Information processing device, information processing method and information processing program

Abstract

To perform highly accurate prediction at the smallest unit required by a user.SOLUTION: An information processing device according to one embodiment of the present invention includes: a reception unit that receives a selection of a prediction unit, which is a unit for predicting an objective variable from among management items having a hierarchical relation, and receives from a user a selection of a prediction target to be a target for the objective variable from among management contents linked to each management item; a generation unit that generates definition information that defines contents of the prediction target on the basis of the management item selected as the prediction unit and the management content selected as the prediction target; and a learning unit that uses learning data extracted based on the definition information to learn a model for predicting the objective variable in the units indicated by the definition information.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] Data analysis using a predictive model has been known. For example, when automatically generating candidate objective variables, a method is known that makes it easy for a user to select candidate objective variables. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2020-135055 Summary of the Invention [Problem to be solved by the invention]

[0004] However, when making predictions in specific, small units, such as SKU (Stockkeeping Unit), it is necessary to perform a large amount of processing, such as aggregating data according to the unit, generating a prediction model, or predicting explanatory variables, etc. Therefore, there is a demand for highly accurate predictions at the smallest unit desired by the user.

[0005] The present invention has been made in view of the above, and proposes an information processing device, an information processing method, and an information processing program that enable highly accurate prediction in the smallest unit desired by the user. [Means for solving the problem]

[0006] In order to solve the above problems, an information processing device of one embodiment of the present invention includes a reception unit that receives a selection of a prediction unit, which is a unit for predicting a dependent variable from management items having a hierarchical relationship, and receives from a user a selection of a prediction target that is the target of the dependent variable from management contents linked to each of the management items; a generation unit that generates definition information that defines the content of the prediction target based on the management items selected as the prediction units and the management contents selected as the prediction target; and a learning unit that uses learning data extracted based on the definition information to train a model that predicts the dependent variable in units indicated by the definition information. [Brief explanation of the drawings]

[0007] [Figure 1] Figure 1 shows an overview of the predictive management DX service. [Figure 2] FIG. 2 is a diagram illustrating an example of the configuration of a server according to the embodiment. [Figure 3] FIG. 3 is a diagram illustrating an example of the organization / item hierarchy master according to the embodiment. [Figure 4] FIG. 4 is a diagram showing a specific example of a method for generating definition information. [Figure 5] FIG. 5 is a diagram illustrating an example of a procedure for generating prediction model data. [Figure 6] FIG. 6 is a diagram illustrating an example of an explanatory variable prediction process procedure. [Figure 7] FIG. 7 is a diagram illustrating an example of a procedure for predicting a response variable. [Figure 8] FIG. 8 is a block diagram illustrating an example of the hardware configuration of the information processing device according to the embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0008] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. Note that the information processing device, information processing method, and information processing program according to the present disclosure are not limited to these embodiments. In the following embodiments, the same components are designated by the same reference numerals, and redundant description will be omitted.

[0009] [Embodiment] 1. Introduction In today's business environment, where dramatic changes occur in a short space of time, companies are required to improve their management decisions and speed. For example, companies are required to improve the accuracy and speed of their objective planning and make quick decisions.

[0010] However, many companies continue to make traditional management decisions based on intuition and experience, or consider preventative measures that analyze results based on past performance.Meanwhile, performance management services that utilize big data and AI (Artificial Intelligence) to forecast demand or market trends are gaining attention.

[0011] Therefore, the inventors of the present invention have come up with the realization of a predictive management DX service as shown in Figure 1. Figure 1 is a diagram showing an overview of the predictive management DX service. The predictive management DX service is a service that promotes advanced management through decision-making based on predictions made by AI, and it is possible to collect and accumulate a wide variety of data, such as external data that affects a company's performance, and combine this data with internal data to perform predictive analysis of the future.

[0012] In this way, predictive management DX services create a complex cycle of AI-based predictions in addition to traditional business management (e.g., PCDA operations), thereby creating the advantage of being able to handle not only traditional forecast-actual analysis but also forecast analysis, which analyzes the difference between plans and forecasts.

[0013] Information processing related to the proposed technology (hereinafter, sometimes referred to as "information processing related to the embodiment") is incorporated into a predictive management DX service as shown in FIG. 1. For example, the predictive management DX service provides business processes (business templates) for supporting forecast analysis. As shown in FIG. 1, the business processes include "data selection, collection, and processing," "AI-based forecasting," and "data analysis and utilization." Of these, the information processing related to the embodiment corresponds to "AI-based forecasting." Furthermore, the information processing related to the embodiment achieves both highly accurate forecasting and reduced operational load related to the forecasting.

[0014] There is a need to create predictive models at finer units, such as SKUs or customers per SKU. However, conventional methods require data aggregation according to the unit, and require a large amount of work, such as aggregating target variable and explanatory variable data, generating predictive models, and predicting the explanatory variables. Furthermore, the large number of models generated must be maintained and managed. For these reasons, conventional methods lack practicality due to the difficulty of operating predictive models. For this reason, current methods involve generating models at coarser units, even if this reduces prediction accuracy, or reducing operational load by limiting the number of items to be predicted.

[0015] To address the above issues, one possible method is to apply more complex models such as neural networks to predict multiple targets with a single model. However, because complex models require a large amount of training data, they are not suitable for forecasting small amounts of data, such as monthly sales volume forecasts.

[0016] In view of the above-mentioned problems, the information processing according to the embodiment is executed for the purpose of achieving both highly accurate prediction and reduced operational load related to the prediction in a predictive management DX service. For example, the information processing according to the embodiment may be performed by a server device 100 described in FIG. 2. The server device 100 is an example of an information processing device.

[0017] The server device 100 is a platform that integrates a data warehouse and a prediction system in a predictive management DX service.

