Information processing system, information processing method, and program

Abstract

The present invention aims to provide a mechanism for improving the accuracy of demand forecasting. [Solution] The present invention is characterized by comprising a selection means that uses time series data related to the object to be predicted and a prediction model for calculating future predicted values ​​to calculate the error between the predicted value based on the prediction model and the time series data, and selects multiple prediction models based on the error, and a calculation means that uses the multiple prediction models to calculate future predicted values ​​related to the object to be predicted.

Description

[Technical Field]

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

[0002] Improving forecast accuracy is one of the major challenges in demand forecasting systems, and improving accuracy can bring about various benefits such as reduced inventory costs, lower stockout risks, and reduced logistics costs. In recent years, as the shortage of logistics resources has become a societal problem, improving forecast accuracy has become increasingly important.

[0003] For example, Patent Document 1 discloses a model acquisition means for acquiring a single item prediction model for predicting demand for a target product and a category prediction model for predicting demand for a category including the target product, a calculation means for calculating the accuracy of the demand prediction for the target product in the single item prediction model and the category prediction model, and an adoption means for adopting either the single item prediction model or the category prediction model based on the accuracy. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2024-77041 DISCLOSURE OF THE INVENTION [Problem to be solved by the invention]

[0005] However, the method of Patent Document 1 only selects either a single-item forecasting model or a category forecasting model based on information about the accuracy of demand forecasting, and is unable to make predictions using multiple models. In practice, it is desirable to evaluate the forecast accuracy of various models based on product characteristics and calculate forecast values ​​with high accuracy. Therefore, an object of the present invention is to provide a mechanism for improving the prediction accuracy in demand forecasting. [Means for solving the problem]

[0006] The present invention is characterized by comprising a selection means that uses time series data related to a prediction target and a prediction model for calculating a future predicted value to calculate an error between the predicted value based on the prediction model and the time series data, and selects a plurality of the prediction models based on the error, and a calculation means that uses the plurality of prediction models to calculate a future predicted value related to the prediction target. [Effects of the Invention]

[0007] According to the present invention, it is possible to provide a mechanism for improving the prediction accuracy in demand forecasting. [Brief explanation of the drawings]

[0008] [Figure 1] 1 is a diagram illustrating an example of a configuration of a supply and demand management system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a diagram illustrating an example of the hardware configuration of a client terminal 110 and a server 130 according to an embodiment of the present invention. [Figure 3] FIG. 1 is a diagram illustrating an example of an outline of processing according to an embodiment of the present invention. [Figure 4] 4A to 4C are diagrams illustrating an example of an AI-DFC model master 400 and an AI-DFC model attribute definition master 410 according to an embodiment of the present invention. [Figure 5] 5A to 5C are diagrams illustrating an example of an AI-DFC statistical method calculation definition master 500 and an AI-DFC model candidate master 510 according to an embodiment of the present invention. [Figure 6] FIG. 6 is a diagram showing an example of an item site master 600 in the embodiment of the present invention. [Figure 7] FIG. 10 is a diagram illustrating an example of a flowchart of a mixture ratio calculation process executed by an information processing device according to an embodiment of the present invention. [Figure 8] FIG. 8 is a diagram showing an example of AI-DFC model mixture ratio data 800 in the embodiment of the present invention. [Figure 9] FIG. 10 is a diagram illustrating an example of a flowchart of a learning data output process executed by an information processing device according to an embodiment of the present invention. [Figure 10] FIG. 1 is a diagram showing an example of learning data 1000 in an embodiment of the present invention. [Figure 11] FIG. 10 is a diagram illustrating an example of a flowchart of a prediction result calculation process executed by an information processing device according to an embodiment of the present invention. [Figure 12] FIG. 10 is a diagram illustrating an example of a prediction result calculation process executed by an information processing device according to an embodiment of the present invention. [Figure 13] FIG. 13 is a diagram showing an example of shipping performance data 1300 in the embodiment of the present invention. [Figure 14] FIG. 14 is a diagram showing an example of a demand characteristic classification master 1400 according to the embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0009] (First embodiment) FIG. 1 is a diagram showing an example of the system configuration of a supply and demand management system 100 according to an embodiment of the present invention.

[0010] A client terminal 110 and a server 130 are connected via a network 120 .

[0011] The client terminal 110 displays the screen sent from the server 130, accepts data input from the user, and sends it to the server 130.

[0012] The server 130 receives an execution request from the client terminal 110 , executes the processing, generates a screen displaying the results, and transmits it to the client terminal 101 .

[0013] The network 120 may be the Internet or a LAN (Local Area Network).

[0014] FIG. 2 is a block diagram showing an example of the hardware configuration of an information processing device that can be used as the server 130 or the client terminal 110 of the present invention.

[0015] As shown in FIG. 2, the information processing device is connected to a CPU (CENTRAL PROCESSING UNIT) 201, a RAM (RANDOM ACCESS MEMORY) 202, a ROM (READ ONLY MEMORY) 203, an input controller 205, a video controller 206, a memory controller 207, and a communication I / F controller 208 via a system bus 204.

[0016] The CPU 201 comprehensively controls each device and controller connected to the system bus 204 .

[0017] The RAM 202 functions as a main memory, a work area, etc. for the CPU 201. The CPU 201 loads programs and the like required for executing processing from the ROM 203 or the external memory 211 into the RAM 202, and executes the loaded programs to realize various operations.

[0018] ROM 203 or external memory 211 stores the BIOS (BASIC INPUT / OUTPUT SYSTEM) and OS (OPERATING SYSTEM), which are control programs executed by CPU 201, computer-readable and executable programs for realizing this information processing method, and various necessary data (including data tables).

[0019] The input controller 205 controls input from input devices such as a keyboard 209 and a pointing device such as a mouse (not shown). If the input device is a touch panel, the user can issue various instructions by pressing (touching with a finger or the like) icons, cursors, or buttons displayed on the touch panel.

[0020] The touch panel may also be a touch panel capable of detecting positions touched by multiple fingers, such as a multi-touch screen.

[0021] The video controller 206 controls the display on an external output device such as a display 210. The display also includes the display of a notebook computer integrated into the main body. Note that the external output device is not limited to a display, and may be, for example, a projector. In addition, for devices capable of receiving the above-mentioned touch operation, an input device is also provided.

[0022] The video controller 206 can control a video memory (VRAM) for display control, and can use part of the RAM 202 as a video memory area, or can provide a separate dedicated video memory.

[0023] The memory controller 207 controls access to the external memory 211. The external memory may be an external storage device (hard disk) that stores a boot program, various applications, font data, user files, edited files, and various data, a flexible disk (FD), or a CompactFlash (registered trademark) memory connected to a PCMCIA card slot via an adapter.

