Demand forecasting apparatus, demand forecasting method, and program

Abstract

To provide a demand forecasting apparatus, and the like, capable of forecasting demand in consideration of an intention of a user related to a target product.SOLUTION: A demand forecasting apparatus is configured to forecast a quantity of target products demanded, on the basis of a prediction model. The prediction model is constructed by integrating multiple kinds of individual learning models with weights applied thereto. The individual learning model is configured to output the quantity of demand in accordance with characteristics of a product, the individual learning model including: identifying means which identifies a similar product which is similar to a target product; calculating means which calculates a base demand for the target product, using the prediction model weighted in accordance with the similar product and the characteristics of the target product; and operation receiving means which receives operation to modify the weight for each of the individual learning models. The calculating means calculates a forecasted demand for the target product, using the prediction model with the operation reflected therein. This disclosure can be used for assisting decision making.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present disclosure relates to a demand forecasting device, a demand forecasting method, and a program. [Background technology]

[0002] Techniques exist for forecasting demand for products using predictive models.

[0003] Patent Document 1 discloses a technique for aggregating actual demand data for products similar to a product to be predicted, and using the actual demand data and a prediction model to predict demand based on the release date of the product to be predicted. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2017-027632 A Summary of the Invention [Problem to be solved by the invention]

[0005] By the way, various goals may be set when selling a product. The person in charge of the product makes decisions to achieve those goals. For example, various actions such as plans for each process in the supply chain and promotion policies are considered. At this time, the results of the demand forecast described above may be used.

[0006] For example, the demand for a target product is predicted using information about similar products that are similar to the target product. Here, the similar products are different from the target product. Therefore, there may be matters that need to be considered for the target product. Or, a person in charge (user) may want to emphasize certain matters for the target product.

[0007] When considering an action, a user has a need to use the result of a demand forecast that reflects such matters. That is, there are cases where a user desires the result of a demand forecast that reflects the user's intention.

[0008] The present disclosure has been made in consideration of the above-mentioned problems, and one of its objectives is to provide a demand prediction device etc. capable of making demand predictions taking into account user intentions regarding a target product. [Means for solving the problem]

[0009] A demand forecasting device according to one aspect of the present disclosure is a device that predicts demand for a target product based on a forecasting model, wherein the forecasting model is a model in which multiple types of individual learning models are integrated and weighted for each individual learning model, and the individual learning models are models that output demand according to the characteristics of the product, and the device is equipped with an identification means for identifying similar products that are similar to the target product, a calculation means for calculating a base demand for the target product using the forecasting model weighted according to the similar products and the characteristics of the target product, and an operation receiving means for receiving an operation to change the weights for each of the individual learning models, and the calculation means calculates the predicted demand for the target product using the forecasting model in which the change operation is reflected.

[0010] A demand forecasting method according to one aspect of the present disclosure is a method for predicting demand for a target product based on a forecasting model, wherein the forecasting model is a model in which multiple types of individual learning models are integrated and weighted for each individual learning model, and the individual learning models are models that output demand based on the characteristics of the product, the method identifies similar products that are similar to the target product, calculates a base demand for the target product using the forecasting model weighted according to the similar products and the characteristics of the target product, accepts a weighting change operation for each of the individual learning models, and calculates a predicted demand for the target product using the forecasting model in which the change operation is reflected.

[0011] A program according to one aspect of the present disclosure is a program that predicts demand for a target product based on a prediction model, wherein the prediction model is a model in which multiple types of individual learning models are integrated and weighted for each individual learning model, and the individual learning models are models that output demand according to the characteristics of the product, and causes a computer to execute a process of identifying similar products that are similar to the target product, a process of calculating a base demand for the target product using the prediction model weighted according to the similar products and the characteristics of the target product, and a process of accepting a weighting change operation for each of the individual learning models, and in the calculation process, the predicted demand for the target product is calculated using the prediction model in which the change operation is reflected. Effect of the Invention

[0012] According to the present disclosure, it is possible to perform demand forecasting taking into account the intentions of users regarding target products. [Brief description of the drawings]

[0013] [Figure 1] FIG. 1 is a diagram illustrating an example of a configuration including a demand prediction device according to the present disclosure. [Diagram 2] 1 is a first block diagram illustrating an example of a functional configuration of a demand prediction device according to the present disclosure. [Diagram 3] 1 is a first flowchart illustrating an example of an operation of a demand prediction device according to the present disclosure. [Figure 4] FIG. 2 is a second block diagram showing an example of a functional configuration of a demand prediction device according to the present disclosure. [Diagram 5] 13 is an example of an input screen according to the present disclosure. [Figure 6] FIG. 13 is a diagram showing a first example of output information according to the present disclosure. [Figure 7] FIG. 13 is a diagram illustrating an example of a parameter input screen according to the present disclosure. [Figure 8] FIG. 1 illustrates an example of a relationship graph according to the present disclosure. [Figure 9]FIG. 13 is a diagram showing a second example of output information according to the present disclosure. [Figure 10] 11 is a second flowchart illustrating an example of the operation of the demand prediction device according to the present disclosure. [Figure 11] FIG. 13 is a diagram showing a third example of output information according to the present disclosure. [Figure 12] FIG. 2 is a block diagram showing an example of a hardware configuration of a computer device that realizes a demand prediction device according to the present disclosure. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings, in which each of the drawings is associated with one or more embodiments.

[0015] <First embodiment> An overview of a demand prediction device according to a first embodiment will be described.

[0016] 1 is a diagram illustrating an example of a configuration including a demand prediction device 100. The demand prediction device 100 is communicably connected to, for example, a terminal device 200 via a wireless or wired network.

[0017] The demand prediction device 100 is a device that predicts the demand for a product. For example, the demand prediction device 100 predicts the demand for a target product by using a prediction model that outputs the demand. The demand is, for example, the sales quantity or sales amount of the product, but is not limited to this example. The demand prediction device 100 is, for example, a server device. The demand prediction device 100 may be realized as a cloud server. Furthermore, the demand prediction device 100 may be realized as a device that constitutes a system. For example, a demand prediction system including the demand prediction device 100 and a terminal device 200 may be realized.

[0018] The predicted demand amount is output to, for example, the terminal device 200. The terminal device 200 is a computer device having an input / output function. The terminal device 200 is, for example, a personal computer. The terminal device 200 is used by a user. An example of a user is a person in charge of a product, such as a demand planner. The user is not limited to this example. The user can check the predicted demand amount. Then, the demand prediction device 100 accepts an operation related to the prediction of the demand amount from the user, for example, via the terminal device 200. Then, the demand prediction device 100 predicts a further demand amount reflecting the operation, and outputs the predicted demand amount to the terminal device 200. Note that, without being limited to this example, the demand prediction device 100 may be realized in various forms.

[0019] Next, an example of the functional configuration of the demand prediction device 100 will be described.

[0020] 2 is a block diagram showing an example of a functional configuration of the demand prediction device 100. The demand prediction device 100 includes an identifying unit 110, a calculating unit 120, and an operation receiving unit .

[0021] The identification unit 110 identifies a similar product that is similar to the target product. For example, the identification unit 110 identifies a similar product by using characteristic information of the product. The characteristic information is information indicating characteristics of the product. The characteristics may include elements of the product itself, such as the category, characteristics, and sales price. The category indicates the division to which the product belongs among the product group. The characteristics indicate the properties of the product. For example, assume that the product is a polo shirt. In this case, the category is tops. The characteristics are color, material, function, and the like. Note that the product is merely an example. The product is not limited to clothing. Furthermore, the characteristics are not limited to this example, and further information may be included in the characteristics. The identification unit 110 may identify a similar product by comparing the characteristic information of the target product with each of the characteristic information of a plurality of products. The method of identifying similar products is not limited to this example.