[0018] The server device 100 provides a user interface that allows a user to select an organizational hierarchy, a product type hierarchy, or the like as a prediction unit. Then, for each prediction target unit under the selected organizational hierarchy or product type hierarchy, data on the objective variable and data on the explanatory variables are extracted and aggregated, and a model is generated and a prediction is made for each prediction target unit. For example, if a user selects the product type hierarchy, a prediction model is trained for each SKU under the product type hierarchy, and a prediction is made using the prediction model. As a result, the user can grasp and analyze changes in the target product and indicators that affect the target product, for example, before grasping actual values, allowing the user to take more specific actions toward achieving their goals.

[0019] The server device 100 also provides a user interface that allows a user to issue an instruction to update multiple models at once. As a result, the user can update multiple models at once, rather than on a model-by-model basis, which provides greater convenience in managing the models.

[0020] [2. Server Device Configuration] The server device 100 according to the embodiment will be described with reference to Fig. 2. Fig. 2 is a diagram showing an example of the configuration of the server device 100 according to the embodiment. As shown in Fig. 2, the server device 100 may include a data warehouse 20 and a control unit 130.

[0021] (About Data Warehouse 20) In the data warehouse 20, data is stored in chronological order. This allows the history of past data to be checked, and at the same time, data for each system can be aggregated and used as overall data. In other words, the data warehouse 20 can be said to be an effective database for corporate decision-making. Also, according to the example of FIG. 2, the data warehouse 20 may include an organization / item hierarchy master DB 21, a forecast unit data DB 22, a forecast target data DB 23, an explanatory variable performance data DB 24, and a target variable performance data DB 25. Note that the data warehouse 20 may exist in a cloud different from the server device 100.

[0022] (About the Organization / Item Hierarchy Master DB21) An example of the organization / item hierarchy master DB 21 according to this embodiment will now be shown with reference to Figure 3. As shown in Figure 3, the organization / item hierarchy master DB 21 is hierarchically organized. For example, as shown in Figure 3, the organization / item hierarchy master DB 21 has the management item "company" in the first hierarchical level, the management item "sales organization" in the second hierarchical level, the management item "product type" in the third hierarchical level, and the management item "item" in the fourth hierarchical level. Note that although the organization / item hierarchy master DB 21 is hierarchically organized into four hierarchical levels in Figure 3, the number of hierarchical levels is not limited, and the types of management items are not limited to the example in Figure 3.

[0023] "Management content" is the specific content of a "management item" and is linked to the subordinate of the "management item." According to the example in FIG. 3, management content "C1," management content "S1," management content "G1," and item "M11" are associated with each other according to the hierarchical relationship of "company" → "sales organization" → "product type" → "item." In this example, company "C1" has sales organization "S1," and sales organization "S1" manages products belonging to product type "G1" and item "M11."

[0024] As will be described later, the server device 100 also accepts the selection of a forecast unit and a forecast target based on the contents stored in the organization / item hierarchy master DB 21.

[0025] (About prediction unit data DB22) 2, the prediction unit data DB 22 may store a management item selected by a user as a prediction unit from among management items configured in a hierarchical structure. The prediction unit here is a unit for predicting a dependent variable.

[0026] (About prediction target data DB23) The prediction target data DB23 may store management contents selected by a user as a prediction target from management contents managed for each management item configured in a hierarchical structure. The prediction target here is the target indicated by the objective variable. For example, if the objective variable is "sales volume," the prediction target indicates the target, i.e., what sales volume it is.

[0027] (About the explanatory variable performance data DB24) The explanatory variable performance data DB 24 stores performance data used to predict explanatory variables. For example, the explanatory variable performance data DB 24 may store in-house data used in the predictive management DX service as performance data. The in-house data referred to here includes ERP (Enterprise Resource Planning) data, various performance data (e.g., sales performance, production performance, and logistics performance), and element data that influences prediction.

[0028] For example, if a user wants to predict sales volume based on the prediction unit and prediction target selected by the user, the user may specify a column from the actual data stored in the explanatory variable actual data DB24 that is linked to factors that affect the sales volume prediction (e.g., sales promotion expenses, number of complaints, import volume, etc.).

[0029] (Regarding DB25 target variable performance data) The objective variable performance data DB 25 stores performance data used for predicting objective variables. For example, similar to the explanatory variable performance data DB 24, the objective variable performance data DB 25 may store in-house data including ERP data, various performance data (e.g., sales performance, production performance, and logistics performance), and element data that may have an effect on making predictions.

[0030] Here, the server device 100 may not only predict the objective variable corresponding to the final result desired by the user, but also predict the explanatory variables used to predict the objective variable. In a simple example, the server device 100 may use performance data corresponding to the objective variable as learning data. However, there are cases where the accuracy of predicting the objective variable can be improved by using the predicted value predicted from the performance data as the explanatory variable (which can be considered as the future value of the explanatory variable).

[0031] Therefore, the server device 100 may also train a prediction model (hereinafter referred to as a "first prediction model") for predicting explanatory variables used to predict the dependent variable.The server device 100 may then train a prediction model (hereinafter referred to as a "second prediction model") for predicting the final dependent variable using explanatory variables extracted from performance data and future explanatory variables predicted by the first prediction model.

[0032] According to this example, learning data for generating a first prediction model is extracted from the performance data stored in the explanatory variable performance data DB 24. For example, a user can specify the learning data to be used for training the first prediction model from the performance data stored in the explanatory variable performance data DB 24. For example, a user can specify a column associated with a factor (e.g., sales promotion expenses, number of complaints, import volume, etc.) that affects the prediction of the explanatory variables from the performance data stored in the explanatory variable performance data DB 24.