[0024] The communication I / F controller 208 connects and communicates with external devices via a network, and executes communication control processing on the network. For example, communication using TCP / IP, telephone lines such as ISDN, and 3G lines for mobile phones are possible.

[0025] The CPU 201 enables display on the display 210 by, for example, executing a process of expanding (rasterizing) an outline font into a display information area in the RAM 202. The CPU 201 also enables user instructions using a mouse cursor (not shown) or the like on the display 210.

[0026] An example of an outline of processing in an embodiment of the present invention is shown in Figure 3. The processing flow will be explained by dividing it into two stages: (1) a learning stage and (2) a prediction stage.

[0027] (1) Learning stages In the learning stage, the (a) attribute information of the product and (b) demand characteristics from past demand data are calculated and used as explanatory variables.Furthermore, the (d) optimal time series forecasting model data selected by the (c) optimal time series forecasting model selection process (described later) is used as the objective variable to generate a trained model.

[0028] First, the definition of (a) attribute information set for each model in the model attribute definition master 410 is obtained. Based on the defined attribute information, attribute information is obtained from a product database or the like. Any attribute information is acceptable, and may include, for example, product classification, characteristics, sales region, sales period, etc. Multiple attributes may be set for each model, and for example, attributes such as product classification and warehouse location may be set for each SKU (stock keeping unit).

[0029] Next, the demand characteristics are calculated from the shipment performance data 1300 (described later in FIG. 13) to obtain (b) demand characteristic data. The demand characteristics are statistical values ​​calculated from the shipment performance data 1300. Therefore, the demand characteristic calculation corresponds to the statistical quantity calculation process (step S908 in FIG. 9). The statistical quantity index can be set for each model in the statistical calculation method definition master 500; for example, the average value, standard deviation, correlation coefficient, etc. are used, and the value of each index during the evaluation period of the shipment performance data is calculated as (b) demand characteristic data.

[0030] Furthermore, using the actual shipment data 1300 and the evaluation index set in the model master table 400 (described later in FIG. 4), the optimal time series prediction model is selected from the models set in the model candidate master 510 (described later in FIG. 5). This process is called (c) optimal time series prediction model selection process, and is a mechanism for calculating the prediction errors over the evaluation period of multiple prediction models based on past actual demand data, and selecting the model with the smallest prediction error. Through the above process, (d) optimal time series prediction model data for each actual shipment data is obtained.

[0031] In a second embodiment described later, the prediction model to be used for predicting future predicted values ​​is determined using the prediction error calculated in (c) the optimum time series prediction model selection process.

[0032] In the first embodiment, the above-mentioned (a) attribute information, (b) demand characteristics, and (d) optimal time series prediction model data are used as training data 1000 (described later in FIG. 10) to generate a trained model. This completes the processing in the training stage.

[0033] (2) Prediction stage In the prediction stage, the demand characteristics calculated from the actual shipment data of the product to be predicted and the product attribute information are input into the trained model generated in (1), thereby calculating candidates for the time series prediction model and the mixture ratio (weight) for each candidate.

[0034] Each statistical value set in the AI-DFC model statistical calculation method definition master (described later in FIG. 5) is calculated from the actual shipment data of the forecast target, and used as demand characteristic data. Furthermore, for the attribute information items set for each model in the model attribute definition master 410, attribute information is obtained from a database of the product to be forecasted, etc.

[0035] By inputting demand characteristic information and attribute information into the trained model, candidate time series prediction models and the blending ratios of each time series prediction model are output. From the candidate models output to the trained model, a model to be used in the prediction value calculation process is selected from the candidate prediction models using the prediction model selection method set in the AI-DFC model master table 400 (described later in FIG. 12). Furthermore, the total blending ratio of the selected models to be used in the prediction value calculation process is adjusted to 1, and output as AI-DFC model blending ratio data 800 (described later in FIG. 8). The processing up to this point is the blending ratio calculation process, the details of which will be described later using FIG. 7.

[0036] The forecast calculation process is performed based on the model used for the forecast calculation process included in the AI-DFC model mix ratio data, the mix ratio of that model, and past shipping performance data. The forecast value is the future demand volume of the product being predicted.

[0037] That is, this is an example of a process for calculating future demand for a product related to a prediction target using multiple prediction models.

[0038] This process corresponds to the predicted value calculation process described in Fig. 11. The above is the process at the prediction stage.

[0039] First, an example of a master used in the mixture ratio calculation process (FIG. 7), the learning data output process (FIG. 9), and the predicted value calculation process (FIG. 11) will be described with reference to FIGS.

[0040] The AI-DFC model master shown in Figure 4(a) is a master that sets the selection method for selecting a model to be used to calculate a predicted value from candidate models set for a group (AI-DFC model ID) that includes each model.

[0041] The AI-DFC model master 400 includes the following items: AI-DFC model ID, forecast model selection category, forecast model upper limit number / cumulative ratio, start index, end index, evaluation period, and forecast error ID.

[0042] The AI-DFC model ID is an ID arbitrarily assigned to a group that includes each model candidate. The AI-DFC model ID may be assigned according to product characteristics (regular demand products, seasonal products, etc.), or may be assigned for each product group or type. Alternatively, the same ID may be assigned to all products.

[0043] For example, by setting an AI-DFC model ID for each of the above-mentioned product characteristics, it is possible to set models that are likely to be selected for each characteristic, such as a product with regular demand or a product with seasonal demand, as model candidates.

[0044] The prediction model selection category can specify the method for selecting candidate models, and specific numerical criteria can be set in the item "Maximum number of prediction models / Cumulative ratio." For example, if the prediction model selection category is "Cumulative ratio" and the cumulative ratio is set to 80% (=0.8), when selecting a model in the blending ratio calculation process of Figure 7, models whose cumulative ratio value, calculated by adding up the blending ratios from the highest model to the lowest, is 0.8 or less will be selected.

[0045] That is, this is an example of a process of selecting a prediction model to be used for calculating a predicted value by the calculation means, based on a condition based on the weight of the prediction model.

[0046] This is also an example of a process of selecting a prediction model from a plurality of prediction models in descending order of weight until the total weight satisfies a predetermined condition.

[0047] Furthermore, for example, if the prediction model selection category is "top designated number" and the upper limit number of prediction models is set to 2, the two models with the highest mixing ratios are selected in order.

[0048] That is, this is an example of a process of selecting a predetermined number of prediction models from among a plurality of prediction models in descending order of weight.

[0049] Alternatively, for example, a model with a blending ratio equal to or greater than a certain value may be selected. In this case, for example, the prediction model selection category can be set to "blend ratio" with a value of 0.8, allowing models with a blending ratio of 0.8 or greater to be selected.

