Anomaly correction device, anomaly correction method, and anomaly correction program

Abstract

We will develop an outlier correction device that can present users with options for more accurate product demand forecasting. [Solution] The abnormal value correction device comprises: an abnormal value detection unit that detects abnormal values ​​from actual values ​​of the volume of goods distributed for each unit period; a reception unit that receives a user's specification of the starting point of a first period corresponding to the abnormal value detected by the abnormal value detection unit; a correction proposal unit that calculates a correction value for the actual values ​​of the first period based on the actual values ​​of a second period prior to the starting point received by the reception unit, and proposes the calculated correction value to the user; and an option presentation unit that presents the user with a first option of not correcting the actual values ​​of the first period, a second option of provisionally correcting the actual values ​​of the first period with the correction value, and a third option of finalizing the correction of the actual values ​​of the first period.

Description

[Technical Field]

[0001] This disclosure relates to an abnormal value correction device, an abnormal value correction method, and an abnormal value correction program. [Background technology]

[0002] Demand forecasting techniques are known. For example, Patent Document 1 discloses a forecasting system that obtains the following input information for a product: the product category, the number of products shipped from wholesalers to retailers one day before to seven days before the target date, and the number of days the product is in stock at the retailer or the average number of products sold. By inputting this information into a learning model generated by performing supervised learning, the system outputs forecast information regarding the number of products that wholesalers will ship to retailers on the target date. [Prior art documents] [Patent Documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-084757 [Overview of the Initiative] [Problems that the invention aims to solve]

[0004] While the technology described in Patent Document 1 can predict the number of goods shipped from wholesalers to retailers, there is room for improvement in terms of more accurate demand forecasting for goods.

[0005] This disclosure has been made in view of the above-mentioned issues, and one exemplary purpose is to provide a technology for more accurately forecasting product demand. [Means for solving the problem]

[0006] An example of an abnormal value correction device relating to this disclosure includes: an abnormal value detection means for detecting abnormal values ​​from actual values ​​of the volume of goods distributed for each unit period; a receiving means for receiving a user's specification of the starting point of a first period corresponding to the abnormal value detected by the abnormal value detection means; a correction proposal means for calculating a correction value for the actual values ​​of the first period based on the actual values ​​of a second period prior to the starting point received by the receiving means, and proposing the calculated correction value to the user; and an option presentation means for presenting the user with a first option of not correcting the actual values ​​of the first period, a second option of provisionally correcting the actual values ​​of the first period with the correction value, and a third option of finalizing the correction of the actual values ​​of the first period.

[0007] An example of an anomaly correction method relating to the present disclosure includes: an anomaly detection process in which at least one processor detects an anomaly from the actual values ​​of the volume of goods distributed for each unit period; an acceptance process in which the at least one processor accepts a user's specification of the starting point of a first period corresponding to the anomaly detected in the anomaly detection process; a correction proposal process in which the at least one processor calculates a correction value for the actual values ​​of the first period based on the actual values ​​of a second period prior to the starting point accepted in the acceptance process, and proposes the calculated correction value to the user; and an option presentation process in which the at least one processor presents the user with a first option of not correcting the actual values ​​of the first period, a second option of provisionally correcting the actual values ​​of the first period with the correction value, and a third option of finalizing the correction of the actual values ​​of the first period.

[0008] An example of an anomaly correction program relating to this disclosure is an anomaly correction program for causing a computer to function as an anomaly correction device, wherein the computer functions as: an anomaly detection means for detecting an anomaly from actual values ​​of the volume of goods distributed for each unit period; a receiving means for receiving a user's specification of the starting point of a first period corresponding to the anomaly detected by the anomaly detection means; a correction proposal means for calculating a correction value for the actual values ​​of the first period based on the actual values ​​of a second period prior to the starting point received by the receiving means, and proposing the calculated correction value to the user; and an option presentation means for presenting the user with a first option of not correcting the actual values ​​of the first period, a second option of provisionally correcting the actual values ​​of the first period with the correction value, and a third option of finalizing the correction of the actual values ​​of the first period. [Effects of the Invention]

[0009] One illustrative effect of this disclosure is that it can provide a technology that can more accurately forecast product demand. [Brief explanation of the drawing]

[0010] [Figure 1] This is a block diagram showing the configuration of the abnormal value correction device related to this disclosure. [Figure 2] This flowchart shows the flow of the abnormal value correction method related to this disclosure. [Figure 3] This figure shows the configuration of the demand forecasting system related to this disclosure. [Figure 4] This is a block diagram showing the configuration of the information processing device related to this disclosure. [Figure 5] This is a flowchart showing the flow of the information processing method related to this disclosure. [Figure 6] This figure shows an example of a screen related to this disclosure. [Figure 7] This figure shows an example of a screen related to this disclosure. [Figure 8] This is a block diagram showing the configuration of a computer that functions as an abnormal value correction device and information processing device related to this disclosure.

Best Mode for Carrying Out the Invention

[0011] Hereinafter, embodiments of the present invention will be exemplified. However, the present invention is not limited to each of the exemplary embodiments shown below, and various modifications are possible within the scope shown in the claims. For example, embodiments obtained by appropriately combining the technologies (part or all of an object or method) adopted in each of the exemplary embodiments shown below may also be included in the scope of the present invention. Also, embodiments obtained by appropriately omitting part of the technologies adopted in each of the exemplary embodiments shown below may also be included in the scope of the present invention. Further, the effects mentioned in each of the exemplary embodiments shown below are merely examples of the effects expected in those exemplary embodiments and do not define the extension of the present invention. That is, embodiments that do not exhibit the effects mentioned in each of the exemplary embodiments shown below may also be included in the scope of the present invention.

[0012] 〔First Exemplary Embodiment〕 A first exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. This exemplary embodiment is a basic form for each of the exemplary embodiments described later. Note that the scope of application of each technology adopted in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technology adopted in this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure as long as there are no particular technical obstacles. Also, each technology shown in the drawings referred to for explaining this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure as long as there are no particular technical obstacles.

[0013] (Configuration of Outlier Correction Device) The configuration of the outlier correction device 1 will be described with reference to FIG. 1. FIG. 1 is a block diagram showing the configuration of the outlier correction device 1. As shown in FIG. 1, the outlier correction device 1 includes an outlier detection unit 11, a reception unit 12, a correction proposal unit 13, and an option presentation unit 14.

[0014] The outlier detection unit 11 detects outliers from the actual values of the commodity circulation volume for each unit period. The reception unit 12 receives a designation of the start point of the first period corresponding to the outlier detected by the outlier detection unit 11 by the user. The correction proposal unit 13 calculates a correction value for the actual value of the first period based on the actual value in the second period before the start point received by the reception unit 12, and proposes the calculated correction value to the user. The option presentation unit 14 presents to the user a first option not to correct the actual value of the first period, a second option to temporarily correct the actual value of the first period with the correction value, and a third option to confirm the correction of the actual value of the first period.

