JP2008123371A - Goods demand predicting device, goods demand predicting method and program thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a goods demand predicting device, demand predicting method and program capable of changing a goods sales plan by flexibly reexamining the goods sales plan. <P>SOLUTION: The goods demand prediction program comprises: a goods information recording means for recording goods attributes and actual sales results to predict goods demand; a correlation coefficient calculating means 110 for calculating a correlation coefficient between actual sales results of a plurality of goods; a graph preparing and displaying means for grouping and displaying shipping flow patterns whose correlation coefficients are approximate in shipping flow patterns of the past actual sales results; a correction rule recording means for recording a correction rule; a graph preparing and displaying means for displaying the past shipping flow patterns as first predicted values by specifying goods attributes; an actual shipping result information recording means for recording actual shipping results of prediction object goods; a graph preparing and displaying means for synthesizing and displaying the actual shipping results so as to be comparable with the first predicted values; and a correcting means 111 for reading the actual shipping results of the prediction object goods from the actual shipping result information recording means, applying the correction rule to correct the first predicted values and obtaining new predicted values. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention shares the process of selecting a reference product to be used as a demand forecast model when forecasting demand for products of various varieties, and enables the review and correction of a short-term cycle of production and sales plans. The present invention relates to a demand forecasting apparatus, a demand forecasting method for merchandise, and a program thereof.

  Conventionally, the person in charge has been cultivating demand forecasts for products with a large number of items, especially cosmetics, which require management for each number of color codes even for the same item, and the product cycle is relatively short, for many years. Based on this experience, it was conducted in consideration of various factors such as the contents of promotional measures and the sales situation by channel.

  Specifically, from the same brand and the same category, taking into account the above factors, select a reference product that should be a model for demand forecast from past sales performance data, and estimate the planned quantity using that performance as a reference. Things were going on.

  According to such a conventional method, since it largely depends on the rule of thumb of the person in charge and there is no selection process of the reference products to be shared, the overall shipping level has been shared as a common recognition among the persons in charge. Even so, it is difficult to make fine adjustments from the initial shipment of the product to the initial, especially the variation in repeat shipments for 1 to 3 months. As a result, there has been a problem that it is difficult to correct the plan particularly at the start-up stage.

  Therefore, a demand prediction program and a demand prediction device as disclosed in Patent Document 1 and Patent Document 2 have been developed in order to share this demand prediction process. However, these conventional technologies are unsuitable for demand prediction of products with a wide variety of items and for reviewing and correcting short-term cycles. In other words, the demand forecasting method cannot be used if there is no sales record for a certain period after the product is released (Reference 2), and the product to be predicted increases as the number of forecast target products and similar similar products increases. Since it is complicated to search for and find the optimal shipping flow pattern (Reference 1), it will be bothersome to review similar products after release and correct the level as appropriate. Reflecting the individuality of the person performing the prediction work, the result would be inconsistent.

JP-A-6-68065 JP-A-2002-351956

  Therefore, the present invention overcomes such drawbacks of conventional demand forecasting systems, grasps and manages systematic sales data and demand forecasts, shares reference product selection processes, and flexibly reviews and changes plans in a short cycle. An object of the present invention is to provide a commodity demand forecasting apparatus, a commodity demand forecasting method, and a program thereof.

  The present invention solves the above-mentioned problems, and in order to predict the demand for goods, the computer records product attributes and past sales results in advance, product information recording means, and between the sales results of a plurality of products. Correlation coefficient calculation means to calculate the correlation coefficient of the past, graph creation display that displays on the monitor, such as a display, a graph of the past sales performance shipment flow patterns that are similar to the correlation coefficient grouping Means, a correction rule recording means for recording correction rules in advance, a graph creation display for graphing the past shipment flow pattern selected by specifying the product attribute as a first predicted value and displaying it on a monitor such as a display Means, shipping result information recording means for recording the actual shipment results of the prediction target product, and the shipment results for comparison with the first predicted value. A graph creation and display means for forming a graph and displaying it on a monitor such as a display, and reading the shipment results of the prediction target product from the shipment result information recording means, and the correction rule recorded in the correction rule recording means based on the shipment results It is a demand forecasting program of goods for applying and correcting the above-mentioned first prediction value and functioning as a correction means for obtaining a new prediction value.