[0050] That is, this is an example of a process of selecting a prediction model whose weight satisfies a predetermined condition as a prediction model to be used for calculating a predicted value.

[0051] Alternatively, a model with a mixing ratio equal to or greater than P% of the highest mixing ratio value may be selected. For example, if P=20% and the highest mixing ratio value is 0.8, other models with a mixing ratio of 0.16 or greater will be selected.

[0052] That is, this is an example of a process of selecting, from among a plurality of prediction models, a prediction model that satisfies a predetermined condition based on the weight of the prediction model with the largest weight, as the prediction model to be used in calculating a predicted value.

[0053] In this way, by allowing users to set the method of selecting a prediction model and the numerical criteria, it is possible to set a model selection method that is appropriate for each model or product group, which is expected to expand the scope of trial and error and improve prediction accuracy.

[0054] The start index and end index can be set to the cycle (period) to be predicted. For example, if the start index and end index are set to 1, only the cycle in which the prediction is performed will be the target of prediction. For example, if the start index and end index are set to 1 and 3, the period from the cycle in which the prediction is performed to two cycles later will be the target of prediction.

[0055] The evaluation period can be set to a period used for evaluating the model. In this embodiment, it is set in units of months, but it may also be set in units of weeks or days.

[0056] The forecast error ID can specify the evaluation index used to evaluate the model. For example, the root mean square error (RMSE) or the mean absolute error (MAE) can be set. The evaluation index set here is used to perform the optimal time series forecast model selection process shown in Figure 1(c). Any index that can evaluate the error can be used as the evaluation index, and other indexes such as the mean percent error (MPE) and the mean absolute error (MAPE) can also be set.

[0057] Fig. 4(b) shows an example of the AI-DFC model attribute definition master 410. The AI-DFC model attribute definition master shown in Fig. 4(b) is a master for setting model attributes for each model.

[0058] The AI-DFC model attribute definition master includes items for AI-DFC data category and AI-DFC attribute name.

[0059] The AI-DFC data classification allows each model set for an AI-DFC model ID to be registered with a distinction as to whether it is for production or training. This allows for model adjustments to be verified using a model registered for training verification, and models that prove to be accurate as a result of training verification can be re-registered as production models, facilitating trial and error.

[0060] The AI-DFC attribute name can be set to the name of the attribute used in each model. Attributes that can be set include, for example, product attributes and the region in which the product is sold. For example, if the user determines that the model selected may change depending on the region in which the product is sold, including region as an attribute allows the trained model to take regional characteristics into account when selecting a model.

[0061] Fig. 5(a) shows an example of the AI-DFC statistical calculation method definition master 500. The AI-DFC statistical calculation method definition master shown in Fig. 5(a) is a data table in which the statistical calculation method for each model can be registered, and includes items such as AI-DFC model ID, AI-DFC data category, statistical ID, and calculation start and end index.

[0062] The statistical ID allows you to select one or more statistical calculation methods for the statistical value for each model: mean, standard deviation, or correlation coefficient. For example, if mean is selected, the mean of the actual demand for the target period will be used as the statistical value, and if standard deviation is selected, the standard deviation of the actual demand for the target period will be used as the statistical value. This allows the user to set statistical indicators that can appropriately express the characteristics of time series data.

[0063] The calculation start index and calculation end index fields allow you to specify the start and end cycles for the period for which statistics are calculated. The average, standard deviation, and correlation coefficient for the period specified in these fields are calculated as statistical values.

[0064] Fig. 5(b) shows an example of the AI-DFC model candidate master 510. The AI-DFC model candidate master shown in Fig. 5(b) can set candidate model parameters for each model, and includes items such as an AI-DFC model ID, an AI-DFC data category, a prediction model parameter number, and a prediction model parameter name.

[0065] The forecast model parameter number is a number arbitrarily assigned to each candidate model. The forecast model parameter name is the specific model parameter name of each candidate model. Various forecasting and estimation methods can be set as model parameters. For example, forecasting methods such as exponential smoothing, Winters' method, and moving average method can be set as model candidates. In addition, autoregressive models and the like may also be used, and any model that can be used for forecasting can be set.

[0066] 6 shows an example of an item site master 600. The item site master allows you to set a forecasting method for each item (product or SKU), and includes fields for item site, forecast calculation method category, AI-DFC model ID, validation AI-DFC model ID, and specified forecast model parameter name.

[0067] The item site is a number assigned to each product or product group, and can be set in combination with the item (product) and site (location) numbers.

[0068] The forecast calculation method can be set for each product. For example, if "AI-DFC Mix" is selected, the future forecast value is calculated using the predicted value using the prediction model candidates output by the AI ​​and the mixture ratio of each model (details will be explained later in Figure 12).

[0069] If "designated model" is specified, the user can set the model to be used for prediction in the item of designated prediction model parameters, and prediction using the designated model becomes possible.

[0070] That is, this is an example of a process for calculating future predicted values ​​related to a prediction target using a prediction model specified by a user.

[0071] As mentioned above, it is possible to set whether to use a model selected using AI output for each product, or to use a model specified by the user, which allows users to customize and set a highly accurate method for each product's characteristics.

[0072] FIG. 7 is a diagram showing an example of a flowchart of processing executed by the information processing device according to the embodiment of the present invention.

[0073] In step S701, the item site master 600 shown in Fig. 6 is obtained. From the item site master 600, information on the AI-DFC model ID set for the product to be predicted is obtained.

[0074] In step S702, information on the pre-set AI-DFC model master 400 shown in Fig. 4(a) is acquired. Based on the settings of the prediction model selection category and the upper limit number of prediction models / cumulative ratio in the AI-DFC model master, a model to be used in the prediction value calculation process described later in Fig. 11 is selected from the candidate models output by the trained model.

[0075] That is, step S702 is a step showing an example of a process of accepting a method for selecting a prediction model to be used for calculating a predicted value by the calculation means from among a plurality of prediction models. In step S703, information is acquired from the pre-set AI-DFC statistical calculation method definition master 500 shown in Fig. 5(a). In the AI-DFC statistical calculation method definition master, the statistical calculation method (average, standard deviation, correlation coefficient, etc.) set for each model is acquired, and in step S708, statistical calculation processing is performed based on the shipping performance data using the acquired statistical calculation method.

[0076] In step S704, information is acquired from the pre-set AI-DFC model candidate master 510 shown in Fig. 5(b). In the AI-DFC model candidate master, model candidates (prediction model parameters) set for each AI-DFC model ID are acquired.

[0077] In step S705, information is acquired from the pre-set AI-DFC model attribute definition master 410 shown in Fig. 4(b). Based on the attribute information defined for each AI-DFC model ID in the AI-DFC model attribute definition master, attribute information for each product is acquired.