[0033] Furthermore, learning data for generating a second prediction model is extracted from the performance data stored in the objective variable performance data DB 25. For example, the user can specify learning data to be used for training the second prediction model from the performance data stored in the objective variable performance data DB 25. For example, the user can specify a column associated with a factor (e.g., sales volume) that affects the prediction of the objective variable from the performance data stored in the objective variable performance data DB 25.

[0034] (Regarding the control unit 130) The control unit 130 is realized by a CPU (Central Processing Unit), an MPU (Micro Processing Unit), or the like executing various programs (for example, the information processing program according to the embodiment) stored in a storage device inside the server device 100 using RAM as a work area. The control unit 130 is also realized by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

[0035] As shown in Fig. 2, the control unit 130 has a reception unit 131, a model data generation unit 132, a response variable control unit 133, and an explanatory variable control unit 134, and realizes or executes the functions and actions of the information processing described below. Note that the internal configuration of the control unit 130 is not limited to the configuration shown in Fig. 2, and may have other configurations as long as they perform the information processing described below. Furthermore, the connection relationships between the processing units included in the control unit 130 are not limited to the connection relationships shown in Fig. 2, and may be other connection relationships.

[0036] (Regarding the reception unit 131) The receiving unit 131 receives from the user a selection of a prediction unit, which is a unit for predicting a dependent variable from among management items having a hierarchical relationship. The receiving unit 131 also receives from the user a selection of a prediction target, which is a target of the dependent variable, from among the management contents linked to each management item. For example, the receiving unit 131 may receive the selection of a prediction unit or a prediction target based on information stored in the organization / item hierarchy master DB 21.

[0037] For example, the reception unit 131 may receive a selection of a prediction unit from among management items that are classifications in a business management database (i.e., organization / item hierarchy master DB21) owned by the organization to which the user belongs, and may also receive a selection of a prediction target from among management contents linked to each management item in the database. A specific example of a selection reception method will be described later.

[0038] The receiving unit 131 may further receive a request to update a prediction model. For example, the receiving unit 131 may receive an update request that is an instruction to update multiple second models collectively.

[0039] (Regarding the model data generation unit 132) The model data generation unit 132 generates definition information that defines the contents of the prediction target based on the management item selected as the prediction unit and the management content selected as the prediction target. Specifically, the model data generation unit 132 searches the organization / item hierarchy master DB 21 for a combination of management contents that satisfies both the management item selected as the prediction unit and the management content selected as the prediction target, and generates definition information for each combination of management contents found.

[0040] For example, the model data generation unit 132 acquires, as elements of definition information, a combination of management contents linked to the management item selected as the prediction unit among the management contents linked to each management item based on the arrangement of management contents according to the axis of the hierarchical relationship and the management contents selected as the prediction target. More specifically, the model data generation unit 132 acquires, as elements of definition information, a combination of management contents linked to the management item selected as the prediction unit among the management contents constituting the arrangement of management contents according to the axis of the hierarchical relationship, which is an arrangement of management contents according to the axis of the hierarchical relationship and includes the management contents selected as the prediction target. Specific examples of methods for generating definition information will be described later.

[0041] As described above, the server device 100 may also train a first prediction model, which is a prediction model for predicting explanatory variables. The server device 100 does not necessarily need to use future values ​​of the explanatory variables when predicting the dependent variable, but using future values ​​of the explanatory variables is preferable in terms of improving prediction accuracy. Therefore, when training the first prediction model and the second prediction model, the model data generation unit 132 may separately generate definition information to be used in training the first prediction model and definition information to be used in training the second prediction model, based on the management item selected as the prediction unit and the management content selected as the prediction target.

[0042] Furthermore, the model data generation unit 132 may store, in a predetermined storage unit, definition information (data for the explanatory variable prediction model) to be used in training the first prediction model and definition information (data for the dependent variable prediction model) to be used in training the second prediction model. As a result, the dependent variable control unit 133 acquires the definition information stored in the predetermined storage unit and uses it for training and prediction related to the dependent variables. Furthermore, the dependent variable control unit 134 acquires the definition information stored in the predetermined storage unit and uses it for training and prediction related to the dependent variables.

[0043] (Regarding the objective variable control unit 133) As shown in FIG. 2, the objective variable control unit 133 may include an objective variable data extraction unit 133a, a learning unit 133b, and a prediction unit 133c.

[0044] (Regarding the objective variable data extraction unit 133a) The objective variable data extraction unit 133a extracts learning data for each definition information. The objective variable data extraction unit 133a may extract, as learning data, performance data related to the management content selected as the prediction target, which satisfies the definition information from performance data corresponding to a period specified by the user. Specifically, the objective variable data extraction unit 133a may extract learning data from the objective variable performance data DB 25. The learning data may be composed of a combination of explanatory variables and objective variables. The explanatory variables may include only performance values ​​of the explanatory variables, or, if a first prediction model has been generated, may further include future values ​​of the explanatory variables predicted using the first prediction model.

[0045] (Regarding the learning unit 133b) The learning unit 133b uses the learning data extracted by the dependent variable data extraction unit 133a to learn a model for predicting the dependent variable for each piece of definition information. Since the definition information defines the specific content of the prediction unit for predicting the dependent variable, the learning unit 133b learns a second prediction model for predicting the dependent variable in the prediction unit defined in the definition information.

[0046] (Regarding the prediction unit 133c) The prediction unit 133c predicts a dependent variable corresponding to the prediction target selected by the user based on the second prediction model. As an example, the prediction unit 133c inputs prediction data into the second prediction model to predict the sales volume of the prediction target selected by the user.

[0047] (Regarding the explanatory variable control unit 134) As shown in FIG. 2, the explanatory variable control unit 134 may include an explanatory variable data extraction unit 134a, a learning unit 134b, and a prediction unit 134c.