[0015] (Effect of the Outlier Correction Device) As described above, in the outlier correction device 1, there are provided an outlier detection unit 11 that detects outliers from the actual values of the commodity circulation volume for each unit period, a reception unit 12 that receives a designation of the start point of the first period corresponding to the outlier detected by the outlier detection unit 11 by the user, a correction proposal unit 13 that calculates a correction value for the actual value of the first period based on the actual value in the second period before the start point received by the reception unit 12, and proposes the calculated correction value to the user, and an option presentation unit 14 that presents to the user a first option not to correct the actual value of the first period, a second option to temporarily correct the actual value of the first period with the correction value, and a third option to confirm the correction of the actual value of the first period. For this reason, according to the outlier correction device 1, an effect can be obtained that an option for more accurately performing demand prediction of a commodity can be presented to the user.

[0016] (Flow of the Outlier Correction Method) The flow of the outlier correction method S1 will be described with reference to FIG. 2. FIG. 2 is a flowchart showing the flow of the outlier correction method S1. As shown in FIG. 2, the outlier correction method S1 includes an outlier detection process S11, a reception process S12, a correction proposal process S13, and an option presentation process S14.

[0017] In the anomaly detection process S11, at least one processor detects anomalies from the actual values ​​of the volume of goods distributed for each unit period. In the reception process S12, the at least one processor accepts the user's specification of the starting point of a first period corresponding to the anomaly detected in the anomaly detection process S11. In the correction proposal process S13, the at least one processor calculates a correction value for the actual values ​​of the first period based on the actual values ​​of a second period prior to the starting point accepted in the reception process S12, and proposes the calculated correction value to the user. In the option presentation process S14, the at least one processor presents the user with a first option of not correcting the actual values ​​of the first period, a second option of provisionally correcting the actual values ​​of the first period with the correction value, and a third option of confirming the correction of the actual values ​​of the first period.

[0018] (Effect of outlier correction method) As described above, the anomaly correction method S1 employs a configuration that includes: an anomaly detection process S11 in which at least one processor detects an anomaly from the actual values ​​of the volume of goods distributed for each unit period; a reception process S12 in which the at least one processor receives a user's specification of the starting point of a first period corresponding to the anomaly detected in the anomaly detection process S11; a correction proposal process S13 in which the at least one processor calculates a correction value for the actual values ​​of the first period based on the actual values ​​of a second period prior to the starting point received in the reception process S12, and presents the calculated correction value to the user; and a choice presentation process S14 in which the at least one processor presents the user with a first option of not correcting the actual values ​​of the first period, a second option of provisionally correcting the actual values ​​of the first period with the correction value, and a third option of finalizing the correction of the actual values ​​of the first period. Therefore, the outlier correction method S1 has the effect of presenting users with options for more accurately forecasting product demand.

[0019] [Second exemplary embodiment] A second exemplary embodiment, which is an example of an embodiment of the present invention, will be described in detail with reference to the drawings. Components having the same function as those described in the above-described exemplary embodiment are denoted by the same reference numerals, and their descriptions are omitted as appropriate. The scope of application of each technology adopted in this exemplary embodiment is not limited to this exemplary embodiment. That is, each technology adopted in this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure, to the extent that no particular technical problems arise. Furthermore, each technology shown in the drawings referenced to describe this exemplary embodiment can also be adopted in other exemplary embodiments included in this disclosure, to the extent that no particular technical problems arise.

[0020] (Configuration of the demand forecasting system) The configuration of the demand forecasting system 100A related to this disclosure will be described with reference to Figure 3. Figure 3 is a block diagram showing the configuration of the demand forecasting system 100A. The demand forecasting system 100A is a system that forecasts the demand for a product and comprises an information processing device 1A and a user terminal 2A. The information processing device 1A and the user terminal 2A are connected to communicate via a communication line N. The specific configuration of the communication line N is not limited to this embodiment, but examples of communication line N include wireless LAN (Local Area Network), wired LAN, WAN (Wide Area Network), public telephone network, mobile data communication network, or a combination thereof.

[0021] Information processing device 1A is a device equipped with the function of predicting product demand, and is, for example, a general-purpose server. Information processing device 1A may also be a personal computer such as a laptop PC or tablet terminal. User terminal 2A is a terminal used by a user (for example, a manufacturer's planner), and is, for example, a personal computer such as a laptop PC or tablet terminal. In the example in Figure 3, one user terminal 2A is illustrated, but two or more user terminals 2A may be included in the demand forecasting system 100A.

[0022] (Configuration of information processing device) The configuration of the information processing device 1A will be explained with reference to Figure 4. Figure 4 is a block diagram showing the configuration of the information processing device 1A. The information processing device 1A comprises a control unit 10A, a storage unit 20A, a communication unit 30A, an input unit 40A, and an output unit 50A. The communication unit 30A communicates with external devices (user terminal 2A, etc.) of the information processing device 1A via a communication line N. The communication unit 30A transmits data supplied from the control unit 10A to other devices and supplies data received from other devices to the control unit 10A.

[0023] (Input section / Output section) The input unit 40A is configured to receive input to the information processing device 1A, and may include, for example, an input device such as a keyboard, mouse, touch panel, camera, or microphone. The input unit 40A may also be configured to receive data from the input device via an interface such as USB (Universal Serial Bus). The output unit 50A is configured to output from the information processing device 1A, and may include, for example, an output device such as a display, printer, touch panel, or speaker. The output unit 50A may also be configured to have an interface such as USB and output data to the output device via that interface.

[0024] (Storage part) The memory unit 20A stores various types of information that are referenced by the control unit 10A. Examples of such information include distribution performance data 201, demand forecast data 202, and the machine learning model LM. Here, when we say that the machine learning model LM is stored in the memory unit 20A, we mean that the parameters that define the machine learning model LM are stored in the memory unit 20A.

[0025] (Distribution history data, demand forecast data) Distribution performance data 201 is time-series data of actual distribution volume of goods for each unit period. Here, the unit period can be, for example, one month, one week, or one day. Actual values ​​can be, for example, the number of goods sold (individuals, boxes, etc.) or the sales amount. Distribution performance data 201 may include, as an example, POS data (Point of Sale system), e-commerce shipment data, wholesale shipment data, manufacturer shipment data, etc. POS data is data that aggregates sales performance of goods at retailers on an individual item basis. POS data may include, as an example, product name, price, quantity, store of purchase, date and time of purchase, and customer information (age group, gender, date and time of visit, etc.). E-commerce shipment data is data that shows the sales performance of goods through e-commerce. Wholesale shipment data is data that shows the shipment performance of goods by wholesalers. Manufacturer shipment data is data that shows the shipment performance of goods by manufacturers. Demand forecast data 202 is data that shows the forecast results of demand for goods. Demand forecast data 202 may include, as an example, time-series data of forecast values ​​of demand for goods for each unit period.