[0078] The attribute information is, for example, "item numeric string attribute 1" and "site string attribute 1," and information about the product and location is obtained by referencing a product database (not shown) to determine the characteristics and values ​​of each attribute for each product. For example, if "item numeric string attribute 1" is information about the product group number (for electrical appliances, this could be copiers, home appliances, audio equipment, etc.), a numeric value related to the product's group number is obtained from the product database. For example, if "site string attribute 1" is information about the product's sales region, a string related to the product's sales region (such as "KANTO" or "KANSAI") is obtained from the product database.

[0079] The order of the processes in steps S701 to S705 described above may be reversed, or the information of each master may be acquired all at once.

[0080] In step S706, the process of steps S706 and subsequent steps, which will be described later, is repeated for each AI-DFC model ID.

[0081] In step S707, actual shipping data for the product to be predicted is acquired. FIG. 13 shows an example of actual shipping data 1300. The actual shipping data is time-series data that summarizes information on actual shipping for each item site cycle. In this embodiment, the data is aggregated for each specified month, but it may also be aggregated for each specified period such as week, day, or quarter.

[0082] In step S708, a calculation process is performed for the shipping result statistics based on the shipping result data of the product to be forecasted acquired in step S707 and the information in the AI-DFC statistics calculation method definition master 500 acquired in step S703. For example, if "average" is set as the statistic, the average of the demand results for the target period is calculated as the statistic value.

[0083] In step S709, a mixing ratio calculation process is executed. In the mixing ratio calculation process, demand characteristic information, which is a statistical value of the demand characteristic of each product and each group calculated from the shipping performance data calculated in step S708, and attribute information of each product and each group acquired in step S705 are given to a trained model that has been trained in advance using the training data shown in Fig. 10, to output model candidates and their mixing ratios.

[0084] The model is selected based on the candidate models and their mixture ratios output by the trained model, and the information on the prediction model selection category selected in the AI-DFC model master acquired in step S702. Details will be described later with reference to FIG.

[0085] That is, step S709 is a step that shows an example of a process of inputting time series data related to the prediction target into a trained model that has been trained using time series data and a prediction model for calculating future predicted values, thereby obtaining multiple prediction models for calculating future predicted values ​​related to the prediction target.

[0086] In step S710, the repeated process performed for each AI-DFC model ID is terminated.

[0087] In step S711, the mixture ratio of the model selected in step S709 is output, and the AI-DFC model mixture ratio data 800 shown in FIG. 8 is output, and this process ends.

[0088] Figure 8 shows an example of AI-DFC model mixing ratio data 800. The AI-DFC model mixing ratio data includes the following fields: item site, monthly cycle number, forecast model parameter number, forecast model parameter name, and AI-DFC mixing ratio. This data is output in step S711, and for each cycle to be predicted for each item site, the forecast model actually used to calculate future forecast values ​​is selected from the candidate forecast models output by the trained model based on the selection method set in the AI-DFC model master table in Figure 4(a).

[0089] The monthly cycle number is a number that identifies the cycle in which a prediction was made using the trained model. The forecast model parameter number and forecast model parameter name are information about the model that was output as the forecast model actually used to calculate future forecast values ​​from the candidate forecast models output by the trained model.

[0090] For example, for the item site "C02431-A100," the models "M1 exponential smoothing with 36-month fluctuations_estimate" and "M1 Winters model with 36-month fluctuations_estimate (no seasonal consideration)" are selected, and the respective blending ratios are output as "0.286" and "0.714." For each cycle to be forecast for each item site, the forecast model used to calculate future forecast values ​​and the probability that that model will be selected are output as the "blend ratio."

[0091] An example of a flowchart of the learning data output process executed by the information processing device is shown in Fig. 9. The learning data output process described in Fig. 9 is a process for outputting learning data 1000 shown in Fig. 10.

[0092] In step S901, the item site master 600 shown in Fig. 6 is obtained. From the item site master 600, information on the AI-DFC model ID set for the product to be learned is obtained.

[0093] In step S902, the shipping performance data of the product to be used for learning is acquired.

[0094] In step S903, the information of the AI-DFC model master 400 shown in Fig. 4(a) is acquired. In step S910, the prediction error calculation process (described later) is performed using the information of the start index, end index, evaluation period, and prediction error ID set in the AI-DFC model master.

[0095] In step S904, information is acquired from the preset AI-DFC model candidate master 510 shown in Fig. 5(b). In the AI-DFC model candidate master, model candidates set for each AI-DFC model ID are acquired.

[0096] In step S905, information is acquired from the pre-set AI-DFC model attribute definition master 410 shown in Fig. 4(b). Based on the attribute information defined for each model in the AI-DFC model attribute definition master, attribute information for each product is acquired.

[0097] In step S906, information is acquired from the pre-set AI-DFC statistical calculation method definition master 500 shown in Fig. 5(a). In the AI-DFC statistical calculation method definition master, the statistical calculation method (average, standard deviation, correlation coefficient) set for each model is acquired, and in step S908, statistical calculation processing is performed based on the shipping performance data.

[0098] The order of the processes in steps S901 to S906 described above may be reversed, or the master information may be acquired all at once.

[0099] In step S907, an optimum time series prediction model selection process is started for each cycle, in which a prediction model with the smallest prediction error is selected from the prediction model candidates.

[0100] In step S908, a statistical value calculation process is performed. The statistical values ​​used are the average, standard deviation, correlation coefficient, etc. set for each model in the AI-DFC statistical calculation method definition master acquired in step S906.

[0101] In step S909, the forecast value of each forecast model is calculated using the forecast model candidates set in the AI-DFC model candidate master. Specifically, by using the forecast model (e.g., exponential smoothing, Winters method, moving average method, etc.) set in the forecast model parameter name field and past actual demand data, it is possible to calculate past forecast values ​​when applying that model to forecasts for the target product. A known method may be used to calculate forecast values ​​from the forecast model and past actual demand.

[0102] In step S910, the prediction error between the prediction results of each prediction model calculated in step S909 and past demand actual data is calculated based on the evaluation period and prediction error ID set in the AI-DFC model master. The prediction model parameters of the model with the smallest prediction error are obtained, and that model is determined as the optimal time series prediction model.

[0103] In this embodiment, the model with the smallest prediction error is selected as the optimal model, but multiple models may be selected and included in the training data based on the prediction error results.

[0104] Alternatively, in step S910, multiple models may be selected based on the prediction errors, and the predicted values ​​of the multiple models and a value based on the prediction errors may be used to calculate future predicted values ​​without using a trained model. This method will be described later as a second embodiment.