[0048] (Regarding the explanatory variable data extraction unit 134a) The explanatory variable data extraction unit 134a extracts learning data for each definition information. The explanatory variable data extraction unit 134a may extract, as learning data, performance data related to the management content selected as the prediction target, which performance data corresponds to a period specified by the user and satisfies the definition information. Specifically, the explanatory variable data extraction unit 134a may extract the learning data from the explanatory variable performance data DB 24.

[0049] (Regarding the learning unit 134b) The learning unit 134b uses the learning data extracted by the explanatory variable data extraction unit 134a to learn a model for predicting explanatory variables for each piece of definition information. Since the definition information defines the specific content of the prediction target for predicting explanatory variables, the learning unit 134b learns a first prediction model for predicting explanatory variables in a prediction unit corresponding to the definition information (a prediction unit selected by the user).

[0050] (Regarding the prediction unit 134c) The prediction unit 134c predicts explanatory variables according to the prediction target selected by the user based on the first prediction model. As an example, the prediction unit 133c inputs prediction data into the first prediction model to predict the sales promotion expenses (which may be the number of complaints, the import volume, etc.) of the prediction target selected by the user.

[0051] [3. Specific examples of reception methods] A specific example of the reception method will be described with reference to Fig. 3. Fig. 3 is a diagram showing a specific example of the reception method for receiving the selection of a prediction unit and a prediction target.

[0052] For example, the reception unit 131 may generate a selection screen G1 for receiving a selection of a forecast unit from among the management items based on the hierarchical structure stored in the organization / item hierarchy master DB21, and deliver the selection screen G1 to the user's terminal device. For example, as shown in FIG. 3, the reception unit 131 may generate a selection screen G1 including a management item list L1 in which an "input field" is associated with each "management item." FIG. 3 shows an example in which, of the four management items, "company," "sales organization," "product type," and "item," check marks are entered in the input field corresponding to "company," the input field corresponding to "sales organization," and the input field corresponding to "item." This example means that the user has selected "company," "sales organization," and "item" as forecast units.

[0053] Furthermore, the selection screen G1 is provided with a button B1, and when the button B1 is pressed, the screen transitions to a selection screen G2. For example, assume that the user selects the button B1. In this case, the reception unit 131 generates the selection screen G2 for receiving from the user the selection of a prediction target from among the management contents associated with each management item, based on the hierarchical structure stored in the organization / item hierarchy master DB21. For example, as shown in FIG. 3, the reception unit 131 may generate the selection screen G2 including a management content list L2 that lists the management contents existing under each management item.

[0054] According to the example of Fig. 3, the selection screen G2 is provided with input fields F21 and F22 for inputting a prediction target selected from the management contents, and a button B2 for confirming the contents selected on the selection screen G1 and the selection screen G2. Fig. 3 shows an example in which, of the management contents displayed in the management content list L2, management content "C1" (i.e., "company C1") under the management item "company" and management content "G1" (i.e., "product type G1") under the management item "product type" have been selected as prediction targets.

[0055] The following describes an example in which processing is performed based on the selections shown in Fig. 3. Specifically, the operation of the server device 100 will be explained using an example in which a user selects the management item "company," the management item "sales organization," and the management item "product type" as the prediction unit, and selects "company C1" and "product type G1" as the prediction targets.

[0056] [4. Method for generating objective variables] Next, a method for generating definition information will be described. In the example of FIG. 3 , the user's selection indicates a request to "predict the sales volume of product G1 handled by company C1 at each company, sales organization, and item level." In this case, a conventional method would require data aggregation for all items under product G1 for all sales organizations within company C1 that handle product G1, at each company, sales organization, and item level. Furthermore, a large amount of work would be required, including aggregation of objective variable and explanatory variable data, model generation, and predicting the explanatory variables. Therefore, the proposed method of the present invention redefines the prediction target for a combination of existing data. More specifically, the proposed method of the present invention redefines the prediction target according to the combination of data corresponding to the management item selected as the prediction unit. That is, the proposed method of the present invention automatically defines the content of the prediction target based on the management item selected as the prediction unit and the management object selected as the prediction target. Then, training data is extracted based on the definition information, and a model that predicts the objective variable for the defined prediction unit is trained using the extracted training data.

[0057] As a result, it is possible to achieve both highly accurate prediction and reduced operational load related to the prediction. A method for generating definition information will be described below with reference to Fig. 4. Fig. 4 is a diagram showing a specific example of a method for generating definition information.

[0058] For example, the model data generation unit 132 searches for management contents by comparing the organization / item hierarchy master DB 21 with the selection contents selected by the user, and generates definition information for each combination of management contents obtained as the search results. Specifically, the model data generation unit 132 searches the organization / item hierarchy master DB 21 using both the management item selected as the prediction unit and the management contents selected as the prediction target as search conditions, and obtains combinations of management contents that satisfy the search conditions as elements of the definition information. The model data generation unit 132 then generates definition information by combining the elements of the definition information with the management contents selected as the prediction target. A specific example of such a generation method is shown below.

[0059] First, according to the organization / item hierarchy master DB 21 shown in Figure 4, management contents are not only vertically linked as subordinates of management items, but also have an arrangement relationship according to the horizontal hierarchical axis of "first level," "second level," "third level," and "fourth level." Figure 4 shows arrays H1, H2, H3, H4, and H5 as arrangements of management contents according to the hierarchical axis. For example, array H1 is composed of management content "C1" (company C1), management content "S1" (sales organization S1), management content "G1" (product type G1), and management content "M11" (item M11).

[0060] In this state, the model data generation unit 132 extracts arrays that include the management details (C1, G1) selected as the prediction target from among the arrays (H1, H2, H3, H4, H5) corresponding to the hierarchical axes (step S41). According to the organization / item hierarchy master DB 21 shown in Fig. 4, the model data generation unit 132 can extract arrays H1, H2, H4, and H5 as arrays that include the management details (C1, G1) selected as the prediction target.

