Information processing device, predication model, information processing method, and program

Abstract

The purpose of the present disclosure is to provide an information processing device with which it is possible to appropriately detect abnormality included in commodity handling histories. An information processing device according to the present disclosure comprises: an acquisition unit that acquires the history of handling of a commodity of interest that indicates the history of handling of the commodity of interest that is to be monitored; a prediction unit that predicts a quantity after handling in relation to the handling included in the history of handling of the commodity of interest and calculates a predicted quantity of the commodity of interest on the basis of the history of handling of the commodity of interest, by using a prediction model having been trained using production information regarding production of a commodity, handing content information that indicates the handling content of the commodity, and quantity information that indicates quantities before and after handling of the commodity, as training data; and an abnormality determination unit that determines, on the basis of the difference between the quantity of the commodity of interest after handling in the history of handling of the commodity of interest and the predicted quantity of the commodity of interest calculated by the prediction unit, whether or not there is abnormality in the quantity change indicated by the history of handling of the commodity of interest.

Description

Information processing device, prediction model, information processing method and program

[0001] The present disclosure relates to an information processing device, a prediction model, an information processing method, and a program.

[0002] There is known technology related to traceability systems for managing goods such as food. A traceability system is a system in which multiple businesses record the handling history of goods and multiple consumers track the goods. For example, when using a traceability system to track the handling history of food recorded by businesses, users of the system can detect food fraud by comparing the food label with the handling history.

[0003] As a related technology, Patent Document 1 discloses an item processing management system that includes a distributed storage system in which each of a plurality of computing devices serves as a node, a tracking information creation unit that creates tracking information, and a storage request unit that outputs a storage request for the tracking information.

[0004] The item processing management system disclosed in Patent Document 1 manages the traceability of items, targeting falsification of handling history related to transactions between businesses. For example, the item processing management system detects handling history with falsified weight by checking whether the weight of an item in a transaction between businesses has increased or decreased beyond a threshold. The item processing management system also manages handling history in a distributed manner on a blockchain, which is a database that can detect and correct tampering of handling history. This prevents tampering of handling history once it has been registered.

[0005] Patent No. 6785701

[0006] Handling histories of goods include histories related to the transactions of goods between businesses, which is the subject of Patent Document 1, as well as histories related to the processing of goods within a business. Examples of falsification of handling histories related to the processing of goods within a business include falsification of values ​​such as weight in processing where the amount of food changes due to waste, and falsification of qualitative data such as place of origin and brand in production processing for registering place of origin, etc. The technology disclosed in Patent Document 1 targets falsification of handling histories related to transactions between businesses, and does not mention such falsification within a business.

[0007] In view of the above-mentioned problems, an object of the present disclosure is to provide an information processing device, an information processing method, and a program that are capable of appropriately detecting anomalies contained in product handling histories.

[0008] The information processing device according to the present disclosure comprises: an acquisition unit that acquires a target product handling history indicating the handling history of a target product that is the subject of an audit; a prediction unit that calculates a target product predicted quantity by predicting the quantity after handling for the handling included in the target product handling history using a prediction model trained using, as learning data, production information regarding the production of the product, handling content information indicating the handling content of the product, and quantity information indicating the quantity of the product before and after handling, based on the target product handling history; and an abnormality determination unit that determines whether there is an abnormality in the quantity change indicated by the target product handling history based on the difference between the quantity of the target product after handling in the target product handling history and the predicted target product quantity calculated by the prediction unit.

[0009] The information processing method disclosed herein comprises an acquisition step of acquiring a target product handling history indicating the handling history of a target product that is the subject of an audit; a prediction step of calculating a target product predicted quantity by predicting the quantity after handling for the handling included in the target product handling history using a prediction model trained using production information regarding the production of the product, handling content information indicating the handling content of the product, and quantity information indicating the quantity of the product before and after handling based on the target product handling history; and an abnormality determination step of determining whether there is an abnormality in the quantity change indicated by the target product handling history based on the difference between the quantity of the target product after handling in the target product handling history and the predicted target product quantity calculated in the prediction step.