[0105] In step S911, the process of selecting the optimal time series prediction model for each cycle ends.

[0106] In step S912, the prediction model parameter number of the model with the smallest prediction error determined for each cycle is obtained, and the learning data 1000 shown in FIG. 10 is output, and the process ends.

[0107] 10 shows an example of learning data 1000. The learning data includes, for example, a demand structure number, a monthly cycle number, a forecast model parameter number, attribute information (item string attribute, site string attribute, item numeric string attribute, site numeric string attribute), statistical values ​​(average, standard deviation, correlation coefficient), etc.

[0108] The demand structure number is a number that uniquely identifies an SKU, and may be a code assigned to each product or warehouse, for example. The monthly cycle number is a number that identifies the cycle for which the prediction was made.

[0109] The forecast model parameter number is the model selected in the optimal time series forecast model selection process in step S910 for the forecast target cycle for that demand structure number.

[0110] The attribute information can be set to item string attributes, site string attributes, item numeric string attributes, and site numeric string attributes, such as product attributes and the region where the product is sold. The attribute information to be included in the learning data can be set by the user in the AI-DFC model attribute definition master 410. The attribute information is acquired in step S905 and output as learning data.

[0111] In this embodiment, the attribute information is set by classifying it into items, sites, character strings, and numerical values, but this is not limitative and it may be registered without classification.

[0112] The statistical values ​​include the items of average, standard deviation, and correlation coefficient. The statistical values ​​here are the average, standard deviation, and correlation coefficient calculated from monthly shipment performance data within a specified period up to the month before the current month (the month in which the prediction was made). In this embodiment, calculations are performed monthly, but calculations may also be performed weekly or daily. The statistical information is calculated in step S908 and output as learning data.

[0113] In this embodiment, training data such as that shown in Figure 9 is used to generate trained models for each product characteristic (regular demand product, seasonal product, etc.) and demand pattern, but this is not limited to this and other attributes may also be used for grouping.

[0114] FIG. 11 shows an example of a flowchart of a prediction result calculation process executed by an information processing device.

[0115] In step S1101, the AI-DFC model master shown in FIG. 4(a) is acquired, and the setting values ​​of the forecast model selection category and the forecast model upper limit number / cumulative ratio are acquired.

[0116] In step S1102, the AI-DFC model candidate master is acquired, and the parameter numbers and prediction model parameter names of the prediction models that are candidates for each model are acquired.

[0117] In step S1103, actual shipping data of the item site that is the target of the prediction is acquired.

[0118] In step S1104, the mixture ratio data is acquired, which allows the model selected by the trained model and the set selection method to be acquired.

[0119] In step S1105, a repeat process of calculating a predicted value for each prediction model is started.

[0120] In step S1106, the prediction model parameters selected for the target cycle of the prediction target item are obtained from the AI-DFC model mixture ratio data shown in Figure 7, and a calculation process for the prediction value is performed based on past demand actual data. This process is the same as the process in step S909.

[0121] In step S1107, the repeated process of calculating the predicted value for each prediction model ends.

[0122] In step S1108, a future predicted value is calculated using the predicted value for each model calculated in step S1106. The predicted value calculated in step S1106 is multiplied by the AI-DFC mixing ratio value of the AI-DFC model mixing ratio data 800 acquired in step S1104, and the sum of these products is used to calculate a final predicted value. Details of the calculation method will be described later with reference to FIG. 12.

[0123] That is, step S1108 is a step showing an example of a process of calculating a future predicted value related to the prediction target using the acquired multiple prediction models.

[0124] In other words, step S1108 is a step showing an example of a process of calculating a predicted value using a prediction model specified by the received selection method.

[0125] In step S1109, the prediction result is displayed, and the process ends.

[0126] 12 shows an example of a prediction result calculation process executed by an information processing device according to an embodiment of the present invention. First, (1) a method for calculating a prediction value using a candidate model and a mixture ratio output by a trained model will be described ((2) will be described in the second embodiment).

[0127] Figure 12(1) shows a specific example of a method for selecting a model to be used in the prediction result calculation process of Figure 11 from the candidate models output by the trained model in the mixing ratio calculation process of step S709, based on the prediction model selection method selected in the AI-DFC model master 400 obtained in step S702.

[0128] As shown in the table on the left, the output result of the trained model is the blending ratio value of each model candidate for each SKU. The trained model is trained using product demand characteristic information and attribute information as explanatory variables and the model selected in the optimal time series prediction model selection process as the objective variable. The trained model is then given demand characteristic information and attribute information of the product to be predicted, and the appropriate prediction model candidate and its blending ratio are output. In other words, the blending ratio can be thought of as the reliability of the model evaluated by the AI ​​or the probability that the model will be selected. The trained model may output the blending ratio of all model candidates set in the AI-DFC model candidate master 510, or it may select an appropriate candidate model from the model candidates and output the blending ratio of that model.

[0129] In other words, this is an example of a process in which the trained model outputs multiple prediction models from among pre-set prediction models for calculating future predicted values ​​for the prediction target based on time series data related to the prediction target.

[0130] From the output results of the trained models, a model to be used for calculating the predicted value is selected based on the selection method set in the AI-DFC model master 400, and output as AI-DFC model mixture ratio data 800.

[0131] For example, if the prediction model selection category is "cumulative ratio" and the cumulative ratio value is set to "0.9" in the AI-DFC model master 400, the model whose cumulative mixing ratio value is 0.9 or less will be selected from the AI ​​output results in the left table in descending order of mixing ratio. Since the mixing ratios of Model A and Model B output by the AI ​​are 0.60 and 0.20, respectively, the cumulative ratio value is 0.80, and Model A and Model B are selected.

[0132] The mixing ratio value output by the AI ​​can be used as is to calculate the predicted value, which will be described later, or it can be adjusted so that the mixing ratio of Model A and Model B becomes 1. When making adjustments, for example, it is possible to leave the ratio of the mixing ratio values ​​of Model A and Model B unchanged and adjust them so that the total becomes 1. When adjusting in this way, the mixing ratio values ​​of Model A and Model B after adjustment will be 0.75 and 0.25, respectively.

[0133] Equation (1) shows how to calculate future predicted values ​​in the above example. In the AI-DFC model mixing ratio data 800, if the mixing ratio for Model A is output as 0.75 and the mixing ratio for Model B is output as 0.25, the predicted values ​​when Model A and Model B are used are used to calculate the predicted value as shown in Equation (1). If the predicted values ​​when Model A and Model B are used are a and b, respectively, the final predicted value is calculated as 0.75a + 0.25b.

[0134] The future forecast value is the sum of the forecast value when the selected model is used and the value multiplied by the blending ratio of that model, for each selected model. In other words, the forecast values ​​for each selected prediction model are summed according to the weight ratio between the prediction models.