[0061] The model data generation unit 132 also acquires a combination of management contents linked to the management items (company, sales organization, item) selected as the forecast unit for each of the arrays (H1, H2, H4, H5) extracted in step S42 (step S42). According to the organization / item hierarchy master DB 21 shown in FIG. 4, the model data generation unit 132 can acquire a combination CB1 of management contents "C1," "S1," and "M11" for array H1. The model data generation unit 132 can acquire a combination CB2 of management contents "C1," "S1," and "M12" for array H2. The model data generation unit 132 can acquire a combination CB4 of management contents "C1," "S2," and "M11" for array H4. Furthermore, the model data generation unit 132 can acquire a combination CB5 of management contents "C1", "S2", and "M12" for the array H5.

[0062] In addition, combination CB1, combination CB2, combination CB4, and combination CB5 are each elements of definition information, and the model data generation unit 132 generates data that combines these elements of definition information with the management content (C1, G1) selected as the prediction target as definition information (step S43).

[0063] Specifically, the model data generation unit 132 generates definition information D1 by combining the combination CB1 (C1, S1, M11) and the management content (C1, G1). As shown in FIG. 4, the definition information D1 is composed of the management content (C1, G1) selected as the prediction target and the management content (C1, S1, M11) distinguished according to the management item (company, sales organization, product type) selected as the prediction unit. The definition information D1 means "to generate a second prediction model M21 that predicts the sales volume of a product of product type G1 handled by company C1, for each of company C1, sales organization S1, and product type M11."

[0064] The model data generation unit 132 also generates definition information D2 by combining the combination CB2 (C1, S1, M12) and the management content (C1, G1). As shown in FIG. 4, the definition information D2 is composed of the management content (C1, G1) selected as the prediction target and the management content (C1, S1, M12) distinguished according to the management item (company, sales organization, product type) selected as the prediction unit. The definition information D2 means "to generate a second prediction model M22 that predicts the sales volume of a product of product type G1 handled by company C1, for each of company C1, sales organization S1, and product type M12."

[0065] The model data generation unit 132 also generates definition information D4 by combining the combination CB4 (C1, S2, M11) with the management content (C1, G1). As shown in FIG. 4, the definition information D4 is composed of the management content (C1, G1) selected as the prediction target and the management content (C1, S2, M11) differentiated according to the management item (company, sales organization, product type) selected as the prediction unit. The definition information D4 means "to generate a second prediction model M24 that predicts the sales volume of a product of product type G1 handled by company C1, for each of company C1, sales organization S2, and product type M11."

[0066] The model data generation unit 132 also generates definition information D5 by combining the combination CB4 (C1, S2, M12) and the management content (C1, G1). As shown in FIG. 4, the definition information D5 is composed of the management content (C1, G1) selected as the prediction target and the management content (C1, S2, M12) distinguished according to the management item (company, sales organization, product type) selected as the prediction unit. The definition information D5 means "to generate a second prediction model M25 that predicts the sales volume of a product of product type G1 handled by company C1, for each of company C1, sales organization S2, and product type M12."

[0067] [5. Processing Procedure] From here, the operation procedure of the server device 100 will be described with reference to Fig. 5 to Fig. 7. The procedure of the prediction model data generation process will be described with reference to Fig. 5. The procedure of the explanatory variable prediction process will be described with reference to Fig. 6. Furthermore, the procedure of the response variable prediction process will be described with reference to Fig. 7.

[0068] [5-1. Prediction model data generation process] 5 is a diagram illustrating an example of a procedure for a prediction model data generation process. The prediction model data generation process may be performed based on the definition information generation method described with reference to FIG.

[0069] First, the server device 100 processes a processing request event requesting generation processing for generating prediction model data (step S51). The prediction model data here refers to definition information (objective variable prediction model data) to be used for training the second prediction model and definition information (explanatory variable prediction model data) to be used for training the first prediction model.

[0070] In step S51, when the control unit 130 is started by the scheduler, it acquires information indicating unprocessed processing request events (step S51-1). Note that this information may also include information about processing request events that were not processed due to a timeout. Then, the control unit 130 calls the prediction model data generation AP1 (step S51-2). That is, the control unit 130 outputs information indicating the unprocessed processing request events to the model data generation unit 132, and instructs the model data generation unit 132 to generate prediction model data corresponding to the unprocessed processing request events.

[0071] The model data generating unit 132 performs a prediction model data generating process under the control of the control unit 130 (step S52).

[0072] In step S52, the model data generation unit 132 generates dependent variable prediction model data using the method described with reference to FIG. 4, and registers the generated data in a table (step S52-1).

[0073] Next, the model data generation unit 132 searches the feature store to determine whether or not a predicted value of the explanatory variable exists for the registered dependent variable prediction model data (step S52-2). If the search results in a predicted value of the explanatory variable, the model data generation unit 132 updates the information so that the predicted value is used for training the second prediction model. For example, if the feature store is available (if a predicted value of the explanatory variable exists), the model data generation unit 132 updates the physical table name and item name of the record of the explanatory variable in the prediction model variable table and prediction condition explanatory variable table of the dependent variable prediction model data to those in the feature store table.

[0074] If the search result shows that there is no predicted value for the explanatory variable, the model data generation unit 132 generates explanatory variable prediction model data using the method described in FIG. 4 and registers the generated data in a table (step S52-3).

[0075] Next, the model data generation unit 132 registers processing requests in the ML (machine learning) processing request table and the ML processing request detail table for the dependent variable prediction model data generated in step S52-1 and the explanatory variable prediction model data generated in step S52-3 (step S52-4).