[0026] (Machine learning model) The machine learning model LM is a machine learning model that the training unit 17A, described later, trains using training data. Examples of machine learning model LM include, but are not limited to, neural network models such as convolutional neural networks or recurrent neural networks. The machine learning model LM learns the factors behind past outliers and predicts the degree of future demand fluctuations. The input to the machine learning model LM may include, as an example, actual values ​​of the volume of goods distributed and information indicating the factors causing fluctuations. The input to the machine learning model LM may also include information related to the sale of goods (main sales channels, number of distribution stores, scale of shipments and sales), information indicating the category of goods, information related to the marketing of goods, information indicating the external environment, etc. The output of the machine learning model LM may, as an example, include data showing the rate of change in the demand forecast due to the factors causing fluctuations (e.g., what percentage will demand fall?).

[0027] (Control Unit) The control unit 10A includes an abnormal value detection unit 11A, a reception unit 12A, a correction proposal unit 13A, a choice presentation unit 14A, a demand forecasting unit 15A, a training data generation unit 16A, a training unit 17A, and a display control unit 18A. The abnormal value detection unit 11A, the reception unit 12A, the correction proposal unit 13A, the choice presentation unit 14A, the training data generation unit 16A, and the training unit 17A are examples of the abnormal value detection means, reception means, correction proposal means, choice presentation means, training data generation means, and training means according to this disclosure, respectively. The demand forecasting unit 15A is an example of the forecasting means, second forecasting means, and third forecasting means according to this disclosure.

[0028] (Anomaly detection unit) The anomaly detection unit 11A detects anomalies from the distribution performance data 201. An anomaly is, for example, a value that shows unusual behavior. The anomaly detection unit 11A detects anomalies by, for example, comparing the deviation of the performance value from the mean with the standard deviation. More specifically, the anomaly detection unit 11A, for example, when the standard deviation of the distribution performance data 201 over a predetermined period is σ and the mean is μ, |x N-i The actual value x for which the value of -μ| / σ is greater than or equal to a predetermined threshold. N-i This is detected as an abnormal value. However, the method by which the abnormal value detection unit 11A detects abnormal values ​​is not limited to the example described above, and the abnormal value detection unit 11A may detect abnormal values ​​by other methods.

[0029] (Reception Department) The reception unit 12A receives various instructions or selections from the user. For example, the reception unit 12A receives the instructions or selections from the user terminal 2A by receiving data indicating the user's instructions or selections. The reception unit 12A may also receive instructions or selections entered by the user into the input unit 40A.

[0030] The reception unit 12A specifically accepts the user's specification of the start date of the period (first period) corresponding to the abnormal value detected by the abnormal value detection unit 11A. Furthermore, if the user selects a third option from among the multiple options presented by the option presentation unit 14A (described later), the reception unit 12A accepts the user's specification of the end date of the period corresponding to the abnormal value detected by the abnormal value detection unit 11A. Furthermore, if the user selects a third option, the reception unit 12A accepts the user's specification of the fluctuation factor. Here, the fluctuation factor is the factor that caused the actual value to change (an abnormal value occurred). Examples of fluctuation factors include, but are not limited to, stockouts, sales of limited-quantity items (such as campaign items), measures, new adoptions, and items being removed from shelves.

[0031] (Proposal Department for Supplementary Budgets) The correction proposal unit 13A calculates a correction value for the actual values ​​of the period corresponding to the anomaly (the first period) based on the actual values ​​of the period prior to the starting point (the second period) received by the reception unit 12A. As an example, the correction proposal unit 13A predicts the actual values ​​after the starting point by performing a time series analysis using the actual values ​​of the period prior to the starting point received by the reception unit 12A, and uses the predicted actual values ​​as the correction value. However, the method by which the correction proposal unit 13A calculates the correction value is not limited to the example described above, and the correction proposal unit 13A may calculate the correction value using other methods.

[0032] Furthermore, the correction suggestion unit 13A proposes the calculated correction value to the user. For example, the correction suggestion unit 13A outputs the calculated correction value to an output device (display, printer, speaker, etc.) and proposes the correction value to the user. For example, the output device is the output device of the user terminal 2A. In this case, the correction suggestion unit 13A transmits the correction value to the user terminal 2A via the communication unit 30A and causes the output device of the user terminal 2A to output the correction value. Alternatively, the correction suggestion unit 13A may propose the correction value to the user by outputting the correction value to an output device (display, speaker, printer, etc.) connected to the output unit 50A.

[0033] (Option presentation section) The option presentation unit 14A presents the user with several options regarding the correction. These options include, for example, three choices: "wait and see," "provisional correction," and "confirm correction." "Wait and see" is the option (first option) in which no correction is made to the actual values ​​for the period corresponding to the anomaly. If this option is selected, the actual values ​​are not corrected, and the corrected values ​​are not reflected in the actual values. In this case, the demand forecasting unit 15A calculates the future demand forecast using the actual values ​​up to the point before the period of the anomaly, excluding the period of the anomaly.

[0034] "Provisional correction" is an option (second option) to provisionally correct the actual values ​​for the period corresponding to the outlier using the correction value mentioned above. Here, provisional correction means not correcting the actual values ​​with the correction value, but retaining both the actual values ​​and the correction value. If this option is selected, the actual values ​​are not corrected, but the correction value is retained in a storage device or the like. In this case, the demand forecasting unit 15A uses the correction value instead of the actual values ​​to perform demand forecasting for the forecasting period.

[0035] "Confirm Correction" is the third option, which confirms the correction of the actual values ​​for the period corresponding to the outlier. Here, confirming the correction means updating the actual values ​​with the corrected values. If this option is selected, the actual values ​​are updated with the corrected values. However, the actual values ​​before correction are also stored separately in the memory unit 20A, etc. In this case, the demand forecasting unit 15A uses the updated actual values ​​(i.e., the corrected values) to perform demand forecasting for the forecasting period.

[0036] If the impact of fluctuating factors such as stock shortages or promotional measures continues, users can select "Provisional Adjustment" without finalizing the adjustment. On the other hand, if the impact of fluctuating factors such as stock shortages or promotional measures has ended, users can finalize the adjustment and use the finalized adjustment value for subsequent demand forecasts.

[0037] (Demand Forecasting Department) The demand forecasting unit 15A forecasts the demand for a product using the distribution history data 201. Examples of demand forecasting methods used by the demand forecasting unit 15A include, but are not limited to, (a) a time series analysis method and (b) a pre-trained model method. (a) The time series analysis method analyzes data that changes over time and predicts how the data will change in the next period; conventional analysis methods can be used.

[0038] (b) The method using a pre-trained model is a method of forecasting demand by inputting various data into a pre-trained model generated by machine learning. In this case, the pre-trained model may be, but is not limited to, a neural network model such as a convolutional neural network or a recurrent neural network. In this case, the input to the pre-trained model may include, as an example, distribution history data. In addition to distribution history data, the input to the pre-trained model may also include other data such as information indicating the category of the product, information related to the marketing of the product, and information indicating the external environment for each unit period. An example of information indicating the category of the product is information indicating the category of the product, such as "drinking water" or "fresh food." An example of information related to the marketing of the product is information indicating the scale and duration of the promotion. An example of information indicating the external environment is the average temperature and the number of foreign visitors to Japan. In addition, an example of the output of the pre-trained model is the forecast result of demand for each unit period of the product during the forecast period. The pre-trained model is stored in a memory device such as the memory unit 20A. Here, when we say that the pre-trained model is stored in the memory device, we mean that the parameters that define the pre-trained model are stored in the memory device.