[0010] The program disclosed herein causes a computer to execute the following steps: an acquisition step of acquiring a target product handling history indicating the handling history of a target product that is the subject of an audit; a prediction step of calculating a target product predicted quantity by predicting the quantity after handling for the handling included in the target product handling history using a prediction model trained using production information regarding the production of the product, handling content information indicating the handling content of the product, and quantity information indicating the quantity of the product before and after handling, based on the target product handling history; and an abnormality determination step of determining whether there is an abnormality in the quantity change indicated by the target product handling history, based on the difference between the quantity of the target product after handling in the target product handling history and the predicted target product quantity calculated in the prediction step.

[0011] The information processing device, information processing method, and program according to the present disclosure can appropriately detect anomalies contained in the handling history of a product.

[0012] FIG. 1 is a block diagram showing a configuration of an information processing device according to the present disclosure. FIG. 2 is a flowchart showing processing performed by an information processing device according to the present disclosure. FIG. 3 is a block diagram showing a configuration of an information processing system according to the present disclosure. FIG. 4 is a block diagram showing a configuration of a handling history shaping unit according to the present disclosure. FIG. 5 is a diagram showing input / output data in a change amount learning data generation unit according to the present disclosure. FIG. 6 is a diagram explaining a method for generating change amount learning data from handling histories according to the present disclosure. FIG. 7 is a diagram showing input / output data in a food feature vector generation unit according to the present disclosure. FIG. 8 is a diagram showing input / output data in a prediction model learning unit according to the present disclosure. FIG. 9 is a flowchart showing learning processing performed by an information processing system according to the present disclosure. FIG. 10 is a block diagram showing a configuration of an anomaly detection unit according to the present disclosure. FIG. 11 is a diagram showing input / output data in an explanatory variable collection unit according to the present disclosure. FIG. 12 is a diagram explaining a method for generating collected data from handling histories according to the present disclosure. FIG. 13 is a diagram showing input / output data in a change amount prediction unit according to the present disclosure. FIG. 14 is a diagram showing input / output data in a prediction error threshold determination unit according to the present disclosure. FIG. 15 is a flowchart showing processing performed by an anomaly detection unit according to the present disclosure. FIG. 16 is a block diagram showing a configuration of an information processing system according to the present disclosure. FIG. 17 is a diagram showing input / output data in a production information combination creation unit according to the present disclosure. FIG. 18 is a block diagram showing a configuration of an anomaly detection unit according to the present disclosure. FIG. 19 is a diagram showing input / output data in a production information combination comprehensive unit according to the present disclosure. FIG. 19 is a diagram showing input / output data in a handling history consistency determination unit according to the present disclosure. FIG. 19 is a flowchart showing processing performed by an anomaly detection unit according to FIG. 1 is a flowchart showing processing performed by an anomaly detection unit according to the present disclosure. FIG. 2 is a block diagram showing the configuration of an information processing system according to the present disclosure. FIG. 3 is a diagram showing input data in a production information grouping unit according to the present disclosure. FIG. 4 is a diagram showing output data in a production information grouping unit according to the present disclosure. FIG. 4 is a block diagram showing the configuration of a handling history shaping unit according to the present disclosure. FIG. 5 is a diagram showing input / output data in a food feature vector generation unit according to the present disclosure. FIG. 5 is a flowchart showing processing performed by an information processing system according to the present disclosure. FIG. 6 is a flowchart showing processing performed by a production information grouping unit according to the present disclosure. FIG. 7 is a diagram showing input data in a production information association unit according to the present disclosure.FIG. 1 is a diagram showing output data in a production information associating unit according to the present disclosure. FIG. 2 is a flowchart showing classification processing performed by an information processing system according to the present disclosure. FIG. 3 is a block diagram showing the configuration of an abnormality determination unit according to the present disclosure. FIG. 4 is a flowchart showing processing performed by an abnormality determination unit according to the present disclosure. FIG. 5 is a block diagram showing the configuration of an information processing system according to the present disclosure. FIG. 6 is a diagram explaining characteristics of handling history stored in a blockchain according to the present disclosure. FIG. 7 is a diagram showing input / output data in a handling history food categorization unit according to the present disclosure. FIG. 8 is a flowchart showing processing by a handling history food categorization unit according to the present disclosure. FIG. 9 is a block diagram illustrating an example hardware configuration of a computer that realizes an information processing system, etc. according to the present disclosure.

[0013] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In each drawing, the same or corresponding elements are designated by the same reference numerals. For clarity of explanation, duplicated explanations will be omitted as necessary.

[0014] First Embodiment First, a description will be given of the first embodiment. Fig. 1 is a block diagram showing the configuration of an information processing device 100 according to the present disclosure. The information processing device 100 includes an acquisition unit 101, a prediction unit 102, and an abnormality determination unit 103.