  Further, the present invention solves the above-described problem, and is an apparatus for predicting the demand for a product, product information recording means for recording product attributes and past sales results in advance, Correlation coefficient calculation means for calculating correlation coefficients between sales results, and a group of the sales flow patterns of past sales results that have similar correlation coefficients, and display them on a monitor such as a display. Graph creation display means, correction rule recording means for recording correction rules in advance, past shipment flow pattern selected by specifying the product attribute is graphed as a first predicted value and displayed on a monitor such as a display Graph creation and display means, shipment result information recording means for recording the actual shipment results of the prediction target product, and the shipment results that can be compared with the first predicted value A graph creation / display unit that synthesizes and displays the result on a monitor such as a display, reads the shipment record of the prediction target product from the shipment record information recording unit, and stores the correction rule recorded in the correction rule recording unit based on the shipment record A commodity demand prediction apparatus comprising correction means for applying and correcting the first predicted value to obtain a new predicted value.

  Furthermore, the present invention sets the most appropriate one according to the attribute of the product from the past sales record as the first predicted value, and appropriately corrects the predicted difference between the predicted value and the actual shipment record of the forecast target product, thereby newly In a demand forecast method for a product that obtains a predictive value, a computer groups a plurality of products into a plurality of product groups according to their attributes in a tree structure and displays it on a monitor such as a display. The past sales information corresponding to the group, which are closely related to the correlation coefficient of the shipping flow pattern, are further grouped together and displayed on a monitor such as a display, and the selected first forecast value and forecast target product The shipping flow patterns of shipping results are combined so that they can be compared and displayed on a monitor such as a display, etc. Serial first predicted value is corrected, a demand prediction method, characterized by graphing display for displaying on a monitor such as a new predicted value comparable graphed and shipping result data display obtained by the correction.

The product demand forecasting apparatus, the product demand forecasting method and the program thereof according to the present invention include:
It is possible to share the process of selecting a reference product, and thus it is difficult for individual differences between persons in charge to occur in the process of demand prediction. In addition, by displaying the shipping flow patterns of the actual data of the reference product and the actual data of the forecast target product in a graph so that they can be compared, it is possible to easily grasp the actual difference between the initial shipment and repeat, and to change the predicted value. When appropriate, the short-term cycle of production and sales plans can be reviewed and revised.

  Hereinafter, the present invention will be described in more detail with reference to the drawings showing an embodiment of the present invention, but the present invention is not limited thereto.

  FIG. 1 is a block diagram showing an embodiment of a system including a demand modeling apparatus according to the present invention. In this figure, 1 is a demand modeling system (hereinafter referred to as “DMS” for short), 2 is a demand planning server, 3 is a production management server, 4 is a sales logistics management server, 5 is a data input terminal, 100 , 200, and 300 indicate networks. The DMS 1 includes an arithmetic processing device 11 and a storage device 12. The arithmetic processing device 11 includes a correlation coefficient calculation unit 110 and a correction unit 111, and the storage device 12 includes a product database 121, product shipping record data 122, Each of the correction rule storage units 123 is configured.

  As shown in FIG. 1, the DMS 1 is connected to the demand planning server 2 via a network 100. The demand planning server 2 is connected to the production management server 3 and the sales logistics management server 4 via the network 200, and the sales logistics management server 4 is further connected to the data input terminal 5 via the network 300. .

  The demand planning server 2 manages data such as shipment / sales of merchandise, and appropriately provides the data necessary for demand prediction to the DMS 1. The data is appropriately input from the data input terminal 5 via the sales / distribution management server 4. Specifically, from the data input terminal 5 installed in the manufacturer's shipping department, the manufacturer's research department, the dealer, etc. Collected by user input. This data collection means may use an existing system such as POS (Point Of Sales).

