Sales forecast adjustment device and sales forecast adjustment method

Abstract

To correct sales forecast in consideration of the influence of an unexpected event.SOLUTION: A sales forecast correction apparatus 100 includes a prediction unit 114 which calculates a degree of influence on sales forecast of a correction target company, using a machine learning technique, based on a result of aggregating relationship attributes between the correction target company for which sales forecast is to be corrected and a news article. The prediction unit 114 calculates the influence on the sales forecast of the correction target company with respect to a customer. The sales forecast correction apparatus 100 corrects sales forecast for the customer so that the sales forecast of the correction target company calculated based on the influence may be coincident with the sum of the sales forecast for the customer.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a sales forecast correction device and a sales forecast correction method for correcting a sales forecast by referring to news.

Background Art

[0002] As a technique for predicting a company's performance, Patent Document 1 discloses a performance prediction device including: a performance information acquisition unit that acquires performance information indicating the past performance of a company; a usage information acquisition unit that acquires usage information of a mobile terminal used by an employee of the company; a prediction model generation unit that generates a prediction model for predicting the future performance of the company from the usage information, with the performance information as the target variable and the usage information as the explanatory variable; and a prediction unit that predicts the future performance of the company from the usage information using the prediction model.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The performance prediction device described in Patent Document 1 does not include sudden events as information for prediction, and it is considered difficult to make a prediction considering the influence of, for example, a change in the performance forecast of a customer company or the occurrence of a disaster. The present invention has been made in view of such a background, and an object thereof is to provide a sales forecast correction device and a sales forecast correction method that enable correction of a sales forecast considering the influence of sudden events.

Means for Solving the Problems

[0005] To solve the above-mentioned problems, the sales forecast correction device according to the present invention includes the aggregated results of the relationship between the target company for correction and news articles as explanatory variables, and calculates the degree of impact on the sales forecast of the target company using a prediction model for correction targets, in which the degree of impact on the sales forecast is the dependent variable. The forecast model for demand destinations, which includes the aggregated results of the relationship between the demand destinations of the adjusted company and news articles as explanatory variables and has the degree of influence of the adjusted company's sales forecast on the demand destinations as the dependent variable, is used to calculate the degree of influence of the adjusted company's sales forecast on the demand destinations. Prediction unit A correction unit calculates a correction rate such that the sum of the sales forecasts of the company to be corrected for the customer, calculated based on the sales forecast of the company to be corrected for the customer and the degree of influence of the company to be corrected for the customer, is equal to the sales forecast of the company to be corrected, calculated based on the sum of the sales forecasts of the company to be corrected for the customer and the degree of influence of the company to be corrected for the customer, and corrects the sales forecast of the company to be corrected for the customer using the correction rate. It is equipped with. [Effects of the Invention]

[0006] According to the present invention, it is possible to provide a sales forecast correction device and a sales forecast correction method that enable the correction of sales forecasts to take into account the impact of unforeseen events. Problems, configurations, and effects other than those described above will be clarified by the following description of embodiments. [Brief explanation of the drawing]

[0007] [Figure 1] This is a functional block diagram of the sales forecast correction device according to this embodiment. [Figure 2] This is a data structure diagram of a news article table to which related attributes have been assigned according to this embodiment. [Figure 3] This is a data structure diagram showing the aggregated results of the news article table according to this embodiment. [Figure 4] This is a flowchart of the sales forecast adjustment process according to this embodiment. [Figure 5] This is a screen configuration diagram of the data input screen for prediction according to this embodiment. [Figure 6] This is a screen configuration diagram of the data input screen for prediction according to this embodiment. [Figure 7] This is a screen configuration diagram of the learning data input screen according to this embodiment. [Figure 8] This is a screen configuration diagram of the learning data input screen according to this embodiment. [Figure 9] This is a flowchart of the prediction model generation process for the target company according to this embodiment. [Figure 10] This is a flowchart of the process for generating a predictive model for customer companies according to this embodiment. [Figure 11] This is a flowchart of a prediction model generation process for the demand target industry type according to this embodiment. [Figure 12] This is a flowchart of a sales forecast calculation process for a company to be corrected according to this embodiment. [Figure 13] This is a flowchart of a sales forecast correction process for a demand target company according to this embodiment. [Figure 14] This is a flowchart of a sales forecast correction process for the demand target industry type according to this embodiment. [Figure 15] This is a flowchart of a news extraction process with a large degree of influence according to this embodiment. [Figure 16] This is a screen configuration diagram of a sales forecast correction result display screen according to this embodiment. [Figure 17] This is a screen configuration diagram of a sales forecast correction result display screen according to this embodiment. [Figure 18] This is a diagram showing the screen configuration of a time change display screen of the influence amount according to this embodiment. [Figure 19] This is a flowchart of a time change display process of the influence amount display screen according to this embodiment.

Mode for Carrying Out the Invention

[0008] ≪Outline of Sales Forecast Correction Device≫ The sales forecast correction device corrects the sales forecast of a company by using a machine learning model that includes the aggregation result of news articles (simply referred to as news) as an explanatory variable. The aggregation result includes the number of news articles related to the company in terms of industry type, region, scale, etc., and the average length of the period from the news release to the sales performance calculation date (for example, the end of the fiscal year).