[0022] In this manner, the identifying unit 110 identifies a similar product that is similar to the target product. The identifying unit 110 is an example of an identifying means.

[0023] The calculation unit 120 calculates the demand for the target product using the prediction model. The prediction model outputs the demand for the product. For example, the prediction model is a model in which multiple types of individual learning models are integrated. The individual learning model outputs the demand according to the characteristics of the product. The individual learning model may be a trained model in which the relationship between the product characteristics and the demand is learned by machine learning. For example, the individual learning model outputs the demand using the product characteristics as input. At this time, a weight is set for each individual learning model. For example, the prediction model may be a model that aggregates values ​​obtained by multiplying the demand output for each individual learning model by the weight, and outputs the value obtained by the aggregation as the demand for the target product. In this way, the prediction model may be a model in which multiple types of individual learning models are integrated and weighted for each individual learning model. Note that the prediction model is not limited to this example.

[0024] The weights set in the individual learning models are determined for each product. For example, the weights set in the individual learning models may be changed for each product. The calculation unit 120 calculates the demand for the target product using a prediction model weighted according to similar products. For example, the calculation unit 120 sets the weights for each individual learning model to weights corresponding to similar products. Next, the calculation unit 120 inputs the features of the target product to the prediction model in which the weights corresponding to similar products are set. The input features are input to each of the individual learning models. Then, the calculation unit 120 may calculate the demand output from the prediction model as the demand for the target product. Hereinafter, the demand calculated using the prediction model is referred to as a base demand.

[0025] In this manner, the calculation unit 120 calculates the base demand for the target product by using the prediction model weighted according to the similar products and the characteristics of the target product. The calculation unit 120 is an example of a calculation means.

[0026] The operation receiving unit 130 receives various operations from the user. Specifically, the operation receiving unit 130 receives operations on parameters for the prediction model. The parameters are information that influences the output of the demand amount by the prediction model. One example of a parameter is a weight for each individual learning model that is used when outputting the demand amount by the prediction model. For example, the operation receiving unit 130 receives an operation to change the weight for each individual learning model.

[0027] In this manner, the operation accepting unit 130 accepts an operation to change the weight for each of the individual learning models. The operation accepting unit 130 is an example of an operation accepting means.

[0028] When the operation of the parameters is accepted by the operation accepting unit 130, the calculation unit 120 calculates the demand amount of the target product in accordance with the operation. Specifically, the calculation unit 120 calculates the demand amount for the target product using the prediction model in which the operation is reflected. Here, the demand amount calculated using the prediction model in which the operation is reflected is referred to as the predicted demand amount.

[0029] Next, an example of the operation of the demand prediction device 100 will be described with reference to Fig. 3. In this disclosure, each step in a flowchart will be expressed using a number assigned to each step, such as "S1".

[0030] 3 is a flowchart explaining an example of the operation of the demand forecasting device. The identification unit 110 identifies similar products similar to the target product (S1). The calculation unit 120 calculates the base demand for the target product using a prediction model weighted according to the similar products and the characteristics of the target product (S2). The operation reception unit 130 receives an operation to change the weight for each individual learning model (S3). Then, the calculation unit 120 calculates the predicted demand for the target product using the prediction model reflecting the operation (S4).

[0031] In this way, the demand prediction device 100 of the first embodiment is a device that predicts the demand for a target product based on a prediction model. In this case, the prediction model is a model in which a plurality of types of individual learning models are integrated and weighted for each of the individual learning models. The individual learning model is a model that outputs the demand amount according to the characteristics of the product. The demand prediction device 100 identifies similar products that are similar to the target product. In addition, the demand prediction device 100 calculates the base demand amount for the target product using the prediction model weighted according to the similar products and the characteristics of the target product. Furthermore, the demand prediction device 100 accepts an operation to change the weight for each individual learning model. Then, the demand prediction device 100 calculates the predicted demand amount for the target product using the prediction model in which the operation is reflected.

[0032] There may be cases where a target product has items that are important or not important in comparison with similar products. The demand prediction device 100 can accept consideration of such items as, for example, an operation to change the weight of an individual learning model. This allows the demand prediction device 100 to calculate the demand amount of the target product when such items are taken into consideration. In other words, the demand prediction device 100 can perform demand prediction that takes into consideration the intentions of users regarding the target product.

[0033] Furthermore, the demand prediction device 100 can present a base demand amount calculated before an operation on the prediction model and a predicted demand amount calculated after the operation. Therefore, the user can compare the base demand amount with the predicted demand amount to consider how much an operation affects the demand amount. In other words, the demand prediction device 100 can also support decision-making regarding actions for selling a product. Second embodiment Next, a demand prediction device according to a second embodiment will be described. In the second embodiment, a further example of the demand prediction device 100 described in the first embodiment will be described. Note that some of the contents overlapping with the first embodiment will not be described. Also, in this embodiment, an example in which the product is clothing will be mainly described, but the product is not limited to this example.

[0034] [Details of the demand forecasting device 100] FIG. 4 is a block diagram showing an example of a functional configuration of the demand prediction device 100. As shown in FIG. 4, the demand prediction device 100 includes an identifying unit 110, a calculating unit 120, an operation receiving unit 130, an acquiring unit 140, and a display control unit 150. The demand prediction device 100 is communicably connected to a terminal device 200 as in the first embodiment. The terminal device 200 is, for example, a personal computer, but is not limited to this example. The terminal device 200 may be a mobile terminal such as a smartphone or a tablet terminal. The terminal device 200 has an input / output function. For example, the terminal device 200 may be equipped with input devices such as a keyboard, a mouse, a touch panel, a camera, a barcode reader, and an IC (Integrated Circuit) reader. The terminal device 200 may also be equipped with output devices such as a display, a printer, a speaker, and a lamp.

[0035] The demand prediction device 100 includes a storage unit 190. The storage unit 190 may be included in an external device capable of communicating with the demand prediction device 100. The storage unit 190 stores various information. For example, the storage unit 190 stores a database. The database includes characteristic information for each product. For example, the user inputs the characteristic information of the product in advance at the terminal device 200. The input characteristic information is then registered in the database. At this time, identification information for identifying the product and the characteristic information of the product are registered in association with each other in the database. Furthermore, the storage unit 190 stores performance information. The performance information is information indicating the sales performance of the product. For example, the performance information indicates information regarding the sales amount and sales quantity of the product sold in the past.

[0036] The acquisition unit 140 acquires feature information indicating features related to the product. The features are, for example, elements of the product itself such as category, characteristics, and sales price, but are not limited to this example. For example, assume that the product is a polo shirt. In this case, the category is tops. The characteristics are color, material, function, etc. If the product is cargo pants, the category is bottoms, and the characteristics are color, material, function, etc. The features may be defined similarly even if the product is various types of pants, jackets, coats, underwear, shoes, etc. The features may include information on measures implemented for the product and information on the external environment at the time of product sales. The information on the measures includes the content of the promotion carried out for sales promotion, the investment scale of the promotion, etc. The content of the promotion includes advertising media such as television commercials, web advertisements, magazines, and flyers, and the location where the promotion is carried out. The information on the external environment is the weather conditions, the occurrence of disasters, the occurrence of infectious diseases, and exchange rates at the time of product sales. The features may include further information.