[0135] In other words, this is an example of a process that obtains weights for each of the multiple prediction models obtained, and calculates future prediction values ​​for the prediction target based on the prediction values ​​calculated using each of the obtained prediction models and the weights associated with the prediction models.

[0136] In other words, this is an example of a process for calculating a predicted value by summing the predicted values ​​of the prediction models selected as the prediction models to be used in calculating the predicted value in accordance with the ratio of the weights of the prediction models.

[0137] In this embodiment, the mixing ratios are output so that the total is 1, but % may be used so that the total is 100, or the maximum or minimum mixing ratio value X may be used as the reference, such as 0.1X or 2X. In other words, any expression method may be used as long as it expresses the likelihood of a candidate model or the probability that it can become a candidate.

[0138] By displaying the mixture ratio for each candidate model, the user can know the information that serves as the basis for which model was judged to be most likely and used to calculate the predicted value.

[0139] In addition, by evaluating the likelihood (probability of selection) of each model as a mixture ratio along with the candidate models from the trained model and using this value to calculate future prediction values, more likely models will have a significant impact on the prediction values, making it possible to make accurate predictions using multiple models.

[0140] The first embodiment has been described above. However, if you want to improve the prediction accuracy of AI, you can consider changing the attribute information and demand characteristics provided to the AI, as well as the prediction model selection method described above, through trial and error. For example, if models A and B are frequently selected for a certain product, and selecting model C also improves prediction accuracy, you can expect to improve accuracy by resetting the selection method to include model C. In other words, by providing various prediction model selection methods, it becomes possible to efficiently adjust and set the model selection method appropriate for each product.

[0141] 6, it is possible to set for each product or SKU whether to use a model selected using AI output or a model specified by the user. Here, it is also possible to display not only whether a prediction model was selected using AI, but also information identifying the person who specified the prediction method.

[0142] Specifically, when a user-specified model is set to be used for prediction, the system displays the name of the user who set the model, the user's attributes, etc. For example, user attributes include the product manager of the company that provides the demand forecasting system, or the customer who uses the demand forecasting system. By visualizing the user's attributes, it becomes possible to identify whether the model was specified by the product manager or the customer, which can be used as a reference when adjusting the model selection method. For example, when adjusting the model selection on the customer side, verification can be performed using the model set by the product manager as a reference.

[0143] Furthermore, when selecting a model using AI, information identifying the person who trained or retrained the AI ​​may be displayed. For example, in this embodiment, the training data 1000 is created and the username and user attributes of the person who provided the model are displayed. This allows a customer to distinguish a new trained model from models created by product personnel and verify it.

[0144] Furthermore, in product demand forecasting, forecast results may be displayed identifiably based on whether or not a forecast model was selected using AI. For example, on a screen that displays forecast results in chronological order, forecast results using AI may be displayed in a different color. Note that as long as the user can identify which selection method was used to generate the forecast results, a pop-up display or other method may also be used.

[0145] The above is an example of a process for distinguishably displaying whether a future predicted value related to a prediction target is a predicted value based on a prediction model acquired using a trained model or a predicted value based on a prediction model specified by a user.

[0146] (Second embodiment) In the first embodiment, a method for calculating a predicted value from the output result of a trained model was shown, whereas in the second embodiment, a method for identifying a prediction model to be used for calculating a predicted result based on a calculated prediction error and calculating a future predicted value is shown.

[0147] 9, the forecast values ​​of the set candidate forecast models are calculated using past demand records, and the forecast error with respect to the forecast values ​​for the evaluation period is calculated to select the optimal model. In the target cycle of the product to be forecasted, the forecast model to be used for calculating the forecast values ​​is selected using the forecast values ​​and forecast error of the forecast model calculated in this optimal time series forecast model selection process, and future forecast values ​​are calculated in the second embodiment.

[0148] 14 shows an example of a demand characteristic classification master 1400 according to an embodiment of the present invention. The demand characteristic classification master 1400 includes items such as a demand characteristic classification ID, a demand characteristic classification name, a DFC forecast category, a forecast error, and an error range rate.

[0149] The demand characteristic classification ID is an arbitrary ID linked to the demand characteristic classification name. The demand characteristic classification name is an item that classifies demand characteristics according to product characteristics. For example, it includes classifications such as seasonal limited items and out-of-stock items. The demand characteristic classification may be automatically assigned based on past demand records, or may be manually set by the user for each item site or product.

[0150] In the DFC forecast category, the user can select either "DFC selection" or "DFC selection mix calculation." If DFC selection is selected, a forecast value is calculated using the forecast model with the smallest forecast error (optimal time series forecast model) selected from the results of the forecast errors calculated in the same manner as in step S912, and this value is used as the future forecast value.

[0151] If the DFC selection mix calculation is selected, the prediction error of the candidate models is calculated, a model is selected based on the prediction error, and future prediction values ​​are calculated based on the prediction value and prediction error of the selected model. Details will be described later using Figure 12.

[0152] In the forecast error section, you can specify the evaluation index used to evaluate the forecast model. For example, you can set the root mean square error (RMSE) or the mean absolute error (MAE). The evaluation index set here is used to evaluate the error between the forecast value and past demand results, and then the forecast model selection process is performed.

[0153] The error margin rate is an item that accepts the model selection method for calculating a predicted value based on the prediction error of candidate models when the DFC selection mix calculation is set. As will be described in detail later with reference to FIG. 12, by using the error margin rate as the selection method, it is possible to select a model that shows an error that falls within a range based on a reference value, using the prediction error value of the model with the smallest prediction error as the reference value. This makes it possible to select models with higher accuracy than simply selecting a specified number of models with small errors.

[0154] That is, this is an example of a process of selecting a prediction model that satisfies a predetermined condition based on the error of the prediction model with the smallest error as the prediction model to be used for calculating a predicted value.

[0155] 12(2) shows a method for calculating a predicted value using the results of the time series prediction model calculation process. That is, this is a method for calculating a predicted value using the results of calculating the prediction error of the prediction model performed in steps S909 to S912 in FIG.

[0156] Figure 12(2) shows an example of processing when "DFC selection mix calculation" is set in the "DFC forecast category" item in Figure 14. This shows an example of a method for calculating a forecast error from the forecast value of a candidate model in the forecast error calculation processing in step S910, and selecting a model to be used in the forecast result calculation processing based on the selection method set in the demand characteristic classification master 1400.

[0157] The forecast error output results show the forecast error values ​​calculated using actual demand data for each candidate model, as shown in the table on the left.