[0009] The sales forecast correction device corrects the sales forecast for each demand target such as customer companies and customer industry types. In addition, it identifies and displays news articles that have a large impact on the sales forecast correction for each demand target. Users of such a sales forecast correction device will be able to correct the sales forecast for unexpected events. Also, by referring to news articles with a high degree of influence on the correction, it will be possible to confirm the likelihood of the correction.

[0010] ≪Configuration of Sales Forecast Correction≫ FIG. 1 is a functional block diagram of a sales forecast correction device 100 according to the present embodiment. The sales forecast correction device 100 is a computer and includes a control unit 110, a storage unit 120, and an input / output unit 180. User interface devices such as a display, a keyboard, and a mouse are connected to the input / output unit 180. The input / output unit 180 includes a communication device and can transmit and receive data to and from devices such as a news providing server. Also, a media drive may be connected to the input / output unit 180 to enable news exchanges using a recording medium.

[0011] ≪Storage Unit≫ The storage unit 120 is configured to include storage devices such as a ROM (Read Only Memory), a RAM (Random Access Memory), and an SSD (Solid State Drive). The storage unit 120 stores a news information database 130, an IR information database 140, a corporate information database 150, a prediction model database 160, and a program 128.

[0012] The news information database 130 stores news articles acquired by the sales forecast correction device 100. The IR information database 140 stores IR information such as the initial sales forecast and the actual sales results of a company. The corporate information database 150 includes information related to the company such as the industry type, product type, location of the head office and factory, business area, and number of employees. The prediction model database 160 includes a machine learning model used for correcting the sales forecast. The program 128 includes a description of a sales forecast correction process (see FIG. 4) described later.

[0013] ≪Control Unit: Information Acquisition Unit≫ The control unit 110 includes a CPU (Central Processing Unit) and comprises an information acquisition unit 111, a news aggregation unit 112, a learning unit 113, a forecast unit 114, a sales forecast correction unit 115, an impact amount calculation unit 116, and a screen control unit 117. The information acquisition unit 111 acquires news articles from, for example, news provider websites and stores them in the news information database 130. The information acquisition unit 111 assigns attributes such as the news article's identification number, media (source), number of characters, acquisition date, and category (for example, new product announcement or accident information) to the news information database 130 before storing it.

[0014] ≪Control Unit: News Aggregation Department≫ The news aggregation unit 112 generates news aggregation results 320 (see Figure 3 below) related to the target company or demand destination (the company, industry, size, or region that is the demand destination). In more detail, the news aggregation unit 112 extracts news articles acquired during a predetermined period from the news information database 130. Next, the news aggregation unit 112 assigns relevant attributes (related information) to each extracted article according to the target company or demand destination and aggregates them.

[0015] Figure 2 is a data structure diagram of the news article table 310 to which the relevant attributes are assigned according to this embodiment. The news article table 310 is tabular data, and each row represents a news article. The columns (attributes) of the news article table 310 are described below.

[0016] Identification information (labeled ID (identifier) ​​in Figure 2), medium, number of characters, and category are obtained from news information database 130. The elapsed days are the number of days from the date the news was obtained to the end of the fiscal year. Related attributes include positive / negative classification, company flag, industry flag, size flag, and region flag. In Figure 2, positive / negative classification, company flag, industry flag, size flag, and region flag are labeled as P / N (positive / negative), company F, industry F, size F, and region F, respectively.

[0017] The affirmative / negative classification indicates whether the news article is positive (P), negative (N), or neutral (-) for the target company or the target company (the company, industry, size, or region that is the target company). The company flag indicates whether the news article relates to any company and is related to the target company or the target company (Y) or not (N). The industry flag indicates whether the news article relates to any industry and is related to the target company or the target company (Y) or not (N). The size flag indicates whether the news article relates to any size and matches the target company or the target company (Y) or not (N). The region flag indicates whether the news article relates to any region and is related to the target company or the target company (Y) or not (N). The news aggregation unit 112 uses natural language processing technology and artificial intelligence technology to determine whether the news is related to the target company or the target company.

[0018] Next, the news aggregation unit 112 aggregates the news article table 310 to generate the aggregation result 320 (see Figure 3 below). Figure 3 is a data structure diagram of the aggregation result 320 of the news article table 310 according to this embodiment. The total number of items is the total number of news articles, which is the number of rows in the news article table 310. The average number of characters is the average number of characters in the news articles. The average number of days elapsed is the average number of days elapsed for the news articles. The category (financial results, product releases, disasters) is the number of news articles for each category.

[0019] P represents the number of news articles with a positive / negative content classification of P, N represents the number of news articles with a negative / negative content classification of N, F represents the number of news articles with a company flag of Y, F represents the number of news articles with an industry flag of Y, F represents the number of news articles with a size flag of Y, and F represents the number of news articles with a region flag of Y.

[0020] As explained above, the relevant attributes include at least one of the following: the industry related to the content of the news article and its relationship to the company or customer; the scale related to the content of the news article and its relationship to the company or customer; and the region related to the content of the news article and its relationship to the company or customer. Furthermore, the aggregate result 320 includes the average length of the period from the news announcement to the date of sales performance calculation (end of the fiscal year).