[0039] Furthermore, the demand forecasting unit 15A performs demand forecasting according to the option selected by the user. More specifically, for example, if the user selects "wait and see" (first option), the demand forecasting unit 15A forecasts the demand for the product using actual values ​​prior to the first period without correcting the actual values ​​for the first period. On the other hand, if the user selects "preliminary correction" (second option), the demand forecasting unit 15A corrects the actual values ​​for the first period with correction values ​​and forecasts the demand for the product using the corrected actual values.

[0040] Furthermore, if the user selects "Confirm Correction" (the third option), the demand forecasting unit 15A calculates a correction value for the actual values ​​for the first period from the starting point to the ending point received by the receiving unit 12A, corrects the actual values ​​for the above period using the calculated correction value, and forecasts the demand for the product using the corrected actual values.

[0041] (Training data generation unit) The training data generation unit 16A generates training data to be used to train the machine learning model LM. The training data is, as an example, data that associates first data, which includes actual values ​​of the volume of goods distributed and variable factors specified by the user, with second data, which includes actual values ​​corrected by the demand forecasting unit 15A.

[0042] (Training Department) The training unit 17A trains the machine learning model LM using the training data generated by the training data generation unit 16A. As an example, the training unit 17A uses the actual value of the distribution volume of the above product and the fluctuation factors specified by the above user as input data, and the corrected value of the above actual value as output data, and trains the machine learning model LM by relating the two.

[0043] (Display Control Unit) The display control unit 18A outputs data representing various screens to a display (display device) and displays the screen on the display. The display is, for example, the display of a user terminal 2A. In this case, the display control unit 18A transmits data representing the screen to the user terminal 2A via the communication unit 30A and displays the screen on the display of the user terminal 2A. In this disclosure, the act of the display control unit 18A transmitting data representing the screen to the user terminal 2A and displaying the screen on the display of the user terminal 2A is also referred to as "the display control unit 18A displays the screen". Alternatively, the display control unit 18A may display the screen on a display connected to the output unit 50A by outputting data representing the screen to the display.

[0044] (Information processing flow) Figure 5 is a flowchart showing an example of the information processing method performed by the demand forecasting system 100A.

[0045] (Step S101) In step S101, the anomaly detection unit 11A detects anomalies from the distribution performance data 201. As an example, the anomaly detection unit 11A detects anomalies by comparing the deviation of the performance value from the mean with the standard deviation. More specifically, as an example, if there is previous year's performance data, the anomaly detection unit 11A takes the standard deviation of the previous year's performance data over a predetermined period as σ and the mean as μ, and |x N-i The actual value x for which the value of -μ| / σ is greater than or equal to a predetermined threshold. N-i This is detected as an outlier. On the other hand, if there is no data from the previous year, the outlier detection unit 11A, as an example, uses σ as the standard deviation and μ as the mean for the most recent predetermined period, and |x N-i The actual value x for which the value of -μ| / σ is greater than or equal to a predetermined threshold. N-i This is detected as an outlier.

[0046] (Step S102) In step S102, the display control unit 18A presents the abnormal value detected by the abnormal value detection unit 11A to the user. As an example, the display control unit 18A presents the detected abnormal value to the user by displaying a screen indicating the abnormal value on the display.

[0047] (Display Example 1) FIG. 6 is a diagram showing an example of a screen displayed by the display control unit 18A on the display. In the example of FIG. 6, in the display area A11 of the display, the distribution performance data indicating the actual performance of the distribution volume in each of the unit periods N-12, N-11, …, and the demand prediction data indicating the predicted results of the demand in each of the unit periods N, N+1, N+2 are displayed.

[0048] In the example of FIG. 6, the distribution performance data includes a plurality of items such as "Deviation from Average", " / Standard Deviation", "Alert", "Correction Plan", and "Correction Value (Determined Value)". "Deviation from Average" is a value indicating how far the actual value is from the average. In the example of FIG. 6, if the standard deviation in a predetermined period of the distribution performance data is σ and the average is μ, the deviation of the actual value x N-i from the average μ is calculated by |x N-i -μ|. " / Standard Deviation" is the value |x N-i -μ| / σ obtained by dividing the deviation from the average |x N-i -μ| by the standard deviation σ. This value is also referred to as the alert index P N-i .

[0049] "Alert" displays information indicating whether the actual value is an abnormal value. In the example of FIG. 6, when the actual value is an abnormal value, a mark M11 indicating that fact is displayed in this item, while when the actual value is not an abnormal value, nothing is displayed in this item.

[0050] As an example, the alert index P N-i is used to determine whether the actual value is an abnormal value. As an example, the abnormal value detection unit 11A determines that the actual value is an abnormal value when the alert index P N-i is greater than or equal to a predetermined threshold.

[0051] Furthermore, in the display area A11 of Figure 6, the user can specify the starting point of the period corresponding to the abnormal value. More specifically, if the actual value for unit period N-1 is detected as an abnormal value, the user uses the input device of user terminal 2A to specify either unit period N-1 or a unit period prior to unit period N-1 as the starting point.

[0052] (Screen example 2) In addition to the table TBL11 illustrated in Figure 6, the display control unit 18A may also display the graph shown in Figure 7. Figure 7 is a diagram showing an example of a graph that the display control unit 18A displays on the display. In the example in Figure 7, the display area A12 displays a graph of the actual value of the amount of goods distributed, a value indicating the deviation from the average, and the value obtained by dividing the deviation from the average by the standard deviation.

[0053] (Step S103・S104) In step S103, the reception unit 12A accepts the user's specification of the starting point. In step S104, the correction proposal unit 13A calculates a correction value for the actual value of the period corresponding to the abnormal value (first period) based on the actual value of the period prior to the starting point (second period) accepted in step S103.

[0054] (Step S105) In step S105, the correction proposal unit 13A proposes the calculated correction value to the user by displaying it on the display, etc. Specifically, as an example, the correction proposal unit 13A displays the calculated correction proposal in the "Correction Proposal" field F11 of the unit period N-1 in the display area A11 of Figure 6.

[0055] Furthermore, in step S105, the option presentation unit 14A presents the user with multiple options (e.g., "wait and see", "preliminary correction", "confirm correction"). As an example, the option presentation unit 14A displays a GUI (graphical user interface) in display area A11 for selecting one of the multiple options. The user uses the input device of the user terminal 2A to select one of the presented options. Once the user has selected one of the options, the user terminal 2A transmits data indicating the selected option to the information processing device 1A.