[0015] The acquisition unit 101 acquires a target product handling history indicating a history of handling of a target product that is the target of an audit. The prediction unit 102 predicts the post-handling volume for the handling included in the target product handling history using a prediction model based on the target product handling history, and calculates a target product predicted volume.

[0016] The prediction model is trained using, as training data, production information related to the production of the product, handling content information indicating the handling content of the product, and quantity information indicating the quantity of the product before and after handling. The abnormality determination unit 103 determines whether there is an abnormality in the quantity change indicated by the target product handling history based on the difference between the quantity of the target product after handling in the target product handling history and the target product predicted quantity calculated by the prediction unit.

[0017] The information processing device 100 includes a processor, a memory, and a storage device (not shown). The storage device stores a computer program that implements the processing described herein. The processor can load the computer program from the storage device into the memory and execute the computer program. In this way, the processor realizes the functions of an acquisition unit 101, a prediction unit 102, and an abnormality determination unit 103.

[0018] Next, the process performed by the information processing device 100 will be described with reference to Fig. 2. Fig. 2 is a flowchart showing the process performed by the information processing device 100.

[0019] First, the acquisition unit 101 acquires the target product handling history (S1). Next, the prediction unit 102 calculates the target product predicted quantity using a prediction model (S2). Then, the anomaly determination unit 103 determines whether there is an anomaly in the quantity change indicated by the target product handling history based on the difference between the target product's post-handling quantity and the predicted target product quantity.

[0020] With this configuration, the information processing device 100 can appropriately detect abnormalities contained in the handling history of a product.

[0021] Second Embodiment Next, a second embodiment will be described. The second embodiment is a specific example of the first embodiment. First, an overview of the information processing system 10 according to the present disclosure will be described.

[0022] The information processing system 10 is an example of the above-mentioned information processing device 100. The information processing system 10 is an information processing system that can determine whether a change in the amount of goods is normal or abnormal based on the deviation between a predicted value of the amount of goods after handling and an input value, in order to detect counterfeit goods.

[0023] The quantity of a product indicates a quantitative characteristic of the product. The quantity of a product can be expressed using quantitative data such as the weight, volume, length, number, or apparent area of ​​the product. Various quantitative data representing the characteristics of a product may be used as the quantity of a product, without being limited to these. In this embodiment, the weight of a product will be used as an example of the quantity of a product. The information processing system 10 acquires weight information indicating the quantity of a product before and after handling, and performs predetermined processing using the weight information.

[0024] In this disclosure, "handling" refers to various procedures related to the product being tracked. For example, if the product is livestock, examples of handling include birth registration, processing, or trading. Handling may be a process performed between businesses or within a business.

[0025] In this embodiment, a learning process is performed to create a change prediction model that predicts the weight of a product after handling, and the change prediction model is used to determine whether the weight change indicated by the product's handling history is normal or abnormal. The handling history is information about the product's handling history. The change prediction model can be trained using, for example, information about at least one of the product's place of origin, brand, and manufacturer as production information. Note that this embodiment assumes that a sufficient amount of handling history has been accumulated for each place of origin, brand, or manufacturer to train the change prediction model, and that no new types of food are added.

[0026] The information processing system 10 acquires past handling history from the blockchain, collects values ​​such as the handling content and weight after handling from the handling history related to processing within the business, values ​​such as weight before handling from the handling history added immediately before the handling history, and production information such as the place of origin, brand, and production business from the handling history in production processing, and generates change amount learning data.

[0027] Here, production information is information related to the quality or properties of a product. The production information can be expressed using qualitative data that represents the quality or properties of a product. The production information is, for example, information indicating the product's place of origin, brand, production company, producer, gender, rearing method, farming method, feed, grade, award history, certification, or authentication. Not limited to these, various qualitative data related to the quality or properties of a product can be used as production information. The production information is registered, for example, during production processing within a company. The production information may also be registered during processing after the production processing.

[0028] The information processing system 10 uses a change amount prediction model generated based on the change amount learning data to predict the weight after handling in the handling history, and uses the prediction result to detect abnormalities in the weight change indicated by the handling history. This makes it possible to automatically detect weight falsification during processing within a business.