[0021] ≪Control Unit: Learning Unit≫ Returning to Figure 1, let's continue the explanation of the control unit 110. The learning unit 113 generates training data with the aggregated results 320 (see Figure 3) for each of the target companies or customers as explanatory variables and the impact on sales as the dependent variable. The impact on sales is the ratio of the difference between actual sales and sales forecast to the sales forecast, and the impact = (actual sales - sales forecast) / sales forecast. Next, the learning unit 113 uses this training data to train and generate a predictive model using machine learning technology and stores it in the predictive model database 160.

[0022] The forecasting models include models that predict the impact on companies subject to sales forecast adjustments and models that predict the impact on the customers of those companies. There are four types of models for predicting the impact on customers, depending on whether the customers are classified as individual companies, or by industry, size, or region. Combined with the models that predict the impact on companies subject to sales forecast adjustments, there are five types of models. Below, these forecasting models will be referred to as models for companies subject to adjustments, models for customer companies, models for customer industry, models for customer size, and models for customer region.

[0023] As explained above, the demand source is one of the following: a company, a company by industry, a company by size, or a company by region.

[0024] Control Unit: Prediction Unit, Sales Forecast Correction Unit The forecasting unit 114 calculates the impact on sales using a forecasting model. More specifically, the news aggregation unit 112 generates aggregation results 320 (see Figure 3) from news articles from the beginning of the fiscal year, and the forecasting unit 114 uses the aggregation results 320 as explanatory variables (inputs) to calculate the impact using a forecasting model.

[0025] As described above, the sales forecast correction device 100 includes a forecasting unit 114 that calculates the degree of impact on the sales forecast of a company to be corrected using a forecasting model for correction targets (a model for correction target companies) which includes the aggregated results 320 of the related attributes between the correction target company and news articles as explanatory variables and the degree of impact on the sales forecast as the dependent variable. Furthermore, the forecasting unit 114 includes the aggregated results 320 of the relationship attributes between the target company's customers and news articles as explanatory variables, and calculates the degree of impact on the target company's sales forecast to customers using a forecasting model for customers (model for customer companies, model for customer industries, model for customer size, and model for customer regions), with the degree of impact on the target company's sales forecast to customers as the dependent variable.

[0026] The sales forecast adjustment unit 115 adjusts the sales forecast based on the impact calculated by the forecasting unit 114. More specifically, the sales forecast adjustment unit 115 adjusts the sales forecast of a company based on the impact of that company calculated by the forecasting unit 114. Next, the sales forecast adjustment unit 115 adjusts the sales forecast of the customer based on the impact of the customer of the company to be adjusted, calculated by the forecasting unit 114, and the adjusted sales forecast of that company. The processing of the sales forecast adjustment unit 115 will be described later with reference to Figures 13 and 14.

[0027] ≪Control Unit: Impact Amount Calculation Unit, Screen Control Unit≫ The impact calculation unit 116 calculates the impact of news articles on companies, industries, sizes, or regions that are the target of demand. The processing of the impact calculation unit 116 will be explained with reference to Figure 19. The screen control unit 117 controls the display of various screens (see Figures 5 to 8 and 16 to 18 below).

[0028] <<Sales forecast adjustment process>> Figure 4 is a flowchart of the sales forecast correction process according to this embodiment. The sales forecast correction process, including the generation of the forecast model, the correction of the sales forecast, and the display of the correction results, will be explained with reference to Figure 4. In the drawings, including Figure 4, news articles will also be simply referred to as news.

[0029] In step S11, the prediction unit 114 instructs the screen control unit 117 to display the prediction data input screens 410 and 420 (see Figures 5 and 6 below) and acquire prediction data. Figure 5 is a screen configuration diagram of the prediction data input screen 410 according to this embodiment. At the top of the prediction data input screen 410 is an area 411 for entering the names of companies that are subject to sales forecast correction and are the companies for which sales are forecasted.

[0030] Below area 411 is area 412, where the input type for the sales forecast of the customer before correction is selected. The input types include sales forecast by customer company, sales forecast by customer industry (by industry of the customer company), sales forecast by customer size (by size of the customer company), and sales forecast by customer region (by region of the customer company). Users of the sales forecast correction device 100 select one of these four to input the sales forecast. On the forecast data input screen 410, sales forecast by customer company is selected, and the sales forecast by customer company is entered in area 413 below area 412. The sum of these sales forecasts by customer company becomes the sales forecast for the company subject to sales forecast correction (the company entered in area 411).

[0031] Figure 6 is a diagram of the screen configuration of the forecast data input screen 420 according to this embodiment. On the forecast data input screen 420, the sales forecast by industry of the demand destination is selected (see area 422), and the sales forecast by industry of the demand destination company is entered in area 423. The sum of these industry-specific sales forecasts becomes the sales forecast of the company subject to sales forecast correction (the company entered in area 421).