[0037] The acquiring unit 140 may acquire information input by a user in the terminal device 200 as the characteristic information of the target product. In this case, the user inputs the characteristic information of the target product to the terminal device 200. Then, the acquiring unit 140 may acquire the input information from the terminal device 200.

[0038] Alternatively, the acquisition unit 140 may acquire the characteristic information of the target product from another device having the characteristic information of the product. For example, the acquisition unit 140 may acquire the characteristic information of the target product from a database stored in the storage unit 190.

[0039] In this embodiment, the acquisition unit 140 acquires the feature information from a database. Specifically, the acquisition unit 140 acquires the identification information of the target product from the terminal device 200. The acquisition unit 140 acquires the feature information of the target product from the database based on the acquired identification information. At this time, the identification information may be input in the terminal device 200 by a user operation.

[0040] In this manner, the acquiring unit 140 acquires characteristic information indicating characteristics of the target product. The acquiring unit 140 is an example of an acquiring means.

[0041] The identification unit 110 identifies similar products that are similar to the target product. The identification unit 110 may use characteristic information of the target product. Specifically, the identification unit 110 identifies one or more similar products from a database that has characteristic information of multiple products. At this time, the identification unit 110 may identify the similar products by calculating the similarity between the target product and products shown in the database using the characteristic information.

[0042] For example, the identification unit 110 compares the feature information of the target product with the feature information of each product shown in the database. At this time, it is not necessary to compare all features. For example, the identification unit 110 may compare only the elements of the products themselves. The identification unit 110 calculates a similarity value that increases as the number of matching features increases. Then, the identification unit 110 may identify a product whose similarity exceeds a threshold as a similar product.

[0043] Alternatively, the identification unit 110 may calculate the distance between products by converting the features into coordinates. More specifically, the identification unit 110 calculates the distance between the coordinates based on the features of the target product and the coordinates based on the features of each product shown in the database. The identification unit 110 calculates a similarity that indicates a value that increases as the calculated distance decreases. Then, the identification unit 110 may identify products whose similarity exceeds a threshold as similar products.

[0044] Alternatively, the identification unit 110 may identify similar products using a similarity calculation model generated by machine learning. The similarity calculation model may be a model generated by supervised learning. In this case, the similarity calculation model is generated by, for example, learning feature information of a plurality of products as learning data and information indicating whether the products are similar to each other as a correct answer. Note that the method of generating the similarity calculation model is not limited to this example. For example, the identification unit 110 inputs feature information of the target product into the similarity calculation model generated in this way. The similarity calculation model calculates the similarity between the target product and the product shown in the database. Then, the identification unit 110 may identify a product whose calculated similarity exceeds a threshold as a similar product.

[0045] The display control unit 150 displays various information on the display of the terminal device 200. Specifically, the display control unit 150 displays a list of similar products identified by the identification unit 110. At this time, the display control unit 150 may display information in which the identified similar products are arranged in order of degree of similarity. In other words, the display control unit 150 may display a ranking of the similar products according to their degree of similarity.

[0046] In this manner, the display control unit 150 displays a list of one or more identified similar products. The display control unit 150 is an example of a display control means.

[0047] Furthermore, the display control unit 150 displays the demand amount calculated by the calculation unit 120. Examples of display of various information by the display control unit 150 will be described later.

[0048] The calculation unit 120 uses a prediction model to calculate the demand for the target product. The prediction model is a model that outputs the demand for the product. The demand may be the sales volume or the sales amount. The prediction model may be a model that outputs the demand for the product in a predetermined period. For example, the calculation unit 120 calculates the total demand in the predetermined period. Furthermore, the calculation unit 120 may calculate the demand for each unit period in the predetermined period. That is, the fluctuation in the demand in the predetermined period may be output. An example of the predetermined period is a period from the start of sales to a predetermined number of days after the product is released. The predetermined period is not limited to this example, and may be a period from any date after the start of sales to a predetermined number of days after the product is released. A specific example of the prediction model will be described later.

[0049] The operation receiving unit 130 receives various operations from the user. Specifically, the operation receiving unit 130 receives operations on parameters that affect the output of the demand amount for the prediction model. The demand amount output by the prediction model changes as the parameters are operated. Specific examples of parameter operations will be described later together with specific examples of the prediction model.

[0050] The calculation unit 120 recalculates the demand for the target product using the prediction model whose parameters have been manipulated by the user. The demand for the target product calculated using the prediction model whose parameters have been manipulated by the user in this manner is referred to as the predicted demand. In other words, the predicted demand can be said to be the demand calculated using a prediction method that reflects the user's intention. On the other hand, the base demand is the demand for the target product calculated using a prediction model whose parameters have not been manipulated by the user.

[0051] [Examples of prediction models] Next, a specific example of a prediction model will be described.

[0052] The prediction model is a model that outputs a demand amount based on similar products similar to the target product. The similar products are identified by the identification unit 110. Then, the calculation unit 120 may calculate a base demand amount based on the identified similar products.

[0053] An example of a model that outputs demand based on similar products is a trained model weighted according to similar products. The trained model is a model in which multiple types of individual learning models are integrated. The trained model is referred to as an integrated learning model. The integrated learning model is a model weighted for each individual learning model. The integrated learning model takes the product features as input and outputs the demand. Here, the individual learning model is a trained model in which the relationship between the product features and the demand is learned by machine learning. For example, when the product features are input to the integrated learning model, the features are input to each of the individual learning models. At this time, the integrated learning model aggregates values ​​obtained by multiplying the demand output for each individual learning model by the weight, and outputs the value obtained by the aggregation as the demand for the target product.

[0054] The individual learning model may be a regression equation generated by regression analysis with the demand as the objective variable and the explanatory variables as features. For example, the individual learning model may be generated by sparse linear regression such as LASSO (Least Absolute Shrinkage and Selection Operator) regression. Note that the machine learning method for generating the individual learning model is not limited to this example.

[0055] Each of the individual learning models is a different type of trained model. Specifically, each of the individual learning models may be a model having different explanatory variables or coefficients of explanatory variables. In this case, each of the individual learning models may be a model corresponding to factors that affect the demand amount. Examples of factors that affect the demand amount are the appearance of the product, the contents of the product, the external environment when the product is sold, and measures implemented for the product. Each of the factors has a related product feature. For example, if the product is clothing, the "appearance" is related to features such as the category, material, color, and the presence or absence of a logo. In addition, the "contents" is related to features such as the selling price, production location, function, and brand. The "external environment" is related to, for example, the weather conditions, the occurrence of disasters, the occurrence of infectious diseases, and exchange rates at the time of the product sale. The "measures" are related to features such as the content of promotions carried out for sales promotion and the investment scale of the promotion. In other words, the individual learning model can be said to be a model that outputs the demand amount by inputting features related to the corresponding factor. Note that the same feature may correspond to multiple factors. For example, characteristics such as materials and functions may correspond to "appearance" and "contents", respectively. Although "appearance", "contents", "external environment", and "measures" are given as factors, the factors are not limited to these examples. The factors may differ depending on various products and various industries to which the demand prediction device 100 is applied.