[0029] The information processing system 10 collects information on the handling history of products from, for example, production companies that perform production processing of the products, and processing companies that perform subsequent processing. Products are items that can be tracked, such as food and medicine. Products can also include livestock. In this embodiment, explanations will be given using hogs as an example of a product. Note that in this disclosure, the tracking target will be referred to as a "product" in the various handling procedures that take place from the time of production until the product reaches the final consumer.

[0030] The handling history is, for example, information that associates a product, handling details, the weight of the product after handling, the handling business, and the handling date and time. The handling history is not limited to these, and may include various other information related to the product.

[0031] For example, a pig producer registers information about the birth of a pig in the information processing system 10. A processing company that processes the pigs also registers handling information related to the processing. The information processing system 10 accumulates the handling history and manages the handling history. This allows the information processing system 10 to manage the traceability of products.

[0032] Furthermore, a user of the information processing system 10 can track the tracking target product by tracing the handling history. The user of the information processing system 10 is, for example, a manufacturer, a processor, an auditor, or an end user (final consumer).

[0033] (Configuration of Information Processing System 10) The configuration of the information processing system 10 will be described with reference to Fig. 3. Fig. 3 is a block diagram showing the configuration of the information processing system 10.

[0034] The information processing system 10 includes a handling history shaping unit 11, a prediction model learning unit 12, a prediction model storage unit 13, an anomaly detection unit 14, a handling history creation unit 15, a handling history distributed management unit 16, a handling history storage unit 17, and a handling history tracking unit 18. In the figure, the left side of the dashed dotted line indicates the learning processing function, and the right side of the dashed dotted line indicates the traceability system.

[0035] In this embodiment, the information processing system 10 is described as having both a learning processing function and a traceability system function. The configuration of the information processing system 10 may be changed as appropriate. For example, the information processing system 10 may be configured to have only the traceability system function, and another information processing device may have the learning processing function.

[0036] The handling history formatting unit 11 formats information related to handling histories. The prediction model learning unit 12 trains a change amount prediction model. The prediction model storage unit 13 stores the change amount prediction model. The anomaly detection unit 14 detects anomalies contained in the handling history. The handling history creation unit 15 creates the handling history. The handling history distributed management unit 16 manages the handling histories in a distributed manner. The handling history storage unit 17 stores the handling histories. The handling history tracking unit 18 tracks the handling history according to user input. Details of the main functional units are explained below.

[0037] 4 is a block diagram showing the configuration of the handling history shaping unit 11. The handling history shaping unit 11 includes a variation amount learning data generation unit 111 and a food feature vector generation unit 112.

[0038] The variation learning data generation unit 111 uses input data acquired from the handling history storage unit 17 and outputs data to the food feature vector generation unit 112 and the prediction model learning unit 12 .

[0039] 5 is a diagram showing input / output data in the variation learning data generation unit 111. As shown in the input data, the handling history storage unit 17 stores handling histories P1, P2, ..., P10 in chronological order. Each of the handling histories P1, P2, ..., P10 indicates a handling history registered by a business operator. In the example shown in the figure, "birth registration" shown in handling histories P1 and P2 indicates information registered at the time of the pig's birth.

[0040] Furthermore, the "transactions" shown in the handling histories P3 and P9 indicate information registered during transactions between businesses. The "processing" shown in the handling history P10 indicates information registered during processing such as processing and sales within a business. This also applies to the following figures.

[0041] The handling history related to birth registration includes, for example, information indicating the food ID, place of origin, brand, sex, producer name, and weight at time of birth registration. The food ID is identification information for identifying the food. The place of origin, brand, sex, producer name, and weight at time of birth registration are examples of production information. For example, in handling history P1, the food ID is "001", the place of origin is "Kagawa Prefecture", the brand is "Brand B1", the gender is "Male", the producer name is "A Production", and the weight at time of birth registration is "3 kg".

[0042] Furthermore, the handling history for transactions between businesses includes, for example, information indicating the food product ID, transaction details, business name, and weight after handling. In the example of handling history P3, the food product ID is "001," the transaction details are "shipment," the business name is "Production A," and the weight after handling is "100 kg."

[0043] The handling history relating to processing within a business includes, for example, information indicating the food product ID, processing details, business name, and weight after handling. In the example of handling history P10, the food product ID is "001," the transaction details are "sale," the business name is "Sales A," and the weight after handling is "50 kg." The handling history storage unit 17 stores handling history in chronological order, so information on different products can be registered consecutively, as shown in handling histories P1 and P2, for example.