[0076] Finally, the model data generation unit 132 registers information about a processing request event that requests an explanatory variable prediction process (step S52-5).

[0077] [5-2. Explanatory variable prediction processing procedure] 6 is a diagram illustrating an example of a procedure for an explanatory variable prediction process. In the explanatory variable prediction process, a first prediction model for predicting explanatory variables is trained, and the explanatory variables are predicted using the first prediction model.

[0078] First, the server device 100 processes a processing request event requesting explanatory variable prediction processing (step S61). In step S61, when the control unit 130 is activated by the scheduler, it acquires information indicating an unprocessed processing request event (step S61-1). Then, the control unit 130 registers the information indicating the processing request event in a queue (step S61-2).

[0079] The explanatory variable control unit 134 receives control from the queue and executes a startup process to start the explanatory variable prediction process (step S62). In step S62, the explanatory variable control unit 134 acquires data to be processed from the ML processing request table and the ML processing request detail table based on the information of the processing request event (step S62-1). Then, the explanatory variable control unit 134 outputs the processing request ID registered in the ML processing request table and the ML processing request detail ID registered in the ML processing request detail table to the explanatory variable data extraction unit 134a (step S62-2). This starts the explanatory variable prediction process.

[0080] The explanatory variable control unit 134 executes explanatory variable prediction processing in response to the startup processing of step S62 (step S63). In step S63, the explanatory variable data extraction unit 134a extracts learning data to be used for training the first prediction model from the explanatory variable performance data DB 24 based on the explanatory variable prediction model data (definition information) generated by the model data generation unit 132 in step S52-3 (step S63-1). For example, the explanatory variable data extraction unit 134a extracts performance data that satisfies the explanatory variable prediction model data and corresponds to a period specified by the user from the explanatory variable performance data DB 24. Then, the explanatory variable data extraction unit 134a generates learning data by processing the performance data based on the explanatory variable prediction model data. In the explanatory variable prediction processing, the explanatory variables are treated as response variables in the learning data.

[0081] The learning unit 134b executes a learning process to learn a first prediction model based on the learning data extracted in step S63-1 (step S63-2). For example, when multiple pieces of definition information (explanatory variable prediction model data) have been generated as in the example of Fig. 4, the learning unit 134b learns a first prediction model for each piece of definition information. For example, when four sets of definition information exist, the learning unit 134b learns four first prediction models.

[0082] The prediction unit 134c predicts the explanatory variables using the first prediction model trained in step S63-2 (step S63-3). For example, the prediction unit 134c predicts the explanatory variables by inputting data corresponding to an array of future dates (an array of management contents) for the prediction period from the end date when the training of the first prediction model is completed.

[0083] Here, in the subsequent dependent variable prediction process, a Granger causality test is used to determine whether each explanatory variable according to the user's selection contributes to the prediction of the dependent variable. However, performing a Granger causality test requires that the data is stationary. Therefore, the explanatory variable control unit 134 performs a stationarity test (step S63-4). In step S63-4, the explanatory variable control unit 134 may exclude explanatory variables determined to be non-stationary and perform control so that processing is performed using explanatory variables that are stationary.

[0084] Next, the explanatory variable control unit 134 registers the learning data extracted in step S63-1 and the predicted value predicted in step S63-3 as a feature store (step S63-5).

[0085] Furthermore, the explanatory variable control unit 134 rewrites the explanatory variables of the dependent variable prediction model data into the feature store (step S63-6).

[0086] Finally, the explanatory variable control unit 134 registers information about a processing request event that requests a response variable prediction process (step S63-7).

[0087] [5-3. Procedure for predicting dependent variables] 7 is a diagram showing an example of a procedure for a dependent variable prediction process. In the dependent variable prediction process, a second prediction model for predicting a dependent variable is trained, and the dependent variable is predicted using the second prediction model.

[0088] First, the server device 100 processes a processing request event that requests a target variable prediction process (step S71). In step S71, when the control unit 130 is started by the scheduler, it acquires information indicating an unprocessed processing request event (step S71-1). Then, the control unit 130 registers the information indicating the processing request event in a queue (step S71-2).

[0089] The objective variable control unit 133 receives control from the queue and executes a startup process to start the objective variable prediction process (step S72). In step S72, the objective variable control unit 133 acquires data to be processed from the ML processing request table and the ML processing request detail table based on the information of the processing request event (step S72-1).

[0090] Furthermore, the response variable control unit 133 determines whether or not all of the preceding explanatory variable prediction processes have been completed (step S72-2). If it is determined in step S72-2 that the explanatory variable prediction processes have not been completed, the process ends at this point.

[0091] On the other hand, if the objective variable control unit 133 determines in step S72-3 that the explanatory variable prediction process has ended, it outputs the processing request ID registered in the ML processing request table and the ML processing request detail ID registered in the ML processing request detail table to the objective variable data extraction unit 133a (step S72-3), thereby starting the objective variable prediction process.

[0092] In response to the startup process of step S72, the response variable control unit 133 executes a response variable prediction process (step S73). In step S73, the response variable control unit 133 determines whether the current response variable prediction process is for newly generating a second prediction model or for updating the second prediction model, and may perform the following process depending on the determination result.

[0093] If the current target variable prediction process is an update of the second prediction model, the target variable control unit 133 determines whether there is any change in the performance data for a specific period used when the second prediction model was previously generated (step S73-1).

[0094] If the objective variable control unit 133 determines in step S73-1 that there is a change in the performance data for a specific period, it performs a Granger causality test again, rather than applying the Granger causality test result (stationarity test result) from the previous time the second prediction model was generated. That is, if the objective variable control unit 133 determines in step S73-1 that there is a change in the performance data for a specific period, it treats the update of the second prediction model as the generation of a new second prediction model. Specifically, the objective variable control unit 133 extracts data for Granger causality testing (step S73-2), in the same way as if it were determined in step S73-1 that there is no change in the performance data for a specific period.