  Details of the DMS 1 of the present invention will be further described below. FIG. 2 shows a hardware configuration diagram of an embodiment of the DMS of the present invention.

  The DMS 1 includes an input device 201, a display device 202, a drive device 203, a recording medium 204, an auxiliary storage device 205, a memory device 206, an arithmetic processing device 207, which are mutually connected via a bus B. The interface device 208 is configured to have a database (hereinafter referred to as DB) 209.

  The input device 201 includes a keyboard and a mouse, and is used to input various operation instructions. The display device 202 displays various windows and data necessary for operation. The interface device 208 is an interface for connecting to the network 100, and includes, for example, a LAN adapter.

  The DB 209 manages various information (for example, product information, sales information, etc.) necessary for processing such as correlation coefficient calculation of the present invention. A DB server may be separately provided to manage information stored in the DB 209.

  The calculation program of the present invention is provided by a recording medium 204 such as a CD-ROM. The recording medium 204 on which the calculation program is recorded is set in the drive device 203, and the calculation program is installed from the recording medium 204 to the auxiliary storage device 205 via the drive device 203.

  The auxiliary storage device 205 stores the installed calculation program and files, data, and the like necessary for the calculation program processing. The memory device 206 reads and stores the calculation program from the auxiliary storage device 205 at the time of activation. The arithmetic processing unit 207 executes processing according to a calculation program stored in the memory device 206.

  Below, the outline of each step of the demand prediction by DMS of this invention is demonstrated with FIG. First, prior to the prediction work, a product grouping step (S1) is performed. That is, it refers to the work of grouping various products according to their attributes, and further grouping those grouped according to their attributes according to past shipment data to systematically form a tree structure. Details will be described later.

  Next, a prediction step (S2) is performed. Here, a product having a shipping flow pattern that is most satisfactory for the target product is selected as a reference product from the product group classified by the grouping step (S1). Here, the “target product” is a product that is a target of demand prediction, and the “reference product” is a product selected for reference in performing the demand prediction. The “shipment flow pattern” refers to a past shipment pattern of a product with a sales record, which is divided into initial shipment and repeat and is graphed in time series. Details will be described later.

  Next, an evaluation step (S3) is performed. In other words, the initial delivery period of the target product and the flow pattern of the reference product are compared, and the level of the flow pattern of the reference product is corrected or changed to a pattern with higher suitability according to the difference between the actual results. Details will be described later.

  Hereinafter, each of the above steps will be described in more detail with reference to FIGS. 4 to 6 and the DMS display screen of the present invention.

  FIG. 4 is a flowchart of the grouping step (S1). The processing in this step is performed according to the following procedure. First, registration and editing of product attributes such as a product master, product characteristics, and sales characteristics are performed (S101). Here, the “product master” refers to information for specifying a product such as a product name, a sales start date, and a price. “Product characteristics” refers to information corresponding to characteristics of a product such as a product category (point makeup or base makeup, etc.) and a functional class (such as lotion or foundation). “Sales characteristics” refers to sales information related to products such as product input characteristics (new series or series added, etc.) and presence / absence of promotions.

  Next, the past shipment result data of each product is registered (S102). Such shipment data is provided as appropriate from the sales management server 2, but of course, it may be directly input to the DMS 1 without using the demand planning server 2. Since the shipment data directly reflects the trend of products in the market, it is possible to predict future demand by collecting and analyzing such data. It should be noted that the past shipment data accumulation period can be arbitrarily determined, but it is desirable that it is 12 months or longer in order to improve the accuracy of demand prediction.

  Next, the pattern tree structure is defined and registered according to the attribute (S103). That is, on the pattern tree definition screen (FIG. 7), products are systematically classified according to their attributes, and are subdivided from higher concepts to lower concepts. FIG. 8 shows an example of such a tree structure. Hereinafter, in this example, each group classified from the top-level brand B to the level of <release date> is referred to as “group A”.