[0056] The individual learning model may be a demand prediction model that emphasizes a specific factor. For example, an individual learning model that emphasizes appearance may be a demand output model that uses features corresponding to appearance, such as category, material, color, and the presence or absence of a logo. Alternatively, an individual learning model that emphasizes appearance may be a demand output model in which features corresponding to appearance are weighted more heavily than other features. Each of the individual learning models may be a model that inputs a set of features that are different from each other. The integrated learning model is, for example, a model that integrates multiple such individual learning models, and is a demand prediction model weighted for each individual learning model. One individual learning model is X n(n is a natural number greater than or equal to 1), the integrated learning model can be expressed, for example, as in Equation 1 below.

[0057] f(X1,X2,...,X n )=a1X1+a2X2+···+a n X n (Formula 1) a n is a coefficient indicating the weight. n Regarding , for example, the following formula 2 holds true.

[0058] a1+a2+···+a n =1 (formula 2) In this way, the integrated learning model may be a model that calculates a value obtained by multiplying the demand amount output for each individual learning model by a weight set, and outputs the demand amount by adding the calculated value. From formula 2, it can be said that the weight is the ratio of the individual learning model in the integrated learning model. The larger the weight, the greater the influence of the individual learning model corresponding to the weight on the demand amount output by the integrated learning model. In other words, the larger the weight, the more importance is attached to the factor corresponding to the individual learning model in calculating the demand amount. In other words, it can be said that the weight for each individual learning model indicates the degree of influence of the factor on the demand amount output by the integrated learning model. It should be noted that formulas 1 and 2 are merely examples. The integrated learning model is not limited to this example.

[0059] The weights are set in advance. The weights may be set for each product. For example, the user may input a weight according to the product in advance. Alternatively, the weights may be set according to the sales record of the product. Specifically, the weights may be set so that the total demand amount for a predetermined period output by the integrated learning model and the total demand amount for a predetermined period based on the sales record of the product are the same value. The weights may be set by various methods without being limited to this example.

[0060] The calculation unit 120 inputs the acquired feature information of the target product into the integrated learning model. The features indicated in the feature information are input into the individual learning model. At this time, the weight for each individual learning model is set to a value corresponding to one similar product. Then, the calculation unit 120 calculates the demand amount output by the integrated learning model as the base demand amount of the target product for a predetermined period of time.

[0061] The operation receiving unit 130 may receive an operation to change the weights of the individual learning models for the integrated learning model. At this time, the calculation unit 120 calculates the predicted demand amount of the target product in a predetermined period by inputting the characteristic information of the target product to the integrated learning model in which the weights of the individual learning models have been changed.

[0062] Furthermore, the operation receiving unit 130 may receive an operation to change the individual learning model itself. Specifically, the operation receiving unit 130 may receive an operation to change a formula indicating the individual learning model. For example, the individual learning model is a regression formula generated by regression analysis. In this case, the operation receiving unit 130 may receive an operation to change the coefficients and constant terms in the regression formula. When the individual learning model itself is changed, the possible range of values ​​output by the individual learning model is changed.

[0063] Furthermore, when the individual learning model is generated by regression analysis, there is a data group showing the sales performance for each set of product features. Then, by learning about the data group, the individual learning model is generated as a regression equation showing the relationship between the features and the demand. The operation reception unit 130 may receive an operation of the data group to be learned at this time. For example, a user may perform an operation to remove specific data learned in the generation of a specific individual learning model from the learning target, or to change various values ​​indicated in the specific data. The operation reception unit 130 may receive a selection of an individual learning model from the user, and may receive a change operation for the data group learned in the generation of the selected individual learning model. In this way, the operation reception unit 130 may receive a change operation for the data group used in the machine learning performed when generating the individual learning model. The calculation unit 120 calculates the predicted demand of the target product in a predetermined period by inputting the feature information of the target product into the integrated learning model in which such a change operation is reflected. The parameters to be changed are not limited to this example.

[0064] [Examples of various information displayed] Next, a description will be given of various display examples displayed by the display control unit 150. In these examples, the display control unit 150 causes the display of the terminal device 200 to display various types of information.

[0065] When the acquisition unit 140 acquires the characteristic information of the target product, the display control unit 150 displays an input screen that prompts the user to input information about the target product. FIG. 5 is an example of the input screen. For example, the user selects "long-sleeved polo shirt" under "tops." The acquisition unit 140 acquires identification information corresponding to the selected "long-sleeved polo shirt." The acquisition unit 140 may then acquire the characteristic information of the "long-sleeved polo shirt" from a database based on the identification information corresponding to the "long-sleeved polo shirt."

[0066] In the example of FIG. 5, when "Product Registration" is selected, the display control unit 150 may display an input form for product characteristics on the display of the terminal device 200. For example, the user inputs the product characteristics into the input form. The acquisition unit 140 acquires information indicating the input product characteristics. Then, the acquisition unit 140 may register the acquired information as characteristic information in a database in association with the product identification information.

[0067] Furthermore, the display control unit 150 displays output information including the calculated demand amount. FIG. 6 is a diagram showing a first example of output information. In the example of FIG. 6, output information is displayed when the target product is a "long-sleeved polo shirt." The output information shows a ranking of similar products. The similar product ranking is a list of similar products identified by the identification unit 110. Information indicating similar products is arranged in descending order of similarity. The user selects one of the similar products shown in the similar product ranking. The operation reception unit 130 receives the selection. In the example of FIG. 6, "short-sleeved polo shirt" is selected.

[0068] For example, when the operation receiving unit 130 receives a selection of a similar product, the calculation unit 120 uses a prediction model to calculate a base demand amount of the target product. In the example of FIG. 6, a base demand amount calculated for a target period of three months from April 1, 2023 is shown. The prediction model is an integrated learning model. In this example, the integrated learning model outputs a demand amount based on the selected similar product "short-sleeved polo shirt."

[0069] Furthermore, in this example, the total demand for the target period and a line graph showing the fluctuation of the demand for the target period are shown. The total demand is the sales volume. For example, the total demand for the target period calculated by the integrated learning model is 5,300 points. The fluctuation of the demand for the target period calculated by the integrated learning model is also shown by a line graph.

[0070] In addition, the user operates parameters for the prediction model. For example, the user selects the "input" icon shown in the block of the integrated learning model. When the icon is selected, the operation reception unit 130 displays a parameter input screen for the integrated learning model.

[0071] FIG. 7 is a diagram showing an example of a parameter input screen. The parameter input screen in FIG. 7 corresponds to an integrated learning model. In this example, the integrated learning model is composed of four individual learning models, namely, an appearance-emphasis model, a content-emphasis model, an external environment-emphasis model, and a policy-emphasis model. The pre-input ratio indicates a weight set in advance. That is, the pre-input ratio is a weight corresponding to a similar product selected by a user. The importance ratio is a weight input by a user. That is, the input of the importance ratio is a weight change operation for an individual learning model. For example, in the example of FIG. 7, the weight of the appearance-emphasis model corresponding to a similar product is 20%. And, the weight of the appearance-emphasis model is changed to 30% by a change operation by a user. That is, the calculation unit 120 calculates a predicted demand amount using an integrated learning model in which the weight of the appearance-emphasis model is set to 30%.

[0072] Moreover, the box plot in the example of FIG. 7 shows the distribution of a data group showing the sales performance for each set of product features used in generating the individual learning model. For example, the user performs an operation to change the numerical values, etc., for the box plot. This changes the individual learning model itself. For example, if the individual learning model is expressed by a regression equation represented by a linear function, the change operation causes the linear function to move in parallel or changes the slope of the linear function. In other words, the range of values ​​that the individual learning model can take is changed.

