulative graph of each standard transition is created (step S234). When the ratio (%) of each week of the standard transition of the transition patterns α, β, and γ shown in FIG. 8 is accumulated, a cumulative graph as shown in FIG. 22 (1) can be created.
[0062]
In step S236 of FIG. 20, an approximate line based on a primary expression or a quadratic expression is estimated from the cumulative graph created in step S234, and a model expression for calculating the 52-week cumulative predicted sales number of the prediction target vehicle type is generated. In the example of FIG. 22 (1), an approximate line based on a quadratic expression is estimated for the pattern α, an approximate line based on the primary expression is estimated for the pattern β, and an approximate line based on the linear expression is estimated for the pattern γ. The model expression shown is generated.
The generated model formula is stored in the model formula storage unit 178a in association with the transition pattern (step S238). The model formula created based on the standard transition stored in the transition storage unit 176b is stored in the model formula storage unit 178b (also in step S228).
[0063]
A transition pattern selection process executed by the transition pattern selection unit 152 of the sales number prediction unit 150 using the model formula prepared as described above will be described with reference to the flowchart of FIG.
First, various information necessary for the sales number prediction process is input from the required information input unit 140 (step S422). As the indispensable information, the actual sales number for four weeks of the vehicle type X shown in FIG. At this time, instruction information for seasonality addition processing can also be input as arbitrary information.
[0064]
Next, a transition pattern i to be processed first is set (step S424). The transition pattern to be initially set can be arbitrarily selected. However, in this embodiment, α is first set as processing is performed in the order of α → β → γ for convenience. In step S426, the transition pattern i of the prediction target vehicle type is assumed to be α.
In step S428, the model formula of the assumed transition pattern is read from the model formula storage unit 178a. Alternatively, if a seasonality addition processing instruction is input in step S422, the model formula of each transition pattern is read from the model formula storage unit 178b (also step S428).
In step S430, using the model expression of the transition pattern read in step S428, the cumulative predicted sales number for 52 weeks is calculated from the cumulative actual sales number for four weeks of the prediction target vehicle type input in step S422. The model formula of each transition pattern in FIG. 22 (2) is to obtain the cumulative ratio of each week. For example, when the cumulative ratio of the fourth week is obtained, and the cumulative number of actual sales up to four weeks of the prediction target vehicle model X shown in FIG. Can be calculated. The calculation result is illustrated in FIG. First, the cumulative actual number of sales is calculated by the model formula of the pattern α.
[0065]
In step S432 of FIG. 15, the weekly actual rate of sales for 4 weeks is calculated by dividing the weekly actual sales number of 4 weeks by the cumulative estimated sales number calculated in step S430. When the weekly actual sales number for four weeks shown in (1) is divided by the cumulative predicted sales number for 52 weeks of the vehicle type X by the model formula of the pattern α illustrated in FIG. 23 (2), the example is shown in (3). The percentage transition is calculated (8.5% for the first week, 3.0% for the second week, 2.5% for the third week, 2.0% for the fourth week, rounded to the first decimal place) ).
[0066]
In step S434 in FIG. 15, it is determined whether or not the next transition pattern remains. If NO, the process proceeds to step S438. If YES, β is set as the next transition pattern i in step S436, and the process returns to step S426 to execute a series of processes up to step S432. FIG. 23 (3) shows the transition of the ratio of the pattern β.
If YES is determined again in step S434, γ is set as the next transition pattern i in step S436, and the process returns to step S426 to execute a series of processes up to step S432. FIG. 23 (3) shows a change in the ratio of the pattern γ.
In the subsequent step S434, it is determined as NO, and the process proceeds to step S438.
[0067]
In step S438 in FIG. 15, the assumed standard transition of each transition pattern is read from the transition storage means 76a. Alternatively, when a seasonality addition process instruction is input in step S402, the standard transition of each transition pattern assumed from the transition storage unit 76b is read.
In step S440, the transition pattern having the highest similarity between the standard transition read in step S438 and the ratio transition is selected. The absolute error between the standard transition shown in FIG. 8 and the weekly ratio transition of four weeks shown in FIG. 23 (3) is that pattern α is 0.9, pattern β is 4.3, and pattern γ is 1.2. Therefore, since the pattern α has the highest similarity, the pattern α is selected.
[0068]
Using the standard transition selected as described above, the predicted sales number calculation unit 154 calculates the predicted sales number for each week over 52 weeks of the prediction target vehicle type. Further, the seasonality addition processing means 156 adds seasonality to the predicted sales number calculated in this way.