[0032] Returning to Figure 4, let's continue the explanation of the sales forecast correction process. In step S12, the forecasting unit 114 proceeds to step S16 if there is a forecasting model in the forecasting model database 160 (step S12 → YES), and to step S13 if there is no forecasting model (step S12 → NO). The forecasting unit 114 determines the presence or absence of a forecasting model depending on the input of the sales forecast before correction selected on the forecasting data input screens 410 and 420 (see areas 412 and 422). If a sales forecast by customer company is selected, the forecasting unit 114 determines whether the model for the company to be corrected and the model for the customer company exist in the forecasting model database 160. If a sales forecast by industry of customer company is selected, the forecasting unit 114 determines whether the model for the company to be corrected and the model for the customer industry exist in the forecasting model database 160. The same applies if a sales forecast by size of customer company or a sales forecast by region of customer company is selected.

[0033] In step S13, the learning unit 113 instructs the screen control unit 117 to display the learning data input screens 430 and 440 (see Figures 7 and 8 below) and acquire learning data. Figure 7 is a screen configuration diagram of the learning data input screen 430 according to this embodiment. On the learning data input screen 430, sales forecasts for each customer company are selected (see area 432), and the sales forecasts and actuals for each customer company are entered in area 433. The sum of these sales forecasts and the sum of the actuals become the sales forecasts and actuals for the companies subject to sales forecast correction (companies entered in area 431).

[0034] In Figure 7, the forecast and actual results for fiscal year 2020 are entered, but it is desirable to also enter the forecast and actual results for fiscal year 2020 and earlier to increase the amount of training data. Figure 8 is a diagram of the screen configuration of the learning data input screen 440 according to this embodiment. On the learning data input screen 440, the sales forecast for each industry of the demand destination is selected (see area 442), and the sales forecast and actual sales for each industry of the demand destination are entered in area 443.

[0035] Returning to Figure 4, let's continue the explanation of the sales forecast adjustment process. In step S14, the learning unit 113 generates a model for the company to be adjusted. Details of step S14 will be described later with reference to Figure 9. In step S15, the learning unit 113 generates a model for the customer company, a model for the customer industry, a model for the customer size, or a model for the customer region, depending on the input type of sales forecast for the customer (see area 412 in Figure 5). Details of step S15 will be described later with reference to Figures 10 and 11.

[0036] In step S16, the forecasting unit 114 calculates the degree of impact using the model for the company to be corrected, and uses this degree of impact to calculate the sales forecast for the company to be corrected. Details of step S16 will be described later with reference to Figure 12. In step S17, the forecasting unit 114 calculates the impact using a model for the customer company, a model for the customer industry, a model for the customer size, or a model for the customer region, depending on the input type of sales forecast for the customer (see area 412 in Figure 5). Next, the sales forecast correction unit 115 corrects the sales forecast for the customer company, customer industry, customer size, or customer region using this impact. Details of step S17 will be described later with reference to Figures 13 and 14.

[0037] In step S18, the impact calculation unit 116 extracts news articles with a high impact for each company that is a customer. Details of step S18 will be described later with reference to Figure 15. In step S19, the screen control unit 117 displays the sales forecast correction result display screen 450 (see Figure 16 below).

[0038] <<Predictive model generation process for companies subject to correction>> Figure 9 is a flowchart of the prediction model generation process for the company to be corrected according to this embodiment. The process of generating the model for the company to be corrected in step S14 (see Figure 4) will be explained with reference to Figure 9. In step S31, the learning unit 113 starts a process that repeats steps S32 to S35 for each past fiscal year (hereinafter referred to as "the relevant fiscal year") related to the company subject to correction (for which sales forecasts and actual sales data can be obtained).

[0039] In step S32, the news aggregation unit 112 starts the process of repeating step S33 for each news article. This news article (hereinafter referred to as "the news article") is a news article acquired within the relevant fiscal year. In step S33, the news aggregation unit 112 assigns attributes to the news article for the company to be corrected (refer to the relevant attributes, elapsed days shown in Figure 2, affirmative / negative classification, company flag, industry flag, size flag, and region flag). For example, the elapsed days is the number of days from the news acquisition date to the end of the fiscal year of the company to be corrected, and the affirmative / negative classification indicates whether the article is positive / negative / neutral for the company to be corrected.

[0040] In step S34, the news aggregation unit 112 aggregates news articles and generates aggregation results 320 (see Figure 3). In step S35, the learning unit 113 calculates the degree of impact on the sales forecast for the fiscal year of the company to be corrected ((actual sales - forecast) / forecast). The aggregated result 320 calculated in step S34 is used as the explanatory variable, and the degree of impact is used as the dependent variable. Each combination of the aggregated result 320 and the degree of impact becomes one training data point. In step S36, the learning unit 113 generates a model for the target company using training data in which the aggregated results 320 generated in step S34 are used as explanatory variables and the influence level calculated in step S35 is used as the target variable.

[0041] <<Process for generating predictive models for customer companies>> Figure 10 is a flowchart of the predictive model generation process for customer companies according to this embodiment. The process of generating the model for customer companies in step S15 (see Figure 4) will be explained with reference to Figure 10. In step S41, the learning unit 113 starts the process of repeating steps S42 to S46 for each company that will be the customer. Here, the companies that will be the customer are the companies located in area 433 as shown in Figure 7.