[0073] The operation receiving unit 130 can receive such a change operation. The parameter input screen for receiving the change operation is not limited to this example. For example, the parameter input screen may display sliders for operating various parameters, or options for various parameters may be displayed.

[0074] By performing such various change operations, the demand prediction device 100 can calculate the predicted demand amount that takes into account various conditions related to the target product as thought by the user. In other words, the demand prediction device 100 can present the predicted demand amount that reflects the user's intention.

[0075] Furthermore, the display control unit 150 may display a graph showing the relationship between the demand amount and the individual learning model. A graph showing the relationship between the demand amount and the individual learning model is called a relationship graph. FIG. 8 is a diagram showing an example of a relationship graph. The relationship graph graphically shows how the demand amount output by the integrated learning model affects the individual learning model. In the example of FIG. 8, the icons of "appearance emphasis", "content emphasis", "external environment emphasis", and "measure emphasis" are connected with arrows to the icon of "demand amount". The icons directly connected to "demand amount" respectively indicate individual learning models. In addition, the thickness of the arrow indicates the weight. In this way, the icon corresponding to the individual learning model may be highlighted according to the weight. In addition, an icon indicating a feature related to the individual learning model is further connected to the icon indicating the individual learning model.

[0076] The display control unit 150 may display the relationship graph together with the parameter input screen of the integrated learning model. Therefore, the user can operate the parameters while checking the relationship graph. Also, for example, the thickness of the arrow changes according to the weight changed by the user. Therefore, the user can intuitively understand how the change operation is affected.

[0077] Then, the display control unit 150 displays the demand calculated by the prediction model in which the change operation is reflected. FIG. 9 is a diagram showing a second example of the output information. Specifically, FIG. 9 shows the output information when the parameter operation is performed on the integrated learning model. In this example, the line graph shows the base demand amount, which is the demand amount before the parameter operation (before the intention is input), and the predicted demand amount, which is the demand amount after the parameter operation (after the intention is input). In addition, the change in the total demand amount before and after the parameter operation is performed is also shown. In this way, the display control unit 150 may display both the base demand amount and the predicted demand amount. This allows the user to confirm the change in the demand amount before and after the parameter operation.

[0078] In this way, the demand prediction device 100 can present to the user what products exist as similar products to the target product, and how the demand amount was calculated based on which similar product. In other words, the demand prediction device 100 can visualize the demand prediction process for the target product. Therefore, compared with the case where only the result of the demand prediction is presented to the user, the demand prediction device 100 can present information that is convincing to the user. In addition, the demand prediction device 100 can accept user operations in each process in the process, such as selection of similar products and operation of parameters for the prediction model, while visualizing the demand prediction process. Therefore, for example, the user can perform input operations while understanding which of the information to be input is used in the demand prediction process.

[0079] [Operation example of the demand forecasting device 100] Next, an example of the operation of the demand prediction device 100 will be described with reference to FIG.

[0080] Fig. 10 is a flowchart illustrating an example of the operation of the demand prediction device 100. The acquisition unit 140 acquires feature information of a target product (S101). For example, the display control unit 150 displays an input screen as shown in Fig. 5. The user selects a target product. Then, the acquisition unit 140 acquires the feature information of the target product from a database stored in the storage unit 190 based on the identification information corresponding to the selected target product.

[0081] The identification unit 110 identifies similar products similar to the target product by using the characteristic information of the target product (S102). The display control unit 150 displays output information including a ranking of similar products (S103). The operation reception unit 130 receives a selection of similar products from the user (S104).

[0082] The calculation unit 120 calculates the base demand amount (S105). For example, the calculation unit 120 calculates the base demand amount using an integrated learning model in which weights corresponding to the selected similar products are set. The display control unit 150 displays output information indicating the calculated base demand amount (S106). The display control unit 150 displays output information such as that shown in FIG. 6, for example.

[0083] The operation accepting unit 130 displays a parameter input screen corresponding to the prediction model (S107). In the example of FIG. 6, the user selects an input icon, causing the operation accepting unit 130 to display the parameter input screen. The operation accepting unit 130 accepts parameter operation (S108). The calculation unit 120 calculates the predicted demand amount by using the prediction model reflecting the accepted operation (S109). Then, the display control unit 150 displays output information indicating the predicted demand amount (S110).

[0084] Note that this operation example is merely an example, and the operation of the demand prediction device 100 is not limited to the above example.

[0085] In this way, the demand prediction device 100 of the second embodiment is a device that predicts the demand for a target product based on a prediction model. In this case, the prediction model is a model in which a plurality of types of individual learning models are integrated and weighted for each of the individual learning models. The individual learning model is a model that outputs the demand amount according to the characteristics of the product. The demand prediction device 100 identifies similar products that are similar to the target product. In addition, the demand prediction device 100 calculates the base demand amount for the target product using the prediction model weighted according to the similar products and the characteristics of the target product. Furthermore, the demand prediction device 100 accepts an operation to change the weight for each individual learning model. Then, the demand prediction device 100 calculates the predicted demand amount for the target product using the prediction model in which the operation is reflected.

[0086] Here, each of the individual learning models corresponds to a different factor that affects the demand amount. Also, the individual learning models are models that input features related to the corresponding factors and output the demand amount. The prediction model outputs the demand amount by aggregating values ​​obtained by multiplying the demand amount output for each individual learning model by a weight.

[0087] There may be cases where a target product has some features that are important or not important compared to similar products. For example, a user may want to emphasize the appearance of a product, which is one of the factors that affect the demand amount, more than similar products. The demand prediction device 100 can accept the consideration of such features as, for example, an operation to change the weight of an individual learning model. This allows the demand prediction device 100 to calculate the demand amount of the target product when such features are taken into consideration. In other words, the demand prediction device 100 can perform demand prediction that takes into consideration the user's intention regarding the target product.

[0088] Furthermore, the demand prediction device 100 may accept a change operation for a formula indicating an individual learning model. At this time, the individual learning model is a model in which the relationship between the product features and the demand amount is learned by machine learning, and the demand prediction device 100 accepts a change operation for a data group used in the machine learning.

[0089] This allows the demand prediction device 100 to accept changes to not only the weights for each individual learning model but also the individual learning itself, allowing the demand prediction device 100 to perform demand prediction that reflects more detailed user intent.

[0090] Furthermore, the demand prediction device 100 may display a list of one or more identified similar products. Then, the demand prediction device 100 may accept a selection of one similar product, and calculate a base demand amount for the target product by using a prediction model weighted according to the selected one similar product and the characteristics of the target product. In this way, the demand prediction device 100 can present the demand amount calculated in consideration of the similar products desired by the user.

[0091] Furthermore, the demand prediction device 100 may display a graph showing the relationship between the demand amount and the individual learning model, in which an icon corresponding to the individual learning model is highlighted according to the weight. At this time, the demand prediction device 100 may display both a parameter input screen and the graph. Then, when the demand prediction device 100 receives an operation to change the weight for each individual learning model, it updates the graph according to the changed weight. This allows the user to operate the parameters while checking the relationship graph, and to intuitively understand how the change operation will affect the parameters.

[0092] [Application example of demand forecasting device] In the present embodiment, the demand prediction device 100 has been mainly described using clothing products as an example. The demand prediction device 100 is applicable to various products and various fields, without being limited to this example. For example, the demand prediction device 100 is applicable as a device for predicting the demand for products related to accessories and decorative items such as bags. That is, the demand prediction device 100 can calculate the demand for decorative items using the features of the product. For example, if the product is a backpack of a specific brand, the category is backpack, and the characteristics are color, material, function, brand name, capacity, and use. In this case, the product features may also include information on measures to be implemented for the product, and information on the external environment at the time of product sales. For example, the demand prediction device 100 can predict the demand for a backpack of a specific brand before its release from information on similar products.