[0069]
Thus, in this embodiment, a model formula is generated for each transition pattern, and the cumulative predicted sales number for 52 weeks of the prediction target vehicle type is calculated for each transition pattern. Therefore, it is possible to obtain the cumulative predicted sales number with less error than the cumulative predicted sales number calculated by generating the model formulas for all vehicle types, and to obtain the predicted sales number with higher accuracy.
[0070]
Specific examples of the present invention have been described in detail above, but these are merely examples and do not limit the scope of the claims. The technology described in the claims includes various modifications and changes of the specific examples illustrated above.
In the present embodiment, the first predetermined period is 4 weeks from the start of sales, but is not limited to this. If the data for one week can be sufficiently predicted, one week from the start of sales may be set as the first predetermined period, and if further prediction accuracy is desired, longer than 8 weeks, 12 weeks, etc. from the start of sales. The period may be the first predetermined period, and an arbitrary period can be set.
In the present embodiment, the second predetermined period is set to 48 weeks, and the first predetermined period + the second predetermined period is set to 52 weeks (about one year). However, the present invention is not limited to this. If you want to make a short-term prediction, the first predetermined period + the second predetermined period may be 12 weeks (about March), 26 weeks (about half a year), or 104 weeks if you want to make a longer-term prediction (About 2 years), 156 weeks (about 3 years) may be set, and an arbitrary period can be set.
-In this example, the data of all products was used without considering the sales area, but only the products sold in the same area as the forecast measurement products were extracted, and the actual number of sales of these products It is also possible to use only the data. Thereby, the sales number prediction process in consideration of regional characteristics can be performed. For example, in FIG. 4, since the sales regions of the prediction target vehicle type X are a and d, only the vehicle regions A to J whose sales region is a or d are extracted and used. For example, for the vehicle type A, only the sales region a is extracted, and for the vehicle type B, only the sales region d is extracted. For the vehicle types C to J, only those in the sales region a or d are extracted.
The technical elements described in this specification or the drawings exhibit technical usefulness alone or in various combinations, and are not limited to the combinations described in the claims at the time of filing. In addition, the technology illustrated in the present specification or the drawings achieves a plurality of objects at the same time, and has technical utility by achieving one of the objects.
[Brief description of the drawings]
FIG. 1 is a diagram for explaining a relationship between a sales period of a prediction target vehicle type and a prediction target period and a relationship between a sales period of a data providing vehicle type.
FIG. 2 is a block diagram of a main configuration of a sales number prediction apparatus according to the first embodiment.
FIG. 3 is a block diagram of a main configuration of data preparation means of the first embodiment.
FIG. 4 is a diagram illustrating a data configuration example of actual sales numbers.
FIG. 5 is a diagram illustrating a data configuration example of a seasonal index.
FIG. 6 is a diagram illustrating an example of a ratio transition.
FIG. 7 is a diagram illustrating pattern classification by cluster analysis.
FIG. 8 is a diagram illustrating an example of standard transition.
FIG. 9 is a diagram illustrating an example of a correlation graph and a model formula.
FIG. 10 is a diagram showing an example of actual sales number, cumulative predicted sales number, and ratio transition of a prediction target vehicle type.
FIG. 11 is a flowchart of main processing of merchandise sales quantity prediction processing;
FIG. 12 is a flowchart of a seasonal index calculation process.
FIG. 13 is a flowchart of a ratio transition creation process.
FIG. 14 is a flowchart of a model formula generation process.
FIG. 15 is a flowchart of transition pattern selection processing.
FIG. 16 is a flowchart of predicted sales number calculation processing;
FIG. 17 is a flowchart of seasonality addition processing.
FIG. 18 is a block diagram of a main configuration of a sales number prediction apparatus according to a second embodiment.
FIG. 19 is a block diagram of the main configuration of data preparation means of the second embodiment.
FIG. 20 is a flowchart of a model formula generation process.
FIG. 21 is a flowchart of transition pattern selection processing;
FIG. 22 is a diagram illustrating an example of a cumulative graph and a model formula.
FIG. 23 is a diagram showing an example of actual sales number, cumulative predicted sales number, and ratio transition of a prediction target vehicle type.