[0042] Steps S42 to S46 are the same as steps S31 to S35 shown in Figure 9. However, the companies subject to adjustment are read as companies that are customers. In step S47, the learning unit 113 generates a model for customer companies using training data in which the aggregated results 320 generated in step S45 are used as explanatory variables and the influence calculated in step S46 is used as the target variable.

[0043] ≪Predictive model generation process for demand sector, size, and region≫ Figure 11 is a flowchart of the process for generating a forecast model for customer industries according to this embodiment. The process for generating the customer industry model in step S15 (see Figure 4) will be explained with reference to Figure 11. The process for generating the customer size model and the customer region model is similar, provided that customer industries are replaced with customer size and customer region, respectively.

[0044] In step S51, the learning unit 113 starts the process of repeating steps S52 to S57 for each industry. Here, the industry refers to the customer industry located in area 443 as shown in Figure 8. In step S52, the learning unit 113 starts a process that repeats steps S53 to S57 for each company belonging to the industry.

[0045] Steps S53 to S57 are the same as steps S31 to S35 shown in Figure 9. However, the companies subject to correction are read as companies belonging to the industry. In step S58, the learning unit 113 generates a model for the demand-side industry using training data in which the aggregated results 320 generated in step S56 are used as explanatory variables and the influence calculated in step S57 is used as the target variable.

[0046] <<Calculation process for sales forecasts of companies subject to adjustment>> Figure 12 is a flowchart of the sales forecast calculation process for the company to be corrected according to this embodiment. The sales forecast calculation process for the company to be corrected in step S16 (see Figure 4) will be explained with reference to Figure 12. In step S61, the news aggregation unit 112 starts the process of repeating step S62 for each news article. These news articles are news articles acquired in the year for which the sales forecast is revised, or news articles from the past year. Steps S62 and S63 are the same as steps S33 and S34 (see Figure 9).

[0047] In step S64, the prediction unit 114 uses the aggregated results 320 (see Figure 3) generated in step S63 as explanatory variables (inputs) and calculates the degree of impact using the model for the corrected company. In step S65, the sales forecast correction unit 115 calculates a corrected sales forecast, which is the product of the impact calculated in step S64 and the sales forecast (the sum of the forecasts in areas 413 and 423 (see Figures 5 and 6)).

[0048] <<Revision processing of sales forecasts for customer companies>> Figure 13 is a flowchart of the sales forecast adjustment process for customer companies according to this embodiment. The sales forecast calculation process for customer companies in step S17 (see Figure 4) will be explained with reference to Figure 13. In step S71, the forecasting unit 114 starts the process of repeating steps S72 to S76 for each company that will be the customer.

[0049] Steps S72 to S75 are the same processes as steps S61 to S64 shown in Figure 12, respectively. However, the companies to be corrected are replaced with companies that are customers. In step S76, the sales forecast adjustment unit 115 calculates an adjusted sales forecast, which is the product of the impact calculated in step S75 and the sales forecast (see area 413 in Figure 5). This adjusted sales forecast is referred to as the provisional sales forecast.

[0050] In step S77, the sales forecast adjustment unit 115 calculates the sum of the adjusted provisional sales forecasts of each customer company calculated in step S76. Next, the sales forecast adjustment unit 115 calculates an adjustment rate (adjusted sales forecast of the companies to be adjusted / sum) that is equal to the adjusted sales forecast of the companies to be adjusted.

[0051] In step S78, the sales forecast adjustment unit 115 calculates a revised sales forecast for each customer company by multiplying the revised provisional sales forecast calculated in step S76 by the adjustment rate calculated in step S77. The sum of these revised sales forecasts is equal to the revised sales forecast of the companies subject to adjustment.

[0052] <<Adjustment processing for sales forecasts based on customer industry, size, and region>> Figure 14 is a flowchart of the sales forecast adjustment process for customer industries according to this embodiment. The sales forecast adjustment process for customer industries in step S17 (see Figure 4) will be explained with reference to Figure 14. The sales forecast adjustment process for customer size and customer region can be similarly applied by replacing customer industry with customer size and customer region, respectively.

[0053] In step S81, the prediction unit 114 starts the process of repeating steps S82 to S88 for each industry. Here, the industry refers to the demand-oriented industry in the area 423 shown in Figure 6. In step S82, the prediction unit 114 starts a process that repeats steps S83 to S87 for each company belonging to an industry. Steps S83 to S86 are the same as steps S61 to S64 shown in Figure 12. However, the companies to be corrected are replaced with companies belonging to an industry, and the model for companies to be corrected is replaced with the model for the demand industry.

[0054] In step S87, the sales forecast adjustment unit 115 calculates the adjusted sales forecast, which is the product of the impact calculated in step S86 and the sales forecast (see area 423 in Figure 6). This adjusted sales forecast is referred to as the provisional sales forecast for each company. In step S88, the sales forecast adjustment unit 115 calculates the sum of the provisional sales forecasts for each company calculated in step S87. This sum is referred to as the provisional sales forecast by industry.

[0055] In step S89, the sales forecast adjustment unit 115 calculates the sum of the provisional sales forecasts by industry calculated in step S88. Next, the sales forecast adjustment unit 115 calculates an adjustment rate (adjusted sales forecast of the company to be adjusted / sum) that makes the adjusted sales forecast of the company to be adjusted equal to the sum.