  Next, those classified into the group A are finally classified into groups having the same number of color codes having similar repeat flow patterns (S104). Such processing is automatically performed by the arithmetic processing unit 17 of the DMS 1, and specifically, proceeds in the following procedure. That is, when a new grouping is performed, those having similar waveforms (correlation coefficients) of repeat flow patterns are classified into the same group. To determine whether the correlation coefficient is close, for example, the correlation between the initial shipment quantity after release and the cumulative shipment quantity after one year is analyzed, and the correlation coefficient is 0.7 or more. It is defined by defining that the repeat flow patterns are similar, and automatically searching for and registering those corresponding to this definition. In this example, if the correlation coefficient is 0.7 or less, remove the product whose correlation coefficient is lowered from the group and assign it to another group, or create a new group at the end of the tree. .

  By such classification work, group A is classified into a set of color numbers (product numbers indicating product colors) having similar repeat flow patterns. Hereinafter, the group of the lowest color number is referred to as “group B”. Then, each group classified into group B, in the example of FIG. 8, by specifying a group composed of color code numbers # 429, 430, 458, 459, 623, 656, Past repeat flow patterns of products of each color number are synthesized and displayed in a graph (S105). This completes the flow pattern systematization step (S1) performed prior to the prediction work.

  FIG. 5 is a flowchart (S2) of the prediction step. A series of processes will be described with reference to this figure. First, the pattern tree is searched according to the attribute of the target product on the pattern tree search screen (FIG. 9), and the corresponding group B is selected (S201). Specifically, according to the product characteristics, sales characteristics, etc. of the target product, the pattern tree of the group A hierarchy is traced from the upper layer to the lower layer, and finally to a specific group B depending on the number of applicable color codes. Complete the sorting.

  Next, the repeat flow pattern of the group of the corresponding color number is displayed and output as necessary (S202). Here, if the special circumstances due to missing items are reflected as they are in the selected flow pattern, it is desirable to perform data cleansing work (S203). That is, on the pattern correction screen (FIG. 10), it is possible to correct a change in the repeat flow pattern caused by a shortage or the like and eliminate the influence of special circumstances.

  Since the flow pattern of past shipments of reference products is displayed in a graph on the pattern correction screen (Fig. 10), production and sales plans for new products can of course be made based on such graphs, but total amount prediction is required. In this case, the total amount prediction evaluation displayed on the pattern tree evaluation screen (FIG. 11) can be referred to. This total amount prediction is basically performed by regression analysis. That is, a correlation analysis is performed on the relationship between the initial sales results and the results after 3 to 12 months, and the relationship is made a linear equation. By registering the partial regression coefficient calculated by the regression analysis in the tree, the total number of shipments after 3 to 12 months is calculated when the result of the prediction target product occurs.

  With the above operation, the prediction creation step (S2) is completed. As described above, the user can proceed with the selection work while systematically grasping various types of products by the tree display classified by the attribute, so that the selection process of the reference product is shared.

  FIG. 6 is a flowchart (S3) of the evaluation step. In this step, first, shipment result data (first shipment, repeat) of the target product is input (S301). Such performance data is provided as appropriate from the demand planning server 2, but of course, it may be input directly to the DMS 1 without using the sales management server 2.

  The input performance data is graphed and synthesized and displayed together with the repeat flow pattern of the reference product (FIG. 11). Here, the performance of the target product, particularly the performance immediately after the sale and the flow pattern of the reference product are compared (S302), and if there is a difference, the predicted value is corrected (S303). For example, when the level of the performance period of the target product is 40% lower than the flow pattern of the reference product, the reference product flow pattern is not changed and the overall level is lowered by 40% (FIG. 12). In addition, if the flow pattern in the performance of the target product is different from the flow pattern of the reference product, select a pattern with higher suitability, that is, change the reference product to a new predicted value to improve the prediction accuracy. (FIG. 13).