[0093] Furthermore, the demand prediction device 100 can be applied as a device for predicting the demand for food products such as fresh food, processed food, and beverages. That is, the demand prediction device 100 can calculate the demand for food products by using the characteristics of the product. For example, if the product is orange juice, the category is fruit drinks, and the characteristics are ingredients, nutrients, fruit juice ratio, production area of ​​ingredients, content volume, brand name, and the like. In this case, the product characteristics may also include information on measures implemented for the product, and information on the external environment at the time of product sales. For example, the demand prediction device 100 can predict the demand for orange juice before its release from information on similar products.

[0094] In addition, the demand prediction device 100 in the present disclosure can be applied to the medical health care field. For example, the demand prediction device 100 can calculate the demand for a health food product by using the product's features. In this case, the categories are tablets, snacks, jellies, beverages, and the like. The characteristics are the product's efficacy, ingredients, nutrients, and content. The characteristics may also include classifications of health functional foods, such as functional foods, functional foods, and foods for specified health uses. In addition, for health food products, the characteristics may also include information on measures implemented for the product and information on the external environment at the time of product sales. For example, the demand prediction device 100 can predict the demand for a pre-release health food from information on similar post-release health foods.

[0095] In this way, the demand prediction device 100 can support the decision-making of sellers for various products and various fields.

[0096] [Variation 1] When calculating the demand amount using the integrated learning model, the demand forecasting device 100 may present information regarding the difference between the target product and similar products.

[0097] For example, the calculation unit 120 uses the integrated learning model to calculate the base demand for one similar product. That is, the calculation unit 120 inputs the characteristic information of one similar product into the integrated learning model. At this time, the weight for each individual learning model corresponds to one similar product. Furthermore, the calculation unit 120 uses the integrated learning model to calculate the base demand for the target product. That is, the calculation unit 120 inputs the characteristic information of the target product into the integrated learning model. At this time, the weight for each individual learning model corresponds to one similar product.

[0098] Here, if there is no difference between the base demand of the target product and the base demand of the one similar product, the characteristics of the target product and the one similar product are exactly the same. On the other hand, if there is a difference between the base demand of the target product and the base demand of the one similar product, there is some difference between the characteristics of the target product and the one similar product. For example, the calculation unit 120 calculates information indicating the degree to which the difference in each feature between the target product and the one similar product affects the difference in the base demand. That is, the calculation unit 120 may calculate an influence degree indicating the degree to which the difference in the feature between the products affects the difference in the base demand. At this time, the display control unit 150 may display an indicator indicating the influence degree for each feature of the target product that is different from the one similar product.

[0099] [Variation 2] The demand prediction device 100 may calculate the base demand amount and the predicted demand amount using a prediction model other than the integrated learning model.

[0100] In this case, there are multiple types of prediction models. The calculation unit 120 may calculate the base demand amount for the target product for each prediction model. Furthermore, the operation reception unit 130 may receive a selection of a prediction model from a user, and receive an operation of a parameter that affects the output of the demand amount for the selected prediction model. Then, the calculation unit 120 may calculate the predicted demand amount for the target product using the prediction model whose parameter has been operated.

[0101] The multiple types of prediction models may include a machine learning model that learns the relationship between product features and demand. The machine learning model may be generated by various algorithms such as a neural network, a decision tree, and a random forest. In this case, the calculation unit 120 may input, for example, the acquired feature information of the target product to the machine learning model. Then, the calculation unit 120 may calculate the demand output by the machine learning model as the base demand for the target product in a predetermined period.

[0102] The operation receiving unit 130 may receive a change operation for the machine learning model, such as a change operation for the type of feature used in the machine learning model and a weight for each feature. In this case, the calculation unit 120 may calculate a predicted demand amount of the target product in a predetermined period by inputting the feature information of the target product to the machine learning model in which the change operation is reflected. The parameters to be changed are not limited to this example.

[0103] Further, the multiple types of prediction models may include other models that output demand based on similar products similar to the target product. An example of the other models that output demand based on similar products is a model that outputs demand based on the sales record of the product. This model is referred to as a first record model. The first record model may be a model that inputs the identification information of a product, identifies the sales record of the product of the input identification information, and outputs the demand based on the identified sales record. For example, the calculation unit 120 may input the identification information of one similar product into the first record model, and calculate the demand based on the sales record of the one similar product as the base demand of the target product. The one similar product is any one of the similar products identified by the identification unit 110. The one similar product may be, for example, a similar product with the highest similarity. Alternatively, the one similar product may be a similar product selected by the user. The selection of the similar product is accepted by the operation acceptance unit 130.

[0104] For example, when identification information of one similar product is input, the first performance model identifies the sales performance of the one similar product from the performance information in the storage unit 190. For example, when predicting the demand volume of the target product from the launch to a predetermined number of days after the launch, the first performance model outputs the demand volume from the sales performance of the one similar product to the launch to a predetermined number of days after the launch. Then, the calculation unit 120 calculates the output demand volume as the base demand volume of the target product. In this way, the calculation unit 120 may calculate the base demand volume of the target product based on the sales performance of one similar product.

[0105] The operation receiving unit 130 may receive a change operation for the total demand amount for a predetermined period and the demand amount for a unit period for the first performance model. At this time, the calculation unit 120 may calculate the predicted demand amount of the target product by using the first performance model according to the change operation. For example, assume that a change operation for the total demand amount for a predetermined period is received. In this case, the fluctuation of the demand amount for the predetermined period is changed according to the changed total demand amount. The parameters that are the subject of the change operation are not limited to this example.

[0106] Another example of the model that outputs the demand amount based on similar products is a model that outputs the demand amount based on the sales record of multiple products. This model is referred to as the second performance model. The second performance model may be a model that inputs the identification information of multiple products, identifies the sales record of each product of the input identification information, and outputs the demand amount based on the identified sales record, which is obtained by performing a predetermined calculation. The calculation unit 120 may, for example, input the identification information of multiple similar products into the second performance model, and calculate the demand amount based on the sales record of each of the multiple similar products, which is obtained by performing a predetermined calculation, as the base demand amount of the target product.

[0107] An example of the multiple similar products is a predetermined number of similar products with the highest similarity. For example, the calculation unit 120 inputs identification information of similar products with the first to third similarities among the similar products identified by the identification unit 110 to the second performance model. Furthermore, the multiple similar products may be multiple similar products selected by a user. In this case, the calculation unit 120 inputs identification information of the selected multiple similar products to the second performance model.

[0108] The predetermined calculation may be an average. For example, the second performance model may calculate an average value of the demand amounts based on the sales records of the multiple similar products as the base demand amount of the target product. The predetermined calculation may be a weighted average. For example, the second performance model may calculate an average value of the demand amounts based on the sales records of the multiple similar products, taking into account the weights, as the base demand amount of the target product. In this case, the weight may be a value that increases as the similarity increases.

[0109] For example, when predicting the demand for a target product a predetermined number of days after its release, the second performance model calculates the demand for each similar product from its sales performance from the start of its release until a predetermined number of days after its release. Then, a weighted average is taken of the calculated demand, taking into account the weight according to the degree of similarity. The calculation unit 120 then calculates the output demand as the base demand for the target product. In this way, the calculation unit 120 may calculate the base demand for the target product by taking a weighted average of the sales performances of a predetermined number of similar products that rank top in similarity, based on the weight according to the degree of similarity.