[Explanation of symbols]
20: Sales number prediction device,
30: Data preparation means,
32: Seasonal index calculation means,
34: Seasonal adjustment means,
36: Ratio transition creation means,
38: Model formula generating means,
50: Sales number prediction means,
52: Transition pattern selection means,
54: Expected sales number calculation means,
56: Seasonal addition means,
72a: Actual sales number storage means,
76a: Transition storage means,
78a: Model formula storage means,
80a: Predicted sales quantity storage means,
82c: Seasonal additional forecast sales number storage means

Claims (6)

A device that predicts the number of sales by unit period from the start of sales of a new product until the second predetermined period elapses after the first predetermined period elapses;
Actual sales number storage means for storing a transition of actual sales number of new products by unit period divided into a plurality of first predetermined periods;
Transition that stores the standard transition by transition pattern of the value obtained by dividing the number of actual sales by unit period by the cumulative number of actual sales from the start of sales of existing products to the time when the first and second predetermined periods have elapsed Standard transition storage means by pattern,
A model formula storage means for storing a model formula for calculating an accumulated actual sales number from the cumulative actual sales number until the second predetermined period elapses until the first predetermined period elapses of the existing product;
Cumulative prediction from the model formula stored in the model formula storage means and the cumulative actual sales number stored until the first predetermined period elapses stored in the actual sales number storage means to the second predetermined period elapse of the new product A cumulative forecast sales number calculation means for calculating the number of sales;
A transition calculating means for calculating a transition of a value obtained by dividing the actual sales number for each unit period from the sales start time of the new product to the time when the first predetermined period has elapsed by the cumulative predicted sales number calculated by the cumulative predicted sales number calculating means; ,
Transition pattern selection means for selecting a transition pattern that is most similar to the transition calculated by the transition calculation means among the standard transitions by transition pattern stored in the standard transition storage means by transition pattern;
From the standard transition of the transition pattern selected by the transition pattern selection means and the cumulative predicted sales number of the new product calculated by the cumulative predicted sales number calculation means until the second predetermined period has elapsed, 2 Predicted sales number calculating means for calculating the predicted sales number for each unit period until a predetermined period of time,
Product sales number prediction device comprising: A device that predicts the number of sales by unit period from the start of sales of a new product until the second predetermined period elapses after the first predetermined period elapses;
Actual sales number storage means for storing a transition of actual sales number of new products by unit period divided into a plurality of first predetermined periods;
Transition that stores the standard transition by transition pattern of the value obtained by dividing the number of actual sales by unit period by the cumulative number of actual sales from the start of sales of existing products to the time when the first and second predetermined periods have elapsed Standard transition storage means by pattern,
A model formula storage means by transition pattern that stores model formulas for calculating the cumulative actual sales number from the cumulative actual sales number until the second predetermined period elapses until the second predetermined period elapses, for each transition pattern;
A transition pattern assumption means for assuming a transition pattern of a new product;
The model formula of the transition pattern assumed by the transition pattern assumption means stored in the model formula storage means by transition pattern, and the cumulative actual number of sales until the first predetermined period of time stored in the actual sales number storage means To the cumulative predicted sales number calculating means for calculating the cumulative predicted sales number until the second predetermined period of time for the new product,
A transition calculating means for calculating a transition of a value obtained by dividing the actual sales number for each unit period from the sales start time of the new product to the time when the first predetermined period has elapsed by the cumulative predicted sales number calculated by the cumulative predicted sales number calculating means; ,
A transition similarity calculation means for calculating the similarity of the transition pattern calculated by the transition calculation means, and the standard transition of the transition pattern assumed by the transition pattern assumption means, stored in the standard transition storage means by transition pattern;
A transition pattern selection means for selecting a transition pattern having the highest similarity calculated by the transition similarity calculation means;
From the standard transition of the transition pattern selected by the transition pattern selection means and the cumulative predicted sales number of the new product calculated by the cumulative predicted sales number calculation means until the second predetermined period has elapsed, 2 Predicted sales number calculating means for calculating the predicted sales number for each unit period until a predetermined period of time,
Product sales number prediction device comprising: This is a method for predicting the number of sales by unit period from when the first predetermined period has elapsed since the start of sales of a new product until the second predetermined period elapses from that point.
Reading the transition of the actual sales number of new products by unit period divided into a plurality of first predetermined periods;
Reading the standard transition by transition pattern of the value obtained by dividing the number of actual sales by unit period by the cumulative number of actual sales from the start of sales of existing products to the time when the first predetermined period and the second predetermined period have elapsed;
Reading a model formula for calculating an accumulated actual sales number from the cumulative actual sales number until the second predetermined period elapses until the first predetermined period elapses of the existing product;
Calculating a cumulative estimated sales number until the second predetermined period elapses of the new product from the read model formula and the accumulated actual sales number until the first predetermined period elapses read;
A step of calculating a transition of a value obtained by dividing the actual number of sales by unit period from the start of sales of a new product to the time when the first predetermined period has elapsed by the calculated cumulative sales number of sales;
Selecting a transition pattern that is most similar to the calculated transition among the standard transitions read by transition pattern;
Forecast by unit period from the first predetermined period to the second predetermined period from the standard transition of the selected transition pattern and the cumulative number of predicted sales of the new product until the second predetermined period has elapsed The process of calculating the number of sales;
Product sales number forecasting method to execute. This is a method for predicting the number of sales by unit period from when the first predetermined period has elapsed since the start of sales of a new product until the second predetermined period elapses from that point.