  Selection of a similar flow pattern is obtained by the same correlation analysis as in the shipping flow pattern systematization step (S104). That is, the relationship between the flow pattern of the target product and the flow pattern of the reference product is measured by the correlation coefficient, and the reference product having the highest value, that is, the most similar flow pattern is automatically searched. Here, the search target is not limited to those belonging to the same B group, but may be other B groups, for example, those belonging to other A group systems. That is, even if a product belongs to a different product category, the shipment flow pattern can be selected as a reference product if the correlation with the flow pattern in the performance of the target product is high.

  Furthermore, as a predicted value correction, correction by a distance function using the Euclidean distance can be performed. That is, the already set correction coefficient is automatically recalculated so that the total difference between the monthly or weekly first predicted value and the actual result is minimized, and a new predicted value is calculated.

  Such an evaluation step (S3) is desirably performed periodically. For example, if the prediction value correction (S303) is performed in a weekly cycle of one week or every two weeks, supply and demand monitoring and prediction value correction are performed immediately after release. It is more preferable because the accuracy of work can be improved.

The block diagram which shows one Example of the system containing the demand modeling apparatus in this invention. The hardware block diagram of one Example of the demand modeling apparatus in this invention. Schematic of each step of the demand prediction method in the present invention. The schematic of the grouping step of the demand prediction method in this invention. The schematic of the prediction step of the demand prediction method in this invention. The schematic diagram of the evaluation step of the demand prediction method in the present invention. The figure which showed the pattern tree definition screen in this invention. The figure which showed an example of the tree structure in this invention. The figure which showed the pattern tree search screen in this invention. The figure which showed the pattern correction screen in this invention. The figure which showed the pattern tree evaluation screen in this invention. The figure which showed the example of application of the level correction of the correction rule in this invention. The figure which showed the example of application of the flow pattern change of the correction rule in this invention.

Explanation of symbols

1… Demand Modeling System
2…… Sales management server 3…… Production management server 4…… Logistics management server 5…… Data input terminal 11…… Processing unit 12…… Storage device 100…… Network 200…… Network 300…… Network 110…… Correlation coefficient calculation means 111…… Correction means 121…… Product database 122…… Product shipment record data 123…… Correction rule storage unit 201…… Input device 202…… Display device 203…… Drive device 204…… Recording medium 205 ... ... Auxiliary storage device 206 ... ... Memory device 207 ... ... Arithmetic processing device 208 ... ... Interface device 209 ... ... Database B ... ... Bus
more than

Claims (14)