[0110] The operation receiving unit 130 may also receive a change operation for the total demand amount for a predetermined period and the demand amount for a unit period for the second performance model. In this case, the calculation unit 120 may calculate the forecast demand amount of the target product by using the first performance model according to the change operation. The parameters that are the subject of the change operation are not limited to this example.

[0111] The display control unit 150 may display output information including the demand amount for each calculated prediction model. FIG. 11 is a diagram showing a third example of the output information. In the example of FIG. 11, base demand amounts corresponding to each of a plurality of prediction models calculated for a target period of three months from April 1, 2023 are shown. The plurality of prediction models are an integrated learning model, a first performance model, a second performance model, and a machine learning model. In this example, the integrated learning model and the first performance model output the demand amount based on the selected similar product "short-sleeved polo shirt". The second performance model outputs the demand amount based on similar products ranked 1st to 3rd in the similar product ranking.

[0112] In this way, the demand prediction device 100 can collectively present the base demand amounts calculated by each of a plurality of types of prediction models.

[0113] Furthermore, the user selects at least one of the prediction models and operates the parameters for the selected prediction model. For example, the user selects the "input" icon for a desired prediction model. The operation receiving unit 130 receives the selection of the prediction model as a result of the icon being selected. Then, the display control unit 150 displays a parameter input screen for the selected prediction model. In this manner, the demand prediction device 100 is capable of receiving parameter operations for each prediction model. Then, the demand prediction device 100 is capable of calculating a predicted demand amount for each prediction model according to the operation.

[0114] [Variation 3] The demand prediction device 100 may be an apparatus integrated with the terminal device 200. For example, the demand prediction device 100 may be an apparatus provided in the terminal device 200. Alternatively, the demand prediction device 100 may be an apparatus realized by installing a program in the terminal device 200.

[0115] <Example of hardware configuration for demand forecasting device> The hardware constituting the demand prediction device of the first and second embodiments described above will be described. Fig. 12 is a block diagram showing an example of the hardware configuration of a computer device constituting the demand prediction device in each embodiment. The demand prediction device and the demand prediction method described in each embodiment and each modified example are realized in a computer device 90. For example, the demand prediction device etc. described in each embodiment and each modified example may have the hardware configuration shown in Fig. 12.

[0116] 12, a computer device 90 includes a processor 91, a RAM (Random Access Memory) 92, a ROM (Read Only Memory) 93, a storage device 94, an input / output interface 95, a bus 96, and a drive device 97. Note that the demand forecasting device and the like may be realized by a plurality of electric circuits.

[0117] The storage device 94 stores a program (computer program) 98. The processor 91 executes the program 98 of the present reception support system using the RAM 92. Specifically, for example, the program 98 includes a program that causes a computer to execute the processes shown in FIG. 3, FIG. 10, etc. As the processor 91 executes the program 98, the functions of the components of the present demand forecasting device are realized. The program 98 may be stored in the ROM 93. The program 98 may also be recorded in the storage medium 80 and read out using the drive device 97, or may be transmitted to the computer device 90 from an external device (not shown) via a network (not shown).

[0118] The input / output interface 95 exchanges data with peripheral devices (such as a keyboard, a mouse, and a display device) 99. The input / output interface 95 functions as a means for acquiring or outputting data. The bus 96 connects each component.

[0119] There are various variations in the method of implementing the demand prediction device. For example, each component included in the demand prediction device can be implemented as a dedicated device. Also, the demand prediction device can be implemented based on a combination of multiple devices.

[0120] The scope of each embodiment also includes a processing method for recording a program for implementing each configuration in the functions of each embodiment on a storage medium, reading the program recorded on the storage medium as code, and executing the program on a computer. That is, a computer-readable storage medium is also included in the scope of each embodiment. Also, the storage medium on which the above-mentioned program is recorded and the program itself are included in each embodiment.

[0121] The storage medium is, for example, a floppy disk, a hard disk, an optical disk, a magneto-optical disk, a CD (Compact Disc)-ROM, a magnetic tape, a non-volatile memory card, or a ROM, but is not limited to these examples. Furthermore, the programs recorded on the storage medium are not limited to programs that execute processes independently, but also include programs that operate on an OS (Operating System) and execute processes in cooperation with other software and functions of an expansion board, which are included in the scope of each embodiment.

[0122] Although the present invention has been described above with reference to the embodiment, the present invention is not limited to the above embodiment. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention.

[0123] Furthermore, the above embodiments and modifications can be combined as appropriate.

[0124] A part or all of the above-described embodiments can be described as, but is not limited to, the following supplementary notes.

[0125] [Appendix 1] A device for predicting a demand for a target product based on a prediction model, comprising: The prediction model is a model in which a plurality of types of individual learning models are integrated and weighted for each of the individual learning models, The individual learning model is a model that outputs a demand amount according to a feature of a product, A means for identifying similar products similar to the target product; a calculation means for calculating a base demand amount for the target product by using the prediction model weighted according to the similar products and the characteristics of the target product; and an operation receiving means for receiving an operation to change the weight for each of the individual learning models, The calculation means calculates a predicted demand quantity for the target product by using the prediction model in which the operation is reflected. Demand forecasting device.

[0126] [Appendix 2] Each of the individual learning models corresponds to a different factor affecting the demand amount; The individual learning model is a model that receives a feature related to a corresponding factor as an input and outputs a demand amount, The prediction model outputs the demand amount by aggregating values ​​obtained by multiplying the demand amount output for each of the individual learning models by a weight. 2. A demand forecasting device as described in claim 1.

[0127] [Appendix 3] The operation receiving means receives an operation for changing a formula representing the individual learning model. 3. A demand forecasting device according to claim 1 or 2.

[0128] [Appendix 4] The individual learning model is a model in which a relationship between product features and demand is learned by machine learning, The operation receiving means receives a change operation for a data set used in the machine learning. 4. A demand forecasting device according to claim 3.

[0129] [Appendix 5] a display control means for displaying a list of the one or more identified similar products; The operation receiving means receives a selection of one similar product, The calculation means calculates a base demand amount for the target product by using the prediction model weighted according to the selected one similar product and the characteristics of the target product. 2. A demand forecasting device as described in claim 1.

[0130] [Appendix 6] the display control means displays a graph showing a relationship between the demand and the individual learning model, the graph highlighting an icon corresponding to the individual learning model according to a weight; 6. A demand forecasting device according to claim 5.

[0131] [Appendix 7] the display control means displays both a parameter input screen and the graph; When the operation receiving means receives an operation for changing the weights of the individual learning models, The display control means updates the graph in accordance with the changed weights. 7. A demand forecasting device according to claim 6.

[0132] [Appendix 8] An acquisition means for acquiring characteristic information indicating characteristics of the target product, The identification means identifies the similar product from a database having characteristic information of a plurality of products by utilizing the characteristic information of the target product. 2. A demand forecasting device as described in claim 1.

[0133] [Appendix 9] A method for predicting a demand for a target product based on a prediction model, comprising: The prediction model is a model in which a plurality of types of individual learning models are integrated and weighted for each of the individual learning models, The individual learning model is a model that outputs a demand amount according to a feature of a product, Identify similar products that are similar to the target product, Calculating a base demand for the target product using the prediction model weighted according to the similar products and the characteristics of the target product; Accepting an operation to change a weight for each of the individual learning models; Calculating a predicted demand for the target product using the prediction model in which the operation is reflected; Demand forecasting methods.