[0056] In step S90, the sales forecast adjustment unit 115 calculates the adjusted sales forecast for each industry by multiplying the sales forecast by the adjustment rate calculated in step S89. The sum of these adjusted sales forecasts for each industry is equal to the adjusted sales forecast of the companies subject to adjustment.

[0057] As explained in steps S77, S78, S89, and S90, the correction unit (sales forecast correction unit 115) of the sales forecast correction device 100 calculates a correction rate such that the sum of the sales forecasts of the company to be corrected for each customer, calculated based on the sales forecast of the company to be corrected for each customer and the degree of influence of the company to be corrected for each customer, is equal to the sales forecast of the company to be corrected, calculated based on the sum of the sales forecasts of the company to be corrected for each customer and the degree of influence of the company to be corrected for each customer, and uses the correction rate to correct the sales forecast of the company to be corrected for each customer.

[0058] ≪Extraction process for news with high impact≫ Figure 15 is a flowchart of the high-impact news extraction process according to this embodiment. The process of step S18 described in Figure 4 will be explained with reference to Figure 15. In step S101, the impact calculation unit 116 starts the process of repeating steps S102 to S107 for each company that is a customer.

[0059] In step S102, the prediction unit 114 starts a process that repeats steps S103 to S106 for each news article (hereinafter referred to as "the news article"). In step S103, the forecasting unit 114 starts the process of repeating step S104 for each news article except for the news article in question. The news articles in steps S102 and S103 are news articles acquired in the year in which the sales forecast is revised, or news articles from the past year.

[0060] Steps S104 to S106 are the same processes as steps S73 to S75 (see Figure 13). In step S107, the prediction unit 114 extracts a predetermined number of news articles, for example five, that correspond to the impact levels that differ significantly from the impact level calculated in step S106 when all news articles are included (see step S75).

[0061] As explained in step S107, the prediction unit 114 extracts a predetermined number of news items in order of their impact, starting with those that have the greatest impact on the impact score.

[0062] <<Display of Correction Results>> Figure 16 is a screen configuration diagram of the sales forecast correction result display screen 450 according to this embodiment. The sales forecast correction result display screen 450 is the initial screen for the correction result display process in step S19 (see Figure 4), and is the initial screen when sales forecasts by customer company are selected as the input type for sales forecasts (see area 412 shown in Figure 5). When sales forecasts by customer industry are selected as the input type (see Figure 6), area 454 displays a list of customer industries instead of a list of customer companies.

[0063] The sales forecast adjustment results display screen 450 has two display formats: summary and detail. As shown in area 451, the sales forecast adjustment results display screen 450 is a summary screen. In area 452, the dropdown list allows you to select "All," "Company P," "Company Q," etc., as the companies that will be the target customers for display. In Figure 16, "All" is selected, and all target companies are displayed in area 454.

[0064] The dropdown list in area 453 allows you to select the number of high-impact news articles to display. In Figure 16, "5 articles" is selected, and area 454 displays the impact of the top 5 news articles on sales forecasts for each customer company. These 5 articles correspond to the predetermined number in step S107 (see Figure 15). Area 454 displays the unadjusted sales forecast, the adjustment based on the most influential news article (adjusted sales forecast), and the adjusted sales forecast for each customer company. For example, for "Company P," the unadjusted sales forecast is 600 million yen, and the adjusted sales forecast is 800 million yen, indicating that the sales forecast increased by 100 million yen (impact amount of 100 million yen) due to the news article with the highest impact.

[0065] Figure 17 is a screen configuration diagram of the sales forecast correction result display screen 460 according to this embodiment. The display format of the sales forecast correction result display screen 460 is detailed (see area 461), and the sales forecast correction result for "Company P" (see area 462) is displayed in area 464. Area 464 shows the impact of the five news articles with the greatest impact (see area 463) on the sales forecast as a graph. When a news graph (rectangle, for example, "News AAAA") is selected, the time change display screen 468 (see Figure 18 below) showing the amount of impact the news has on the sales forecast is displayed.

[0066] Incidentally, in area 464, the unadjusted sales forecast was 600 million yen, but due to news article AAAA, it increased by 100 million yen; news article BBBB increased by 80 million yen; news article CCCC decreased by 70 million yen; news article DDDD increased by 50 million yen; news article EEEE increased by 20 million yen; and other news articles increased by 20 million yen, resulting in an adjusted sales forecast of 800 million yen.

[0067] Figure 18 is a diagram showing the screen configuration of the time-change display screen 468 of the impact amount according to this embodiment. The time-change display screen 468 displays news titles and body text, a graph showing the impact amount of the news on sales forecasts, etc. Figure 19 is a flowchart of the display process for the time-change display screen of the impact amount according to this embodiment.

[0068] In step S111, the prediction unit 114 starts the process of repeating step S112 for each news article. Step S112 is the same process as step S33 (see Figure 9). Note that the companies to be corrected should be replaced with the companies displayed on the sales forecast correction result display screen 460 (see Figure 18).