[0056] (Step S106) In step S106, the option presentation unit 14A determines which option the user has selected. If "Wait and See" is selected (step S106: "Wait and See"), the option presentation unit 14A proceeds to step S107. If "Provisional Correction" is selected (step S106: "Provisional Correction"), the option presentation unit 14A proceeds to step S109. If "Confirm Correction" is selected (step S106: "Confirm Correction"), the option presentation unit 14A proceeds to step S111.

[0057] (Step S107・S108) In step S107, the demand forecasting unit 15A forecasts the demand for the product using uncorrected actual values. In step S108, the display control unit 18A presents the demand forecasting result from step S107 to the user by displaying it on a display or the like.

[0058] (Step S109・S110) In step S109, the demand forecasting unit 15A forecasts the demand for the product using the correction value calculated in step S104. In step S110, the display control unit 18A presents the demand forecast result from step S109 to the user by displaying it on a display or the like.

[0059] (Step S111) In step S111, the reception unit 12A receives the specification of the end point of the period corresponding to the abnormal value and the selection of the fluctuation factor. As an example, in the display area A11 of Figure 6, the reception unit 12A displays information on the display for the user to specify the end point of the period corresponding to the abnormal value and accepts the specification of the end point. The reception unit 12A also displays information on the display for selecting the factor of the abnormal value in the display area A11 and accepts the selection of the fluctuation factor. The user specifies the end point and selects the fluctuation factor using the input device of the user terminal 2A. Once the user has specified the end point and selected the fluctuation factor, the user terminal 2A transmits data indicating the end point and data indicating the fluctuation factor to the information processing device 1A.

[0060] (Step S112・S113) In step S112, the demand forecasting unit 15A updates the actual values ​​with corrected values ​​and uses the corrected values ​​to forecast the demand for the product. In step S113, the display control unit 18A presents the demand forecasting results from step S112 to the user by displaying them on a display or the like.

[0061] (Step S114) In step S114, the training data generation unit 16A stores training data in the storage unit 20A, which includes data indicating the variable factors received in step S111, distribution performance data, and data indicating the period from the start point to the end point specified by the user. In addition to the above data, the training data may also include information indicating product attributes (channels, etc.), information indicating product categories, information related to product marketing, information indicating the external environment, etc.

[0062] The training data stored in the memory unit 20A is used for machine learning of the machine learning model LM by the training unit 17A. The machine learning model LM trained by the training unit 17A is used, for example, for demand forecasting. For instance, the demand forecasting unit 15A may perform demand forecasting for a product based on output data obtained by inputting input data, which includes actual values ​​of the volume of goods in circulation and information indicating the factors causing fluctuations, into the machine learning model LM.

[0063] (Responding to changes in levels) As described above, when the user selects "Confirm Correction," the demand forecasting unit 15A corrects the actual values ​​for the period determined by the start and end points specified by the user, and uses the corrected actual values ​​to forecast the demand for the product. This makes it possible to perform demand forecasting that takes into account the impact of temporary demand fluctuations. On the other hand, there are cases where demand fluctuations are not temporary but continuous. For example, if the fluctuation factor is "out of stock," "limited edition," or "promotion," the period of influence of the fluctuation factor is temporary, whereas if the fluctuation factor is "new adoption" or "out of stock," the period of influence of the fluctuation factor is considered to be continuous. For such non-temporary demand fluctuations, the user may specify the change point or rate of change. In this case, for non-temporary demand fluctuations, the user may input the rate of change in the demand level, and the demand forecasting unit 15A may adjust only the level while maintaining the trend and seasonality in the demand forecast. In other words, if the fluctuation factor received by the receiving unit 12A is the cause of continuous demand fluctuations, the demand forecasting unit 15A can also forecast the demand for the product by changing the level of the actual values ​​according to the fluctuation factor received by the receiving unit 12A.

[0064] In this case, the training data generation unit 16A generates training data by associating the first data with the second data when the variable factor received by the reception unit 12A is a variable factor that causes a temporary fluctuation in demand ("out of stock," "limited edition," "promotion," etc.). On the other hand, the training data generation unit 16A does not generate training data when the variable factor received by the reception unit 12A is a variable factor that causes a temporary fluctuation in demand ("newly adopted," "out of stock," etc.).

[0065] (Effects of information processing equipment) As explained above, the information processing device 1A not only detects anomalies from actual transaction volume data, but also proposes correction values ​​for the detected anomalies to the user. This allows the user of the information processing device 1A to decide how to correct (or not correct) the actual transaction data used for demand forecasting, by referring to the proposed correction values. In other words, with the information processing device 1A, the user can easily decide how to correct (or not correct) the actual transaction data by referring to the proposed correction values.

[0066] Furthermore, instead of uniformly reflecting the proposed corrections, the information processing device 1A presents the user with three options: a first option to not correct the actual values, a second option to provisionally correct the actual values ​​with the corrected values, and a third option to finalize the correction of the actual values. This allows the user of the information processing device 1A to review the proposed corrections and choose whether to not make any corrections, make provisional corrections, or finalize the corrections, enabling the user's intentions to be reflected in the demand forecast with simple operation.

[0067] Furthermore, the information processing device 1A is configured to include a demand forecasting unit 15A that, when the user selects a first option, predicts the demand for the product using actual values ​​prior to the first period without correcting the actual values ​​for the first period. Therefore, the information processing device 1A can perform demand forecasting that reflects the user's intentions.

[0068] Furthermore, the information processing device 1A is configured to include a demand forecasting unit 15A that, when the user selects a second option, corrects the actual values ​​for the first period with a correction value and uses the corrected actual values ​​to forecast the demand for the product. Therefore, the information processing device 1A can perform demand forecasting that reflects the user's intentions.

[0069] Furthermore, in the information processing device 1A, if the user selects a third option, the reception unit 12A receives the user's specification of the end point of the period corresponding to the abnormal value detected by the abnormal value detection unit 11A, calculates a correction value for the actual values ​​of the first period from the start point to the end point received by the reception unit 12A, corrects the actual values ​​of the first period using the calculated correction value, and employs a demand forecasting unit 15A that forecasts the demand for the product using the corrected actual values. Therefore, the information processing device 1A can perform demand forecasting that reflects the user's intentions.

[0070] Furthermore, in the information processing device 1A, the reception unit 12A accepts the user's specification of a variable factor when a third option is selected by the user, and employs a configuration in which a training data generation unit 16A generates training data that associates first data, which includes the actual value of the volume of goods distributed and the variable factor specified by the user, with second data, which includes the actual value corrected by the demand forecasting unit 15A. Therefore, the information processing device 1A can generate training data used to train a machine learning model that can reflect the impact of a variable factor in demand forecasting when similar variable factors to those of the past are expected in the future.

[0071] Furthermore, the information processing device 1A employs a configuration that includes a training unit 17A that uses the training data generated by the training data generation unit 16A to train a machine learning model by associating the actual value of the volume of goods distributed and the fluctuation factors specified by the user as input data, and the corrected value of the actual value as output data. Therefore, the information processing device 1A can generate a machine learning model that can reflect the impact of fluctuation factors in demand forecasts when similar fluctuation factors to those of the past are expected in the future. With this machine learning model, for example, a user can simply specify fluctuation factors, and when similar fluctuation factors to those of the past are expected in the future, the impact of those fluctuation factors can be reflected in demand forecasts.