Computer to predict demand for goods,
Product information recording means for recording product attributes and past sales results in advance,
Correlation coefficient calculating means for calculating a correlation coefficient between sales results of a plurality of products,
A graph creation and display means for grouping together those having similar correlation coefficients in the past sales performance shipment flow pattern and displaying it on a monitor such as a display.
Correction rule recording means for recording correction rules in advance;
Graph creation display means for graphing the past shipment flow pattern selected by specifying the product attribute as a first predicted value and displaying it on a monitor such as a display,
Shipment record information recording means for recording the actual shipment record of the forecast target product,
A graph creation and display means for composing and graphing the shipment results so that they can be compared with the first predicted value, and displaying them on a monitor such as a display,
The shipment record of the prediction target product is read from the shipment record information recording unit, the correction rule recorded in the correction rule recording unit is applied based on the shipment record, the first predicted value is corrected, and a new predicted value is obtained. Correction means to obtain,
Demand forecasting program for products to function as. A demand forecasting program for merchandise according to claim 1,
In addition, computers to predict demand for goods,
A tree information recording means for recording a grouped product of a plurality of products according to pre-recorded product attributes and recording the tree structure;
Means for displaying the tree information on a monitor such as a display;
Demand forecasting program for products to function as.   The product according to claim 1, wherein the correction rule for correcting the first predicted value to obtain a new predicted value is correction of a level in the entire predicted value. Demand forecasting program.   The above correction rule for correcting the first forecast value to obtain a new forecast value calculates the correlation coefficient of the shipment flow pattern in the past sales performance of multiple products, and the sales information that the correlation coefficient is more approximate 3. The demand forecasting program for merchandise according to claim 1, wherein the demand forecasting value is selected as a second forecast value.     The above correction rule for correcting the first predicted value to obtain a new predicted value recalculates a preset correction coefficient so that the total difference between the first predicted value and the actual result is minimized. The product demand prediction program according to claim 1, wherein the program is a correction for calculating a new predicted value. An apparatus for predicting the demand for goods,
Product information recording means for recording product attributes and past sales results in advance,
Correlation coefficient calculating means for calculating a correlation coefficient between sales results of a plurality of products,
A graph creation and display means for grouping together those having similar correlation coefficients in the past sales performance shipment flow pattern and displaying it on a monitor such as a display.
Correction rule recording means for recording correction rules in advance;
Graph creation display means for graphing the past shipment flow pattern selected by specifying the product attribute as a first predicted value and displaying it on a monitor such as a display,
Shipment record information recording means for recording the actual shipment record of the forecast target product,
A graph creation and display means for composing and graphing the shipment results so that they can be compared with the first predicted value, and displaying them on a monitor such as a display,
The shipment record of the prediction target product is read from the shipment record information recording unit, the correction rule recorded in the correction rule recording unit is applied based on the shipment record, the first predicted value is corrected, and a new predicted value is obtained. Correction means to obtain,
An apparatus for predicting demand for goods. A tree information recording means for recording a plurality of products grouped according to their attributes in a tree structure and recording;
Means for displaying the tree information on a monitor such as a display;
Graph creation display means for graphing the past shipment flow pattern selected by specifying the product attribute by the tree information as a first predicted value and displaying it on a monitor such as a display,
The demand forecasting device for goods according to claim 6 characterized by things.   The product according to claim 6, wherein the correction rule for correcting the first predicted value to obtain a new predicted value is correction of a level in the entire predicted value. Demand forecasting device.   The above correction rule for correcting the first predicted value to obtain a new predicted value calculates the correlation coefficient between the sales information of a plurality of products, and sets the sales information whose correlation coefficient is more approximate to the second predicted value. The demand prediction apparatus for merchandise according to claim 6 or 7, wherein:     The above correction rule for correcting the first predicted value to obtain a new predicted value recalculates a preset correction coefficient so that the total difference between the first predicted value and the actual result is minimized. The product demand forecasting apparatus according to claim 6, wherein the forecast is a correction for calculating a new forecast value. A product that obtains a new forecast value by appropriately correcting the forecast difference between the forecast value and the actual shipment performance of the forecasted product as the first forecast value from the past sales record that is most appropriate according to the product attributes In the demand forecast method of
The computer pre-groups multiple products into multiple product groups according to their attributes and displays them on a monitor such as a display.
The past sales information corresponding to the product group in the tree information above, which is similar to the correlation coefficient of the shipping flow pattern, are further grouped together and displayed on a monitor such as a display,
The selected first forecast value and the shipment flow pattern of the shipment performance of the forecast target product are combined for comparison and displayed on a monitor such as a display.
According to the shipment results, the first predicted value is corrected by a correction rule recorded in advance,
A demand forecasting method characterized in that a new forecast value obtained by correction is graphed so that it can be compared with shipment result data and displayed on a monitor such as a display.   The correction rule calculates a level difference between the first predicted value and the actual shipping result of the prediction target product, and corrects the level of the shipping actual data of the predicted target product according to the level of the first predicted value. The demand prediction method according to claim 11.   12. The correction rule according to claim 11, wherein the correction rule calculates a correlation coefficient between the first predicted value and the actual shipment of the prediction target product, and extracts a flow pattern of another reference product having a higher correlation. Demand forecast method. From the past sales results, the most appropriate one according to the product attributes is used as the first forecast value, and the forecast value and the actual shipment performance of the forecast target product are appropriately corrected to obtain a new forecast value. In the demand forecast method,
The computer calculates the correlation coefficient between the forecast value and the actual shipment of the forecasted product for each preset period, selects other reference products with higher correlation, and ships the selected shipment flow pattern A demand forecasting method, wherein a forecast value is sequentially updated by repeating a graph creating and displaying step of making a graph comparably with actual data and displaying it on a monitor such as a display.