[0069] In step S113, the impact amount calculation unit 116 sets the elapsed days to 0. In step S114, the impact amount calculation unit 116 starts the process of repeating steps S115 to S117 while increasing the number of days elapsed (see Figure 2) of the news article (hereinafter referred to as "the news") specified on the impact amount time change display screen 468 from 0 to 1.

[0070] Step S115 is the same process as step S34 (see Figure 9). Step S116 is the same process as step S64 (see Figure 12). Note that the companies to be corrected should be replaced with the companies displayed on the sales forecast correction result display screen 460 (see Figure 18).

[0071] In step S117, the impact calculation unit 116 stops the loop and proceeds to step S118 if the news announcement date + elapsed days exceeds the end of the next fiscal year (step S117 → YES). If the news announcement date + elapsed days is before the end of the next fiscal year (step S117 → NO), the impact calculation unit 116 increments the elapsed days by one and returns to step S115.

[0072] In step S118, the impact amount calculation unit 116 adjusts the impact calculated in step S116 to match the fitting curve. For example, the impact amount calculation unit 116 takes t as the number of days elapsed and the increment of the impact is (t 2 / e t Correct it so that it becomes ). In step S119, the impact calculation unit 116 calculates the impact of the news. The impact is the product of the sales forecast for the company being displayed, the degree of impact, and the correction rate (see step S77 in Figure 13). The impact is calculated for each degree of impact (for each number of days elapsed) calculated in step S116.

[0073] In step S120, the screen control unit 117 displays a screen 468 (see Figure 18) showing the time change in the impact amount, including the news title, the main text, and a graph of the impact amount calculated in step S119. The horizontal axis of the impact amount graph represents the number of days elapsed since the announcement date. For example, 0 elapsed days is the announcement date, and (end of fiscal year - announcement date) elapsed days is the end of the fiscal year.

[0074] The process described in Figure 19 calculates the time-dependent change in the impact on the sales forecast of the target company to its customers. The same applies to the customer industry, customer size, and customer region. For example, for the customer industry, the sum of the impact amounts for each company in that industry can be used as the impact amount for that industry.

[0075] In step S117, the determination of when to terminate the iterative process is based on the end of the next fiscal year, but this is not limited to that. For example, the process may be repeated as long as the number of elapsed days is less than one year, or as long as the change in impact (the difference from the impact calculated in the previous step S116, the slope of the graph showing the impact amount) is greater than a predetermined value.

[0076] As explained in step S116, the forecasting unit 114 calculates the impact on the sales forecast of the adjusted company to its customers when the announcement date of news that has a significant impact on the impact changes. As explained in step S119, the impact amount calculation unit 116 calculates the change in the amount of impact of the news on the sales forecast of the company to be corrected to the demand destination, based on the degree of impact, the correction rate, and the sales forecast of the company to be corrected to the demand destination.

[0077] Features of the Sales Forecast Correction Device The sales forecast correction device 100 generates a predictive model using past news and its impact on the sales forecasts of demand customers as training data. The sales forecast correction device 100 corrects the sales forecast based on the input sales forecasts of demand customers and news. The sales forecast correction device 100 also identifies and displays news articles that have a significant impact on the sales forecast correction for each demand customer. Users will be able to adjust their sales forecasts in response to unforeseen events. Furthermore, they will be able to verify the accuracy of these adjustments by referring to news articles that have a significant impact on those adjustments.

[0078] <<Variation example: Calculating the sales forecast of the target company from the sales forecast adjustment of the demand destination>> In the embodiment described above, the sales forecast correction device 100 calculates the corrected sales forecast of the company to be corrected (see step S65 in Figure 12). Subsequently, the sales forecast correction device 100 corrects the sales forecast of the customer to match this corrected sales forecast (see step S78 in Figure 13 and step S90 in Figure 14).

[0079] Alternatively, the sum of the adjusted sales forecasts of the demand destinations may be used as the adjusted sales forecast for the company to be adjusted. This sum is the sum used when calculating the adjustment rate (see steps S77 and S89). In other words, the forecasting unit 114 calculates the degree of influence that the demand destinations have on the adjusted company's sales forecast using a forecasting model. The sales forecast adjustment unit 115 adjusts the adjusted company's sales forecasts for the demand destinations using the degree of influence, and the sum of these sales forecasts is used as the adjusted sales forecast for the company to be adjusted.

[0080] As described above, the sales forecast correction device includes a forecasting unit that calculates the degree of influence of the target company's sales forecast on demand using a demand forecasting model in which the degree of influence of the target company's sales forecast on demand is the dependent variable, and the aggregated results of the relationship between the target company's demands and news articles are included as explanatory variables. The sales forecast correction device also includes a correction unit (sales forecast correction unit 115) that corrects the target company's sales forecast on demand using the degree of influence, and takes the sum of the corrected sales forecasts as the corrected sales forecast for the target company.

[0081] ≪Variation: Number of days elapsed≫ In the embodiment described above, the number of elapsed days is the length of time from the news announcement to the end of the fiscal year of the individual company. Alternatively, the number of elapsed days may be defined as the time from the news announcement to the end of the fiscal year of the company subject to correction. Doing so reduces the workload of the process of generating the prediction model and the process of calculating the impact using the prediction model.