[0158] From the output results of the forecast error, a model to be used for calculating a future forecast value is selected based on the selection method set in the demand characteristic classification master 1400, and the forecast value is calculated. That is, this is an example of a calculation process in which time-series data related to the forecast target and a forecast model for calculating a future forecast value are used to calculate the error between the forecast value based on the forecast model and the time-series data, multiple forecast models are selected based on the error, and a future forecast value related to the forecast target is calculated using the multiple forecast models.

[0159] For example, if the "error margin rate" is set to "0.4" in the demand characteristic classification master 1400, model B, which has an error of 40% or less, is selected from the prediction error value of model A, which has the smallest prediction error in the prediction error output results in the left table. In other words, this is an example of a process for selecting multiple prediction models to be used in the calculation of a prediction value by the calculation means, based on a condition based on the magnitude of the value related to the error of the prediction model.

[0160] The prediction errors of Model A and Model B are 130 and 180, respectively, and the reciprocal of these values ​​is used to calculate the blending ratio of Model A and Model B. Since the smaller the prediction error, the better the prediction accuracy, the reciprocal of the prediction error is used as the blending ratio. In this case, it is also possible to adjust the blending ratio so that the sum of the blending ratios is 1 without changing the ratio of the reciprocals of the prediction errors of Model A and Model B. When adjusted in this way, the blending ratio values ​​of Model A and Model B after adjustment will be 0.58 and 0.42, respectively.

[0161] Equation (2) shows a method for calculating future predicted values ​​in the above example. If the blending ratio for model A is output as 0.58 and the blending ratio for model B is output as 0.42, the predicted values ​​when using model A and model B are used to calculate the predicted value as shown in equation (2). If the predicted values ​​when using model A and model B are a and b, respectively, the final predicted value is calculated as 0.58a + 0.42b. In other words, this is an example of a process for calculating future predicted values ​​for a prediction target based on predicted values ​​calculated using multiple prediction models and values ​​related to the errors of the prediction models.

[0162] In other words, this is an example of a process of receiving a selection method for a prediction model to be used to calculate a predicted value by a calculation means from among the prediction models for which an error has been calculated, and calculating a predicted value using the prediction model identified by the received selection method.

[0163] In other words, the future predicted value is the sum of the predicted value when the selected model is used and the value obtained by multiplying the mixture ratio of the model, and adding the sum for the number of selected models. In other words, this is a process of summing the predicted values ​​of each selected prediction model according to the weight ratio between the prediction models. In other words, this is an example of a process of calculating a predicted value by summing the predicted values ​​of each prediction model selected as the prediction model to be used in calculating the predicted value according to the ratio of the reciprocal of the error between the prediction models.

[0164] In the second embodiment, a method has been described in which a model to be used for calculating future predicted values ​​is selected from among the candidate models by setting an error margin rate, but other selection methods may also be used.

[0165] For example, similar to the selection method in the "prediction model selection category" in the first embodiment, an upper limit on the number of prediction models may be set as the top designated number, and a designated number of models may be selected in order of smallest error. In other words, this is an example of a process of selecting a predetermined number of prediction models in order of smallest error.

[0166] Alternatively, a cumulative prediction error value may be set and models selected so that the cumulative prediction error value is equal to or less than a specified value. Alternatively, a prediction error value may be set and models whose prediction error values ​​are equal to or less than the specified value may be selected. In other words, this is an example of a process of selecting prediction models in order from smallest to largest error until the total prediction error satisfies a predetermined condition.

[0167] This allows the user to set the method for calculating future predicted values ​​and the method for selecting the model to use in calculating future predicted values. By calculating the blending ratio based on the magnitude of the error, models with small errors have a large effect on the predicted value, and models with large errors have a small effect on the predicted value, enabling accurate predictions using multiple models.

[0168] Although the embodiments have been described above, the present invention can be embodied as, for example, a system, an apparatus, a method, a program, a recording medium, etc. Specifically, the present invention may be applied to a system made up of multiple devices, or may be applied to an apparatus made up of a single device.

[0169] The program of the present invention is a program that enables a computer to execute the processing method of the flowchart shown in the figure, and the storage medium of the present invention stores a program that enables a computer to execute the processing method shown in the figure. Note that the program of the present invention may be a program for each processing method of each device in the figure.

[0170] As described above, it goes without saying that the object of the present invention can also be achieved by supplying a recording medium on which a program that realizes the functions of the above-mentioned embodiments is recorded to a system or device, and having the computer (or CPU or MPU) of that system or device read and execute the program stored on the recording medium.

[0171] In this case, the program itself read from the recording medium will realize the novel functions of the present invention, and the recording medium on which the program is recorded will constitute the present invention.

[0172] Examples of recording media for supplying the program include flexible disks, hard disks, optical disks, magneto-optical disks, CD-ROMs, CD-Rs, DVD-ROMs, magnetic tapes, non-volatile memory cards, ROMs, EEPROMs, and silicon disks.

[0173] Furthermore, it goes without saying that not only are the functions of the above-mentioned embodiments realized by the computer executing a program it has read, but also cases are included in which an OS (operating system) running on the computer performs some or all of the actual processing based on the instructions of the program, and the functions of the above-mentioned embodiments are realized through that processing.

[0174] Furthermore, it goes without saying that this also includes cases where a program read from a recording medium is written into a memory provided on a function expansion board inserted into a computer or a function expansion unit connected to the computer, and then a CPU or the like provided on the function expansion board or function expansion unit performs some or all of the actual processing based on the instructions of the program code, thereby realizing the functions of the above-mentioned embodiments.

[0175] Furthermore, the present invention may be applied to a system consisting of multiple devices, or to a device consisting of a single device. It goes without saying that the present invention can also be applied to a case where the present invention is achieved by supplying a program to a system or device. In this case, the system or device can enjoy the effects of the present invention by reading a recording medium containing a program for achieving the present invention into the system or device.

[0176] Furthermore, by downloading and reading a program for achieving the present invention from a server, database, etc. on a network using a communication program, the system or device can enjoy the effects of the present invention. Note that the present invention also includes configurations that combine the above-mentioned embodiments and their modified examples. [Explanation of symbols]

[0177] 100 Supply and Demand Management System 110 Client Terminal 120 Network 130 servers

Claims

1. a first calculation means for calculating a predicted value for an evaluation period using a predetermined prediction method based on time series data that is past actual values ​​related to the product to be predicted, and calculating an error between the calculated predicted value and the time series data that is actual values ​​for the evaluation period; a reception means for receiving a selection method of a prediction method to be used for calculating a future predicted value from among a plurality of predetermined prediction methods; a second calculation means for calculating a future predicted value of the product to be predicted using a plurality of prediction methods identified by the selection method accepted by the acceptance means and the magnitude of the error calculated by the first calculation means; An information processing device comprising:

2. The information processing device according to claim 1, characterized in that the second calculation means calculates a future predicted value for the product to be predicted based on a predicted value calculated using the plurality of prediction methods identified by the selection method accepted by the acceptance means and a value relating to an error of the prediction method.