[0072] Furthermore, in the information processing device 1A, the training data generation unit 16A generates training data in which the first data and the second data are associated when the fluctuation factor received by the reception unit 12A is caused by a temporary demand fluctuation, while not generating training data when the fluctuation factor received by the reception unit 12A is caused by a temporary demand fluctuation. By generating training data for machine learning only for fluctuation factors that cause temporary demand fluctuations in this way, it is possible to generate a machine learning model LM with higher prediction accuracy.

[0073] Furthermore, in the information processing device 1A, if the variable factor received by the reception unit 12A is caused by continuous demand fluctuations, the demand forecasting unit 15A is configured to forecast the demand for the product by changing the level of the actual value according to the variable factor received by the reception unit 12A.

[0074] [Examples of implementation using software] Some or all of the functions of the abnormal value correction device 1 and the information processing device 1A (hereinafter also referred to as "each of the above devices") may be implemented by hardware such as integrated circuits (IC chips) or by software.

[0075] In the latter case, each of the above devices is implemented, for example, by a computer that executes instructions for a program, which is software that realizes each function. An example of such a computer (hereinafter referred to as Computer C) is shown in Figure 8. Figure 8 is a block diagram showing the hardware configuration of Computer C, which functions as each of the above devices.

[0076] Computer C comprises at least one processor C1 and at least one memory C2. Memory C2 stores a program P that causes computer C to operate as each of the above-mentioned devices. In computer C, processor C1 reads program P from memory C2 and executes it, thereby realizing each of the above-mentioned devices.

[0077] For processor C1, for example, a CPU (Central Processing Unit), GPU (Graphic Processing Unit), DSP (Digital Signal Processor), MPU (Micro Processing Unit), FPU (Floating Point Number Processing Unit), PPU (Physics Processing Unit), TPU (Tensor Processing Unit), quantum processor, microcontroller, or a combination thereof can be used. For memory C2, for example, flash memory, HDD (Hard Disk Drive), SSD (Solid State Drive), or a combination thereof can be used.

[0078] Computer C may also be equipped with RAM (Random Access Memory) for loading program P at runtime and for temporarily storing various data. Furthermore, computer C may be equipped with communication interfaces for sending and receiving data with other devices. Additionally, computer C may be equipped with input / output interfaces for connecting input / output devices such as keyboards, mice, displays, and printers.

[0079] Furthermore, program P can be recorded on a non-temporary, tangible recording medium M that is readable by computer C. Such a recording medium M could be, for example, tape, disk, card, semiconductor memory, or programmable logic circuitry. Computer C can acquire program P via such a recording medium M. Program P can also be transmitted via a transmission medium. Such a transmission medium could be, for example, a communication network or broadcast waves. Computer C can also acquire program P via such a transmission medium.

[0080] Furthermore, each of the above functions of each of the above devices may be implemented by a single processor in a single computer, by multiple processors in a single computer working together, or by multiple processors in each of multiple computers working together. In addition, the programs for implementing each of the above functions in each of the above devices may be stored in a single memory in a single computer, distributed and stored in multiple memories in a single computer, or distributed and stored in multiple memories in each of multiple computers.

[0081] [Additional Note A] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims. (Note A1) An anomaly detection means for detecting anomalies from actual values ​​of the volume of goods distributed over a unit period, A receiving means that receives a user's specification of the starting point of a first period corresponding to the abnormal value detected by the abnormal value detection means, A correction proposal means calculates a correction value for the actual values ​​of the first period based on the actual values ​​of the second period prior to the starting point received by the reception means, and proposes the calculated correction value to the user. A choice-presenting means that presents the user with a first option of not correcting the actual values ​​for the first period, a second option of provisionally correcting the actual values ​​for the first period with the correction value, and a third option of finalizing the correction of the actual values ​​for the first period. An abnormal value correction device equipped with the following features.

[0082] (Appendix A2) If the user selects the first option, a forecasting means predicts the demand for the product using actual values ​​prior to the first period without correcting the actual values ​​for the first period. An abnormal value correction device as described in Appendix A1, further comprising the above.

[0083] (Note A3) If the user selects the second option, a second forecasting means corrects the actual values ​​for the first period with the correction value and uses the corrected actual values ​​to forecast the demand for the product. An abnormal value correction device as described in Appendix A1 or A2, further comprising the above.

[0084] (Note A4) If the user selects the third option, the receiving means accepts the user's specification of the end point of the period corresponding to the abnormal value detected by the abnormal value detection means. The system further comprises a third forecasting means that calculates a correction value for the actual values ​​for the first period from the starting point to the ending point received by the receiving means, corrects the actual values ​​for the first period using the calculated correction value, and forecasts the demand for the product using the corrected actual values. An abnormal value correction device as described in any one of the appendices A1 to A3.

[0085] (Note A5) The receiving means, when the user selects the third option, receives the user's specification of the variable factor. The system further comprises a training data generation means that generates training data in which first data including actual values ​​of the distribution volume of the said product and the fluctuation factors specified by the user and second data including actual values ​​corrected by the third prediction means are associated. An abnormal value correction device as described in Appendix A4.

[0086] (Note A6) The system further comprises a training means that uses the training data generated by the training data generation means to train a machine learning model by associating the actual value of the distribution volume of the product and the fluctuation factors specified by the user as input data, and a corrected value of the actual value as output data. An abnormal value correction device as described in Appendix A5.

[0087] (Note A7) The training data generation means generates training data in which the first data and the second data are associated when the fluctuation factor received by the receiving means is caused by a temporary demand fluctuation, but does not generate the training data when the fluctuation factor received by the receiving means is caused by a temporary demand fluctuation. An abnormal value correction device as described in Appendix A5 or A6.

[0088] (Note A8) If the fluctuation factor received by the receiving means is the cause of continuous demand fluctuations, the third forecasting means predicts the demand for the product by changing the level of the actual value in accordance with the fluctuation factor received by the receiving means. An abnormal value correction device as described in one of the appendices A5 to A7.

[0089] [Additional Note B] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims. (Note B1) At least one processor performs an anomaly detection process that detects anomalies from the actual values ​​of the volume of goods distributed per unit period, The at least one processor includes an acceptance process that accepts a user's specification of the starting point of a first period corresponding to the abnormal value detected in the abnormal value detection process, The at least one processor performs a correction proposal process in which it calculates a correction value for the actual value of the first period based on the actual value of the second period prior to the starting point received in the reception process, and proposes the calculated correction value to the user. The at least one processor performs a choice presentation process that presents the user with a first option of not correcting the actual values ​​for the first period, a second option of provisionally correcting the actual values ​​for the first period with the correction value, and a third option of finalizing the correction of the actual values ​​for the first period. An outlier correction method that includes this.