[0095] Then, the objective variable control unit 133 executes a Granger causality test (step S73-3). Specifically, the objective variable control unit 133 determines whether or not each explanatory variable according to the user's selection contributes to the prediction of the objective variable by the Granger causality test.

[0096] The objective variable data extraction unit 133a extracts learning data to be used for learning the second prediction model from the objective variable performance data DB 25 based on the objective variable prediction model data (definition information) generated by the model data generation unit 132 in step S52-1 (step S73-4). For example, the objective variable data extraction unit 133a extracts performance data that satisfies the objective variable prediction model data and corresponds to a period specified by the user from the objective variable performance data DB 25. Then, the objective variable data extraction unit 133a generates learning data by processing the performance data based on the objective variable prediction model data.

[0097] If it is determined in step S73-1 that there is no change in the performance data for the specific period, the objective variable control unit 133 may skip steps S73-2 and S73-3 and move the process to step S73-4.

[0098] The learning unit 133b executes a learning process to learn a second prediction model based on the learning data extracted in step S73-4 (step S73-5). For example, when multiple pieces of definition information (explanatory variable prediction model data) have been generated as in the example of Fig. 4, the learning unit 133b learns a second prediction model for each piece of definition information. For example, when four sets of definition information exist, the learning unit 133b learns four second prediction models.

[0099] Here, if the current dependent variable prediction process is an update of the second prediction model, the learning unit 133b makes predictions for the test period using the confirmed second prediction model (the second prediction model generated last time) and the second prediction model generated this time, and compares the accuracy of the prediction results between the two second prediction models (step S73-6).

[0100] Next, the prediction unit 133c extracts prediction data (step S73-7). For example, the prediction unit 133c may extract, as prediction data, data corresponding to an array of future dates (an array of management contents) for the prediction period from the end date when the learning of the second prediction model is completed.

[0101] In addition, if the current dependent variable prediction process is not an update of the second prediction model (i.e., if the current dependent variable prediction process is a new generation of the second prediction model), the learning unit 133b may skip step S73-6 and proceed to step S73-7.

[0102] The prediction unit 133c predicts the dependent variable by providing the prediction data to the second prediction model and registers a predicted value indicating the prediction result (step S73-8). For example, if the current dependent variable prediction process is an update of the second prediction model, the prediction unit 133c makes the prediction using the second prediction model with higher accuracy based on the comparison result of step S73-6. Furthermore, if the current dependent variable prediction process is a new generation of the second prediction model, the prediction unit 133c makes the prediction using the newly generated second prediction model. Furthermore, the server device 100 may transmit a predicted value indicating the prediction result to the user's terminal device.

[0103] [6. Summary] So far, specific examples of information processing according to the embodiment have been described. According to the information processing according to the embodiment, highly accurate prediction can be performed in the minimum unit desired by the user. According to the information processing according to the embodiment, the operational load related to prediction can be reduced.

[0104] [7. Hardware Configuration] Next, an example of the hardware configuration of an information processing device (server device 100) according to an embodiment will be described. FIG. 8 is a block diagram showing an example of the hardware configuration of the information processing device according to an embodiment. Referring to FIG. 8, the information processing device includes, for example, a processor 801, a ROM 802, a RAM 803, a host bus 804, a bridge 805, an external bus 806, an interface 807, an input device 808, an output device 809, a storage 810, a drive 811, a connection port 812, and a communication device 813. Note that the hardware configuration shown here is an example, and some of the components may be omitted. Furthermore, the information processing device may further include components other than those shown here.

[0105] (Processor 801) The processor 801 functions, for example, as an arithmetic processing device or control device, and controls the overall operation of each component or part of it based on various programs recorded in the ROM 802, RAM 803, storage 810, or removable recording medium 901.

[0106] (ROM802, RAM803) The ROM 802 is a means for storing programs to be read into the processor 801, data to be used for calculations, etc. The RAM 803 temporarily or permanently stores, for example, the programs to be read into the processor 801, various parameters that change as appropriate when the programs are executed, etc.

[0107] (host bus 804, bridge 805, external bus 806, interface 807) The processor 801, ROM 802, and RAM 803 are interconnected via, for example, a host bus 804 capable of high-speed data transmission. On the other hand, the host bus 804 is connected to an external bus 806, which has a relatively low data transmission speed, via, for example, a bridge 805. Furthermore, the external bus 806 is connected to various components via an interface 807.

[0108] (input device 808) Examples of the input device 808 include a mouse, keyboard, touch panel, button, switch, and lever. Furthermore, a remote controller (hereinafter referred to as a remote control) capable of transmitting control signals using infrared rays or other radio waves may also be used as the input device 808. The input device 808 also includes an audio input device such as a microphone.

[0109] (output device 809) The output device 809 is a device capable of visually or audibly notifying the user of acquired information, such as a display device such as a CRT (Cathode Ray Tube), LCD, or organic EL display, an audio output device such as a speaker or headphones, a printer, a mobile phone, or a facsimile. The output device 809 according to this embodiment also includes various vibration devices capable of outputting tactile stimuli. The output device 809 may also include an AI speaker or a wearable device.

[0110] (Storage 810) The storage 810 is a device for storing various types of data. For example, a magnetic storage device such as a hard disk drive (HDD), a semiconductor storage device, an optical storage device, or a magneto-optical storage device may be used as the storage 810.

[0111] (Drive 811) The drive 811 is a device that reads information recorded on a removable recording medium 901 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, or writes information to the removable recording medium 901 .

[0112] (Connection port 812) The connection port 812 is a port for connecting an external device 902, such as a Universal Serial Bus (USB) port, an IEEE1394 port, a Small Computer System Interface (SCSI), an RS-232C port, or an optical audio terminal.