[0134] [Appendix 10] Each of the individual learning models corresponds to a different factor affecting the demand amount; The individual learning model is a model that receives a feature related to a corresponding factor as an input and outputs a demand amount, The prediction model outputs the demand amount by aggregating values ​​obtained by multiplying the demand amount output for each of the individual learning models by a weight. Demand forecasting method as described in Appendix 9.

[0135] [Appendix 11] Accepting a change operation of the formula representing the individual learning model; 11. The demand forecasting method according to claim 9 or 10.

[0136] [Appendix 12] The individual learning model is a model in which a relationship between product features and demand is learned by machine learning, Accepting changes to the data set used in the machine learning; 12. A demand forecasting method as described in Appendix 11.

[0137] [Appendix 13] Displaying a list of the identified one or more similar products; Accept the selection of one similar product, Calculating a base demand for the target product by using the prediction model weighted according to the selected one similar product and the characteristics of the target product; Demand forecasting method as described in Appendix 9.

[0138] [Appendix 14] displaying a graph showing a relationship between the demand and the individual learning model, the graph being highlighted with an icon corresponding to the individual learning model according to a weight; 14. A demand forecasting method as described in Appendix 13.

[0139] [Appendix 15] Displaying both the parameter input screen and the graph, when a weight change operation for each of the individual learning models is received, the graph is updated according to the changed weights. 15. A demand forecasting method as described in Appendix 14.

[0140] [Appendix 16] acquiring characteristic information indicating characteristics of the target product; Using the characteristic information of the target product, the similar product is identified from a database having characteristic information of a plurality of products. Demand forecasting method as described in Appendix 9.

[0141] [Appendix 17] A program for predicting the demand for a target product based on a prediction model, The prediction model is a model in which a plurality of types of individual learning models are integrated and weighted for each of the individual learning models, The individual learning model is a model that outputs a demand amount according to a feature of a product, A process of identifying similar products that are similar to the target product; A process of calculating a base demand amount for the target product using the prediction model weighted according to the similar products and the characteristics of the target product; and receiving an operation to change the weights of the individual learning models. In the calculation process, a predicted demand quantity for the target product is calculated using the prediction model in which the operation is reflected. program.

[0142] [Appendix 18] Each of the individual learning models corresponds to a different factor affecting the demand amount; The individual learning model is a model that receives a feature related to a corresponding factor as an input and outputs a demand amount, The prediction model outputs the demand amount by aggregating values ​​obtained by multiplying the demand amount output for each of the individual learning models by a weight. 18. The program according to claim 17.

[0143] [Appendix 19] In the receiving process, a change operation of a formula representing the individual learning model is received. 19. The program according to claim 17 or 18.

[0144] [Appendix 20] The individual learning model is a model in which a relationship between product features and demand is learned by machine learning, In the receiving process, a change operation for a data group used in the machine learning is received. 19. The program according to claim 19.

[0145] [Appendix 21] further executing a process of displaying a list of the one or more identified similar products; In the receiving process, a selection of one similar product is received; In the calculation process, a base demand quantity for the target product is calculated using the prediction model weighted according to the one selected similar product and the characteristics of the target product. 18. The program according to claim 17.

[0146] [Appendix 22] In the display process, a graph showing a relationship between the demand and the individual learning model is displayed, in which an icon corresponding to the individual learning model is highlighted according to a weight. 22. The program according to claim 21.

[0147] [Appendix 23] In the process of displaying, a parameter input screen and the graph are displayed together, In the receiving process, when a weight change operation for each of the individual learning models is received, In the process of displaying, the graph is updated in accordance with the changed weights. 23. The program according to claim 22.

[0148] [Appendix 24] and further executing a process of acquiring characteristic information indicating characteristics of the target product; In the identifying process, the similar product is identified from a database having characteristic information of a plurality of products by utilizing characteristic information of the target product. 18. The program according to claim 17. [Explanation of symbols]

[0149] 100 Demand forecasting device 110 Specific section 120 Calculation section 130 Operation reception section 140 Acquisition Department 150 Display control unit 190 Memory section 200 Terminal Equipment

Claims

1. A device for predicting a demand for a target product based on a prediction model, comprising: The prediction model is a model in which a plurality of types of individual learning models are integrated and weighted for each of the individual learning models, The individual learning model is a model that outputs a demand amount according to a feature of a product, A means for identifying similar products similar to the target product; a calculation means for calculating a base demand amount for the target product by using the prediction model weighted according to the similar products and the characteristics of the target product; and an operation receiving means for receiving an operation to change the weight for each of the individual learning models, The calculation means calculates a predicted demand quantity for the target product by using the prediction model in which the operation is reflected. Demand forecasting device.

2. Each of the individual learning models corresponds to a different factor affecting the demand amount; The individual learning model is a model that receives a feature related to a corresponding factor as an input and outputs a demand amount, The prediction model outputs the demand amount by aggregating values ​​obtained by multiplying the demand amount output for each of the individual learning models by a weight. The demand forecasting device according to claim 1 .

3. The operation receiving means receives an operation for changing a formula representing the individual learning model. The demand forecasting device according to claim 1 or 2.

4. The individual learning model is a model in which a relationship between product features and demand is learned by machine learning, The operation receiving means receives a change operation for a data set used in the machine learning. The demand forecasting device according to claim 3.

5. a display control means for displaying a list of the one or more identified similar products; The operation receiving means receives a selection of one similar product, The calculation means calculates a base demand amount for the target product by using the prediction model weighted according to the selected one similar product and the characteristics of the target product. The demand forecasting device according to claim 1 .

6. the display control means displays a graph showing a relationship between the demand and the individual learning model, the graph highlighting an icon corresponding to the individual learning model according to a weight; The demand forecasting device according to claim 5.

7. the display control means displays both a parameter input screen and the graph; When the operation receiving means receives an operation for changing the weights of the individual learning models, The display control means updates the graph in accordance with the changed weights. The demand forecasting device according to claim 6.

8. An acquisition means for acquiring characteristic information indicating characteristics of the target product, The identification means identifies the similar product from a database having characteristic information of a plurality of products by utilizing the characteristic information of the target product. The demand forecasting device according to claim 1 .

9. A method for predicting a demand for a target product based on a prediction model, comprising: The prediction model is a model in which a plurality of types of individual learning models are integrated and weighted for each of the individual learning models, The individual learning model is a model that outputs a demand amount according to a feature of a product, Identify similar products that are similar to the target product, Calculating a base demand for the target product using the prediction model weighted according to the similar products and the characteristics of the target product; Accepting an operation to change a weight for each of the individual learning models; Calculating a predicted demand for the target product using the prediction model in which the operation is reflected; Demand forecasting methods.

10. A program for predicting the demand for a target product based on a prediction model, The prediction model is a model in which a plurality of types of individual learning models are integrated and weighted for each of the individual learning models, The individual learning model is a model that outputs a demand amount according to a feature of a product, A process of identifying similar products that are similar to the target product; A process of calculating a base demand amount for the target product using the prediction model weighted according to the similar products and the characteristics of the target product; and receiving an operation to change the weights for each of the individual learning models. In the calculation process, a predicted demand quantity for the target product is calculated using the prediction model in which the operation is reflected. program.

Images