3. 3. The information processing apparatus according to claim 2, wherein the value relating to the error is a value based on the reciprocal of the error.

4. 2. The information processing apparatus according to claim 1, wherein the selection method is a method of selecting the plurality of prediction methods to be used in the calculation of the predicted value by the second calculation means based on a condition based on the magnitude of a value related to an error of the prediction method.

5. 2. The information processing apparatus according to claim 1, wherein the selection method is a method of selecting the prediction methods in order from the smallest error until a total of prediction errors satisfies a predetermined condition.

6. 2. The information processing apparatus according to claim 1, wherein the selection method selects, from among the prediction methods, a prediction method that satisfies a predetermined condition based on an error of a prediction method with the smallest error as the prediction method to be used for calculating the predicted value by the second calculation means.

7. 2. The information processing apparatus according to claim 1, wherein the selection method is a method of selecting a predetermined number of prediction methods from among the prediction methods in descending order of the error.

8. 2. The information processing apparatus according to claim 1, wherein the second calculation means calculates the predicted value by summing the predicted values ​​of the respective prediction methods selected as the prediction methods to be used in calculating the predicted value in accordance with a ratio of the reciprocals of the errors of the respective prediction methods.

9. 2. The information processing device according to claim 1, wherein the selection method comprises selecting a plurality of prediction methods from among preset prediction methods for calculating a future predicted value for the product to be predicted based on the error.

10. 2 . The information processing apparatus according to claim 1 , wherein the second calculation means calculates the future demand for the commodity related to the commodity to be predicted using a plurality of prediction methods selected by the selection method.

11. a selection receiving means for receiving a selection of whether to use the prediction method selected by the selection method or a prediction method selected by a user as a prediction method to be used in calculation of a predicted value by the second calculation means, 2. The information processing device according to claim 1, wherein, when the selection accepting means accepts a selection to use the prediction method selected by the selection method, the second calculation means calculates a predicted value using the prediction method selected by the selection method, and when the selection accepting means accepts a selection to use the prediction method selected by the user, the second calculation means calculates a predicted value using the prediction method selected by the user.

12. A first calculation means for calculating a predicted value for an evaluation period using a predetermined prediction method based on time series data which are past actual values ​​related to the product to be predicted, and calculating the error between the calculated predicted value and the time series data which are actual values ​​for the evaluation period; a selection means for selecting, from among a plurality of prediction methods for calculating future predicted values, a prediction method from the predetermined prediction methods in order of decreasing error until the sum of the errors satisfies a predetermined condition; a second calculation means for calculating a future predicted value of the product to be predicted using the plurality of prediction methods selected by the selection means; An information processing device comprising:

13. A first calculation means for calculating a predicted value for an evaluation period using a predetermined prediction method based on time series data which are past actual values ​​related to the product to be predicted, and calculating the error between the calculated predicted value and the time series data which are actual values ​​for the evaluation period; a selection means for selecting, from among the plurality of prediction methods for calculating future predicted values, a prediction method that satisfies a predetermined condition based on the error of the prediction method with the smallest error; a second calculation means for calculating a future predicted value of the product to be predicted using the plurality of prediction methods selected by the selection means; An information processing device comprising:

14. A first calculation means for calculating a predicted value for an evaluation period using a predetermined prediction method based on time series data which are past actual values ​​related to the product to be predicted, and calculating the error between the calculated predicted value and the time series data which are actual values ​​for the evaluation period; a selection means for selecting a predetermined number of prediction methods from the predetermined prediction methods in descending order of the error as the plurality of prediction methods for calculating future predicted values; a second calculation means for calculating a future predicted value of the product to be predicted using the plurality of prediction methods selected by the selection means; An information processing device comprising:

15. a first calculation step in which a first calculation means of the information processing device calculates a predicted value for an evaluation period using a predetermined prediction method based on time series data that is past actual values ​​related to the product to be predicted, and calculates an error between the calculated predicted value and the time series data that is actual values ​​for the evaluation period; a receiving step in which a receiving means of the information processing device receives a selection method of a prediction method to be used for calculating a future predicted value from among a plurality of selection methods; a second calculation step in which second calculation means of the information processing device calculates a future predicted value of the product to be predicted using a plurality of prediction methods identified by the selection method accepted in the acceptance step and the magnitude of the error calculated in the first calculation step; 10. A method for controlling an information processing device, comprising:

16. A first calculation step in which a first calculation means of the information processing device calculates a predicted value for an evaluation period using a predetermined prediction method based on time series data which are past actual values ​​related to the product to be predicted, and calculates an error between the calculated predicted value and the time series data which are actual values ​​for the evaluation period; a selection step in which a selection means of the information processing device selects, from among the predetermined prediction methods, a prediction method having a smallest error as a plurality of prediction methods for calculating a future predicted value until the sum of the errors satisfies a predetermined condition; a second calculation step in which second calculation means of the information processing device calculates a future predicted value related to the product to be predicted using the plurality of prediction methods selected by the selection step; 10. A method for controlling an information processing device, comprising:

17. A first calculation step in which a first calculation means of the information processing device calculates a predicted value for an evaluation period using a predetermined prediction method based on time series data which are past actual values ​​related to the product to be predicted, and calculates an error between the calculated predicted value and the time series data which are actual values ​​for the evaluation period; a selection step in which a selection means of the information processing device selects, from among the predetermined prediction methods, a prediction method that satisfies a predetermined condition based on the error of the prediction method with the smallest error, as the prediction method for calculating a future predicted value; a second calculation step in which second calculation means of the information processing device calculates a future predicted value related to the product to be predicted using the plurality of prediction methods selected by the selection step; 10. A method for controlling an information processing device, comprising:

18. A first calculation step in which a first calculation means of the information processing device calculates a predicted value for an evaluation period using a predetermined prediction method based on time series data which are past actual values ​​related to the product to be predicted, and calculates an error between the calculated predicted value and the time series data which are actual values ​​for the evaluation period; a selection step in which a selection means of the information processing device selects, from the predetermined prediction methods, a predetermined number of prediction methods in descending order of the error as the plurality of prediction methods for calculating a future predicted value; a second calculation step in which second calculation means of the information processing device calculates a future predicted value related to the product to be predicted using the plurality of prediction methods selected by the selection step; 10. A method for controlling an information processing device, comprising:

19. A program for causing at least one computer to function as each of the means of the information processing device according to any one of claims 1 to 14.

Images