[0090] (Note B2) If the user selects the first option, the at least one processor performs a forecasting process to predict the demand for the product using actual values ​​prior to the first period, without correcting the actual values ​​for the first period. The abnormal value correction method described in Appendix B1, which further includes the above.

[0091] (Note B3) If the user selects the second option, the at least one processor performs a second forecasting process which corrects the actual values ​​for the first period with the correction value and uses the corrected actual values ​​to forecast the demand for the product. An abnormal value correction method described in Appendix B1 or B2, further including the above.

[0092] (Note B4) If the user selects the third option, the at least one processor, in the reception process, accepts the user's specification of the end point of the period corresponding to the abnormal value detected in the abnormal value detection process. The at least one processor further includes a third forecasting process which calculates a correction value for the actual values ​​for the first period from the starting point to the ending point received in the reception process, corrects the actual values ​​for the first period using the calculated correction value, and forecasts the demand for the product using the corrected actual values. The method for correcting outliers described in one of the appendices B1 to B3.

[0093] (Note B5) In the aforementioned reception process, if the user selects the third option, the at least one processor accepts the user's specification of the variable factor. The at least one processor further includes a training data generation process that generates training data in which first data, including actual values ​​of the volume of goods distributed and the fluctuation factors specified by the user, and second data, including actual values ​​corrected by the third prediction process, are associated. The method for correcting outliers described in Appendix B4.

[0094] (Note B6) The at least one processor further includes a training process that uses the training data generated in the training data generation process to train a machine learning model by associating the actual value of the volume of goods distributed and the fluctuation factors specified by the user as input data, and a corrected value of the actual value as output data. The method for correcting outliers is described in Appendix B5.

[0095] (Note B7) In the training data generation process, the at least one processor generates training data in which the first data and the second data are associated when the variable factor received in the reception process is caused by a temporary demand fluctuation, but does not generate the training data when the variable factor received in the reception process is caused by a temporary demand fluctuation. The abnormal value correction method described in Appendix B5 or B6.

[0096] (Note B8) If the variable factor received in the reception process is caused by continuous demand fluctuations, the at least one processor predicts the demand for the product in the third forecasting process by changing the level of the actual value according to the variable factor received in the reception process. The method for correcting outliers described in one of the appendices B5 through B7.

[0097] [Additional Note C] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims. (Note C1) A program for causing a computer to function as an abnormal value correction device, wherein the computer, An anomaly detection means for detecting anomalies from actual values ​​of the volume of goods distributed over a unit period, A receiving means that receives a user's specification of the starting point of a first period corresponding to the abnormal value detected by the abnormal value detection means, A correction proposal means calculates a correction value for the actual values ​​of the first period based on the actual values ​​of the second period prior to the starting point received by the reception means, and proposes the calculated correction value to the user. A choice-presenting means that presents the user with a first option of not correcting the actual values ​​for the first period, a second option of provisionally correcting the actual values ​​for the first period with the correction value, and a third option of finalizing the correction of the actual values ​​for the first period. An anomaly correction program to enable it to function as such.

[0098] (Note C2) The aforementioned computer, If the user selects the first option, a forecasting means predicts the demand for the product using actual values ​​prior to the first period without correcting the actual values ​​for the first period. An abnormal value correction program described in Appendix C1 further enhances its functionality.

[0099] (Note C3) The aforementioned computer, If the user selects the second option, a second forecasting means corrects the actual values ​​for the first period with the correction value and uses the corrected actual values ​​to forecast the demand for the product. An abnormal value correction program described in Appendix C1 or C2 to further enhance its functionality.

[0100] (Note C4) If the user selects the third option, the receiving means accepts the user's specification of the end point of the period corresponding to the abnormal value detected by the abnormal value detection means. The aforementioned computer, The receiving means calculates a correction value for the actual values ​​for the first period from the starting point to the ending point received by the receiving means, corrects the actual values ​​for the first period using the calculated correction value, and further functions as a third forecasting means that forecasts the demand for the product using the corrected actual values. An abnormal value correction program described in one of the appendices C1 to C3.

[0101] (Note C5) The receiving means, when the user selects the third option, receives the user's specification of the variable factor. The aforementioned computer, The training data generation means further functions as a means for generating training data in which first data, which includes the actual value of the distribution volume of the said product and the fluctuation factors specified by the user, and second data, which includes the actual value corrected by the third prediction means, are associated. The abnormal value correction program described in Appendix C4.

[0102] (Appendix C6) The aforementioned computer, The training means further functions as a training means that uses the training data generated by the training data generation means to train a machine learning model by associating the actual value of the distribution volume of the product and the fluctuation factors specified by the user as input data, and the corrected value of the actual value as output data. The abnormal value correction program described in Appendix C5.

[0103] (Note C7) The training data generation means generates training data in which the first data and the second data are associated when the fluctuation factor received by the receiving means is caused by a temporary demand fluctuation, but does not generate the training data when the fluctuation factor received by the receiving means is caused by a temporary demand fluctuation. An abnormal value correction program as described in Appendix C5 or C6.

[0104] (Note C8) If the fluctuation factor received by the receiving means is the cause of continuous demand fluctuations, the third forecasting means predicts the demand for the product by changing the level of the actual value in accordance with the fluctuation factor received by the receiving means. An abnormal value correction program described in one of the appendices C5 to C7.

[0105] [Additional Note D] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims. (Note D1) It comprises at least one processor, and the at least one processor is An anomaly detection process that detects outliers from actual values ​​of the volume of goods distributed over a unit period, A reception process that accepts a user's specification of the starting point of a first period corresponding to the abnormal value detected in the abnormal value detection process, Based on the actual values ​​for the second period prior to the starting point received in the aforementioned reception process, a correction proposal process is performed to calculate a correction value for the actual values ​​for the first period and propose the calculated correction value to the user. A choice presentation process that presents the user with a first option of not correcting the actual values ​​for the first period, a second option of provisionally correcting the actual values ​​for the first period with the correction value, and a third option of finalizing the correction of the actual values ​​for the first period. An abnormal value correction device that performs this function.

[0106] The abnormal value correction device may also include a memory. Furthermore, the memory may store a program for causing at least one processor to perform each of the aforementioned processes.

[0107] (Note D2) The aforementioned at least one processor, If the user selects the first option, a forecasting process is performed to predict the demand for the product using actual values ​​prior to the first period, without correcting the actual values ​​for the first period. An abnormal value correction device described in Appendix D1 further performs the following.

[0108] (Note D3) The aforementioned at least one processor, If the user selects the second option, a second forecasting process is performed to correct the actual values ​​for the first period with the correction value and to forecast the demand for the product using the corrected actual values. An abnormal value correction device described in Appendix D1 or D2 that further performs the above.

[0109] (Note D4) If the user selects the third option, the reception process accepts the user's specification of the end point of the period corresponding to the abnormal value detected in the abnormal value detection process. The aforementioned at least one processor, A third forecasting process is further executed, which involves calculating a correction value for the actual values ​​for the first period from the starting point to the ending point received in the reception process, correcting the actual values ​​for the first period using the calculated correction value, and forecasting the demand for the product using the corrected actual values. An abnormal value correction device as described in any one of the appendices D1 to D3.