[0082] <<Other variations>> Although several embodiments of the present invention have been described above, these embodiments are merely illustrative and do not limit the technical scope of the present invention. The present invention can take various other embodiments, and furthermore, various modifications such as omissions and substitutions can be made without departing from the spirit of the present invention. These embodiments and their variations are included in the scope and spirit of the invention as described herein, and are also included in the scope of the invention and its equivalents as described in the claims. [Explanation of Symbols]

[0083] 100 Sales forecast correction device 111 Information Acquisition Department 112 News Tally Department 113 Learning Department 114 Prediction Section 115 Sales Forecast Adjustment Department (Adjustment Department) 116 Impact Calculation Section 117 Screen Control Unit 130 News Information Database 140 IR Information Database 150 Corporate Information Databases 160 Predictive Model Databases 320 Summary Results

Claims

1. The degree of impact on the sales forecast of a company subject to adjustment is calculated using a forecasting model for adjustment targets, which includes the aggregated results of the relationship between the adjustment target company and news articles as explanatory variables, and the degree of impact on the sales forecast as the dependent variable. A forecasting unit that calculates the impact on the sales forecast of the adjusted company to the demand destination using a demand destination forecasting model in which the aggregated results of the relationship between the demand destinations of the adjusted company and news articles are included as explanatory variables, and the impact on the sales forecast of the adjusted company to the demand destination is the dependent variable, A correction rate is calculated such that the sum of the sales forecasts of the adjusted company for the said customer, calculated based on the sales forecast of the adjusted company for the said customer and the degree of impact on the sales forecast of the adjusted company for the said customer, is equal to the sales forecast of the adjusted company, calculated based on the sum of the sales forecasts of the adjusted company for the said customer and the degree of impact on the sales forecast of the adjusted company. The system includes a correction unit that uses the correction rate to correct the sales forecast of the company subject to correction for the customer. Sales forecast adjustment device.

2. The aforementioned related attributes are, The relationship between the industry related to the content of the aforementioned news article and the aforementioned company subject to correction or the aforementioned customer, The scale of the content of the aforementioned news article, its relationship to the aforementioned company subject to correction or the aforementioned customer, and The relationship between the region related to the content of the aforementioned news article and at least one of the aforementioned companies subject to correction or the aforementioned customers. The sales forecast correction device according to claim 1.

3. The aforementioned aggregated results are as follows: Includes the average length of time from the news announcement to the date of sales performance calculation. The sales forecast correction device according to claim 1.

4. The aforementioned customer is, It is one of the following: a company, a company by industry, a company by size, or a company by region. The sales forecast correction device according to claim 1.

5. The prediction unit, A predetermined number of news articles are extracted in order of their impact, starting with those that have the greatest impact on the aforementioned level of influence. The sales forecast correction device according to claim 1.

6. It also includes an impact calculation unit, The aforementioned aggregated results are as follows: This includes the average length of time from the news announcement to the date of sales performance calculation. The prediction unit, The impact on the sales forecast of the adjusted company to the aforementioned customers is calculated when the announcement date of news that has a significant impact on the aforementioned impact changes. The aforementioned impact calculation unit, Based on the aforementioned impact level, the aforementioned correction rate, and the sales forecast of the company subject to the correction for the aforementioned demand destination, the change in the amount of impact of the news on the sales forecast of the company subject to the correction for the aforementioned demand destination is calculated. The sales forecast correction device according to claim 1.

7. A forecasting unit calculates the impact on the sales forecast of the adjusted company on the demand destinations, using a demand destination forecasting model in which the impact on the sales forecast of the adjusted company on the demand destinations is the dependent variable, and which includes the aggregated results of the relationship between the demand destinations of the adjusted company and news articles as explanatory variables, and the impact on the sales forecast of the adjusted company on the demand destinations is the dependent variable. The system includes a correction unit that uses the impact to correct the sales forecast of the company subject to correction for the aforementioned demand destination, and takes the sum of the corrected sales forecasts as the corrected sales forecast of the company subject to correction. Sales forecast adjustment device.

8. The sales forecast adjustment device, The steps include: calculating the impact on the sales forecast of a company subject to adjustment using a forecast model for adjustments that includes the aggregated results of the related attributes between the company subject to adjustment and news articles as explanatory variables, and the impact on the sales forecast as the dependent variable; The steps include: calculating the impact on the sales forecast of the adjusted company to the demand destination using a forecasting model for demand destinations, which includes the aggregated results of the relationship between the demand destinations of the adjusted company and news articles as explanatory variables, and the impact on the sales forecast of the adjusted company to the demand destinations as the dependent variable; A step of calculating an adjustment rate such that the sum of the sales forecasts of the adjusted company for the said customer, calculated based on the sales forecast of the adjusted company for the said customer and the degree of impact of the adjusted company's sales forecast on the said customer, is equal to the sum of the sales forecasts of the adjusted company for the said customer and the degree of impact of the adjusted company's sales forecast on the said customer. The steps include: performing the following: adjusting the sales forecast of the company subject to adjustment for the customer using the adjustment rate; Method for adjusting sales forecasts.

Images