[0113] (Communication device 813) The communication device 813 is a communication device for connecting to a network, such as a communication card for wired or wireless LAN, Bluetooth (registered trademark), or WUSB (Wireless USB), a router for optical communication, a router for ADSL (Asymmetric Digital Subscriber Line), or a modem for various types of communication.

[0114] (Removable recording medium 901) The removable recording medium 901 is, for example, a DVD medium, a Blu-ray (registered trademark) medium, an HD DVD medium, various semiconductor storage media, etc. Of course, the removable recording medium 901 may also be, for example, an IC card equipped with a contactless IC chip, or an electronic device.

[0115] (External connection device 902) The externally connected device 902 is, for example, a printer, a portable music player, a digital camera, a digital video camera, or an IC recorder.

[0116] When the information processing device according to the embodiment is a server device 100, the data warehouse 120 is realized by a ROM 802, a RAM 803, and a storage 810. Furthermore, a control unit 130 realized by a processor 801 reads out and executes each control program (for example, an information processing program according to the embodiment) that realizes a reception unit 131, a model data generation unit 132, a response variable control unit 133, and an explanatory variable control unit 134 from the ROM 802, the RAM 803, etc.

[0117] [8. Other] Of the above processes, all or part of the processes described as being performed automatically may be performed manually. Furthermore, all or part of the processes described as being performed manually may be performed automatically using known methods. Furthermore, the information, including the processing procedures, specific names, various data, and parameters shown in the above documents and drawings, may be changed as desired unless otherwise specified. For example, the various information shown in each drawing is not limited to the information shown.

[0118] Furthermore, the components of each device shown in the figure are functional concepts and do not necessarily have to be physically configured as shown. In other words, the specific form of distribution and integration of each device is not limited to that shown. Furthermore, all or part of each component may be functionally or physically distributed and integrated in any unit depending on various loads, usage conditions, etc. Furthermore, each of the processes described above may be executed in appropriate combinations within a range that does not contradict each other.

[0119] The above describes in detail the embodiments of the present application based on several drawings, but these are merely examples, and the present invention can be implemented in other forms that include the embodiments described in the Disclosure of the Invention section and that have been modified and improved in various ways based on the knowledge of those skilled in the art. [Explanation of symbols]

[0120] 100 Server device 130 Control Unit 131 Reception 132 Model data generation unit 133 Objective variable control section 133a Objective variable data extraction section 133b Learning Department 133c Prediction Department 134 Explanatory variable control section 134a Explanatory variable data extraction section 134b Learning Department 134c Prediction Section

Claims

1. a reception unit that receives a selection of a prediction unit, which is a unit for predicting a dependent variable, from among management items having a hierarchical relationship, and receives from a user a selection of a prediction target that is a target of the dependent variable from among management contents linked to each of the management items; A generation unit that generates definition information that defines the content of the prediction target based on the management item selected as the prediction unit and the management content selected as the prediction target; a learning unit that uses learning data extracted based on the definition information to learn a model that predicts the dependent variable in units indicated by the definition information; An information processing device comprising:

2. the reception unit receives a selection of the prediction unit from the management items that are classifications in a business management database of an organization to which the user belongs, and also receives a selection of the prediction target from management contents linked to each of the management items in the database; The generation unit searches the database for a combination of management contents that satisfies both the management item selected as the prediction unit and the management contents selected as the prediction target, and generates the definition information for each combination of management contents found.

2. The information processing apparatus according to claim 1, wherein:

3. The generation unit acquires, as management contents that are elements of the definition information, a combination of management contents that are linked to the management items selected as the prediction unit, among the management contents linked to each of the management items, based on the arrangement of the management contents according to the axis of the hierarchical relationship and the management contents selected as the prediction target.

3. The information processing apparatus according to claim 2, wherein:

4. The generation unit acquires, as elements of the definition information, a combination of management contents associated with a management item selected as the prediction unit from among management contents constituting an array of management contents according to an axis of the hierarchical relationship, the array including the management contents selected as the prediction target.

4. The information processing apparatus according to claim 3,

5. The generating unit generates, as the definition information, data that combines elements of the definition information and the management content selected as the prediction target.

4. The information processing apparatus according to claim 3,

6. The learning unit extracts, as the learning data, performance data that satisfies the definition information from performance data that is related to the definition information and corresponds to a period designated by a user.

2. The information processing apparatus according to claim 1, wherein:

7. the generating unit generates the definition information for each combination of management contents acquired as elements of the definition information; The learning unit learns a model for predicting the dependent variable in units indicated by the definition information for each of the definition information.

2. The information processing apparatus according to claim 1, wherein:

8. An information processing method executed by an information processing device, a receiving step of receiving a selection of a prediction unit, which is a unit for predicting a dependent variable, from among management items having a hierarchical relationship, and receiving from a user a selection of a prediction target that is a target of the dependent variable from among management contents linked to each of the management items; a generating step of generating definition information that defines the content of the prediction target based on the management item selected as the prediction unit and the management content selected as the prediction target; a learning step of learning a model that predicts the dependent variable in units indicated by the definition information, using learning data extracted based on the definition information; An information processing method comprising:

9. a receiving step of receiving a selection of a prediction unit, which is a unit for predicting a dependent variable, from among management items having a hierarchical relationship, and receiving from a user a selection of a prediction target that is a target of the dependent variable from among management contents linked to each of the management items; a generation step of generating definition information that defines the content of the prediction target based on the management item selected as the prediction unit and the management content selected as the prediction target; a learning procedure for learning a model that predicts the objective variable in units indicated by the definition information, using learning data extracted based on the definition information; An information processing program for causing an information processing device to execute the above.

Images