Reading the transition of the actual sales number of new products by unit period divided into a plurality of first predetermined periods;
Reading the standard transition by transition pattern of the value obtained by dividing the number of actual sales by unit period by the cumulative number of actual sales from the start of sales of existing products to the time when the first predetermined period and the second predetermined period have elapsed;
A step of reading a model formula for each transition pattern for calculating an accumulated actual number of sales until the second predetermined period elapses from an accumulated actual number of sales of the existing product until the first predetermined period elapses;
A process that assumes the transition pattern of new products;
Calculating the model expression of the assumed transition pattern and the cumulative actual sales number until the second predetermined period elapses from the read cumulative actual sales number until the first predetermined period elapses;
A step of calculating a transition of a value obtained by dividing the actual number of sales by unit period from the start of sales of a new product to the time when the first predetermined period has elapsed by the calculated cumulative sales number of sales;
Calculating the standard transition of the assumed transition pattern and the similarity of the calculated transition;
Selecting the transition pattern with the highest calculated similarity,
Forecast by unit period from the first predetermined period to the second predetermined period from the standard transition of the selected transition pattern and the cumulative number of predicted sales of the new product until the second predetermined period has elapsed The process of calculating the number of sales;
Product sales number forecasting method to execute. This is a program for predicting the number of sales by unit period from the time when the first predetermined period has elapsed since the start of sales of a new product until the second predetermined period elapses from that point. That is,
A process of reading the transition of the number of actual sales of new products by unit period divided into a plurality of first predetermined periods;
A process of reading the standard transition by transition pattern of the value obtained by dividing the number of actual sales by unit period by the cumulative number of actual sales from the start of sales of existing products to the time when the first predetermined period and the second predetermined period have elapsed;
A process of reading a model formula for calculating an accumulated actual sales number until the second predetermined period elapses from an accumulated actual sales number until the first predetermined period elapses of the existing product;
A process of calculating the read model formula and the cumulative estimated sales number until the second predetermined period elapses of the new product from the read cumulative sales number until the first predetermined period elapses;
A process of calculating a transition of a value obtained by dividing the actual number of sales by unit period from the start of sales of a new product to the time when the first predetermined period has elapsed, based on the calculated cumulative number of sales,
A process of selecting a transition pattern that is most similar to the calculated transition among the standard transitions read by transition pattern;
Forecast by unit period from the first predetermined period to the second predetermined period from the standard transition of the selected transition pattern and the cumulative number of predicted sales of the new product until the second predetermined period has elapsed Processing to calculate the number of sales,
A program for forecasting the number of products sold. This is a program for predicting the number of sales by unit period from the time when the first predetermined period has elapsed since the start of sales of a new product until the second predetermined period elapses from that point. That is,
A process of reading the transition of the number of actual sales of new products by unit period divided into a plurality of first predetermined periods;
A process of reading the standard transition by transition pattern of the value obtained by dividing the number of actual sales by unit period by the cumulative number of actual sales from the start of sales of existing products to the time when the first predetermined period and the second predetermined period have elapsed;
A process of reading a model formula for each transition pattern by transition pattern, calculating a cumulative actual sales number from the cumulative actual sales number until the second predetermined period elapses until the first predetermined period elapses of the existing product;
A process that assumes the transition pattern of new products,
A process of calculating a model expression of the assumed transition pattern and the cumulative actual sales number until the second predetermined period elapses from the read cumulative actual sales number until the first predetermined period elapses;
A process of calculating a transition of a value obtained by dividing the actual number of sales by unit period from the start of sales of a new product to the time when the first predetermined period has elapsed, based on the calculated cumulative number of sales,
A standard transition of the assumed transition pattern, a process of calculating the similarity of the calculated transition,
Processing to select the transition pattern with the highest calculated similarity,
Forecast by unit period from the first predetermined period to the second predetermined period from the standard transition of the selected transition pattern and the cumulative number of predicted sales of the new product until the second predetermined period has elapsed Processing to calculate the number of sales,
A program for forecasting the number of products sold.

Images