[0110] (Note D5) In the aforementioned reception process, if the user selects the third option, the at least one processor accepts the user's specification of the variable factor. The aforementioned at least one processor, Further, a training data generation process is performed to generate training data in which first data, which includes the actual value of the distribution volume of the said product and the fluctuation factors specified by the said user, and second data, which includes the actual value corrected by the third prediction process, are associated. An abnormal value correction device as described in Appendix D4.

[0111] (Note D6) The aforementioned at least one processor, Using the training data generated in the aforementioned training data generation process, a further training process is performed in which the actual value of the product's circulation volume and the fluctuation factors specified by the user are used as input data, and the corrected value of the actual value is used as output data, and the machine learning model is trained by relating the two. An abnormal value correction device as described in Appendix D5.

[0112] (Note D7) In the training data generation process, the at least one processor generates training data in which the first data and the second data are associated when the variable factor received in the reception process is caused by a temporary demand fluctuation, but does not generate the training data when the variable factor received in the reception process is caused by a temporary demand fluctuation. An abnormal value correction device as described in Appendix D5 or D6.

[0113] (Note D8) If the variable factors received in the reception process are caused by continuous demand fluctuations, the third forecasting process predicts the demand for the product by changing the level of the actual value in accordance with the variable factors received in the reception process. An abnormal value correction device as described in any one of the appendices D5 to D7.

[0114] [Additional Note E] This disclosure includes the technologies described in the following appendices. However, the present invention is not limited to the technologies described in the following appendices, and various modifications are possible within the scope of the claims. (Note E1) A program for causing a computer to function as an abnormal value correction device, wherein the computer, An anomaly detection process that detects outliers from actual values ​​of the volume of goods distributed over a unit period, A reception process that accepts a user's specification of the starting point of a first period corresponding to the abnormal value detected in the abnormal value detection process, Based on the actual values ​​for the second period prior to the starting point received in the aforementioned reception process, a correction proposal process is performed to calculate a correction value for the actual values ​​for the first period and propose the calculated correction value to the user. A choice presentation process that presents the user with a first option of not correcting the actual values ​​for the first period, a second option of provisionally correcting the actual values ​​for the first period with the correction value, and a third option of finalizing the correction of the actual values ​​for the first period. A non-temporary recording medium containing a program for correcting abnormal values ​​to execute the program. [Explanation of Symbols]

[0115] 1. Anomaly Correction Device 1A Information Processing Device 2A User Terminal 11, 11A Abnormal Value Detection Unit 12, 12A Reception Desk 13, 13A Correction Proposal Section 14, 14A Option Presentation Section 15A Demand Forecasting Department 16A Training Data Generation Unit 18A Display Control Unit 17A Training Department S11 Anomaly detection process S12 Reception Processing S13 Correction Proposal Processing S14 Option presentation process

Claims

1. An anomaly detection means for detecting anomalies from actual values ​​of the volume of goods distributed over a unit period, A receiving means that receives a user's specification of the starting point of a first period corresponding to the abnormal value detected by the abnormal value detection means, A correction proposal means calculates a correction value for the actual values ​​of the first period based on the actual values ​​of the second period prior to the starting point received by the reception means, and proposes the calculated correction value to the user. A choice-presenting means that presents the user with a first option of not correcting the actual values ​​for the first period, a second option of provisionally correcting the actual values ​​for the first period with the correction value, and a third option of finalizing the correction of the actual values ​​for the first period. An abnormal value correction device equipped with the following features.

2. If the user selects the first option, a forecasting means predicts the demand for the product using actual values ​​prior to the first period without correcting the actual values ​​for the first period. The abnormal value correction device according to claim 1, further comprising:

3. If the user selects the second option, a second forecasting means corrects the actual values ​​for the first period with the correction value and uses the corrected actual values ​​to forecast the demand for the product. An abnormal value correction device according to claim 1 or 2, further comprising the following:

4. If the user selects the third option, the receiving means accepts the user's specification of the end point of the period corresponding to the abnormal value detected by the abnormal value detection means. The system further comprises a third forecasting means that calculates a correction value for the actual values ​​for the first period from the starting point to the ending point received by the receiving means, corrects the actual values ​​for the first period using the calculated correction value, and forecasts the demand for the product using the corrected actual values. An abnormal value correction device according to claim 1 or 2.

5. The receiving means, when the user selects the third option, receives the user's specification of the variable factor. The system further comprises a training data generation means that generates training data in which first data, which includes actual values ​​of the distribution volume of the said product and the fluctuation factors specified by the user, and second data, which includes actual values ​​corrected by the third prediction means, are associated. The abnormal value correction device according to claim 4.

6. The system further comprises a training means that uses the training data generated by the training data generation means to train a machine learning model by associating the actual value of the distribution volume of the product and the fluctuation factors specified by the user as input data, and a corrected value of the actual value as output data. The abnormal value correction device according to claim 5.

7. The training data generation means generates training data in which the first data and the second data are associated when the fluctuation factor received by the receiving means is caused by a temporary demand fluctuation, but does not generate the training data when the fluctuation factor received by the receiving means is caused by a temporary demand fluctuation. The abnormal value correction device according to claim 5.

8. If the fluctuation factor received by the receiving means is the cause of continuous demand fluctuations, the third forecasting means predicts the demand for the product by changing the level of the actual value in accordance with the fluctuation factor received by the receiving means. The abnormal value correction device according to claim 5.

9. At least one processor performs an anomaly detection process that detects anomalies from the actual values ​​of the volume of goods distributed per unit period, The at least one processor includes an acceptance process that accepts a user's specification of the starting point of a first period corresponding to the abnormal value detected in the abnormal value detection process, The at least one processor performs a correction proposal process in which it calculates a correction value for the actual value of the first period based on the actual value of the second period prior to the starting point received in the reception process, and proposes the calculated correction value to the user. The at least one processor performs a choice presentation process that presents the user with a first option of not correcting the actual values ​​for the first period, a second option of provisionally correcting the actual values ​​for the first period with the correction value, and a third option of finalizing the correction of the actual values ​​for the first period. An outlier correction method that includes this.

10. An abnormal value correction program for causing a computer to function as an abnormal value correction device, wherein the computer, An anomaly detection means for detecting anomalies from actual values ​​of the volume of goods distributed over a unit period, A receiving means that receives a user's specification of the starting point of a first period corresponding to the abnormal value detected by the abnormal value detection means, A correction proposal means calculates a correction value for the actual values ​​of the first period based on the actual values ​​of the second period prior to the starting point received by the reception means, and presents the calculated correction value to the user. A choice-presenting means that presents the user with a first option of not correcting the actual values ​​for the first period, a second option of provisionally correcting the actual values ​​for the first period with the correction value, and a third option of finalizing the correction of the actual values ​​for the first period. An anomaly correction program to enable it to function as such.

Images