[0044] As shown in the output data, the change amount learning data generation unit 111 outputs explanatory variables 1001 and dependent variables 1002. The explanatory variables 1001 may include information on the place of origin, brand, sex, producer, handling details, and weight before handling. The dependent variable 1002 includes information on weight after handling.

[0045] The internal processing of the change amount learning data generation unit 111 will be described with reference to Fig. 6. Fig. 6 is a diagram illustrating a method for generating change amount learning data from handling histories. As an example, it is assumed here that handling histories P11, P12, ..., P20 are stored in the handling history storage unit 17. The handling histories P11, P12, ..., P20 indicate handling histories related to a product with food ID "001".

[0046] First, the change amount learning data generation unit 111 acquires handling histories P11, P12, ..., P20 stored in chronological order from the handling history storage unit 17. Here, assume that attention is paid to handling history P14. The change amount learning data generation unit 111 acquires the handling content "dismantling" and the weight after handling of "70 kg" from the handling history of the processing process of interest.

[0047] The variation learning data generation unit 111 acquires the place of origin, brand, gender, etc. from the handling history P11 of the birth registration that has the same food ID as the handling history of the processing process of interest. The variation learning data generation unit 111 references the handling history P13 that has the same food ID and was registered immediately before the handling history P14 of interest. The variation learning data generation unit 111 acquires the weight "100 kg" in the handling history P13 as the pre-handling weight in the handling history P14. In this way, the variation learning data generation unit 111 collects data from the handling history storage unit 17 and acquires the collected data 1003. The variation learning data generation unit 111 repeats the collection process of the production information, handling details, post-handling weight, and pre-handling weight as many times as the number of handling histories related to processing within the business. In this way, the variation learning data generation unit 111 generates variation learning data.

[0048] 5 to the food feature vector generation unit 112. The variation learning data generation unit 111 also transmits the objective variable 1002 to the prediction model learning unit 12.

[0049] The internal processing of the food feature vector generation unit 112 will be described with reference to Fig. 7. Fig. 7 is a diagram showing input and output data in the food feature vector generation unit 112.

[0050] First, the food feature vector generation unit 112 obtains the explanatory variables 1001 from the variation learning data generation unit 111. Next, the food feature vector generation unit 112 performs preprocessing such as one-hot encoding on the qualitative variables included in the explanatory variables 1001. Next, the food feature vector generation unit 112 performs preprocessing such as standardization on the quantitative variables included in the explanatory variables 1001. The food feature vector generation unit 112 transmits the generated food feature vector 1004 to the prediction model learning unit 12.

[0051] The internal processing of the prediction model learning unit 12 will be described with reference to Fig. 8. Fig. 8 is a diagram showing input and output data in the prediction model learning unit 12.

[0052] The prediction model learning unit 12 acquires the food feature vector 1004 from the food feature vector generation unit 112, and acquires the dependent variable 1002 from the change amount learning data generation unit 111. The prediction model learning unit 12 performs multiple regression analysis using a pair of the food feature vector 1004 and the dependent variable 1002. As a result, the prediction model learning unit 12 generates a change amount prediction model that predicts the weight after handling.

[0053] In multiple regression analysis, a response variable Y to be predicted can be expressed by the following equation (1) using explanatory variables X1, X2, X3, . . . and partial regression coefficients b1, b2, b3, . . .

[0054] Y=b1X1+b2X2+b3X3+...+b0 (1)

[0055] Here, the objective variable Y is the post-handling weight to be predicted. Furthermore, X1, X2, X3, ... are scores corresponding to each element included in the food feature vector 1004. For example, X1, X2, X3, ... are scores corresponding to "Kagawa Prefecture," "Tokushima Prefecture," "Kochi Prefecture," ... respectively. Furthermore, the partial regression coefficients b1, b2, b3, ... indicate the weight for each element. Note that b0 indicates the bias.

[0056] In the embodiment, multiple regression analysis is used as the method for learning the change amount prediction model, but other methods such as k-nearest neighbor analysis, support vector regression, and deep learning may also be used, and the method is not limited to these. The prediction model learning unit 12 stores the generated change amount prediction model in the prediction model storage unit 13.

[0057] The learning process performed by the information processing system 10 will be described with reference to Fig. 9. Fig. 9 is a flowchart showing the learning process performed by the information processing system 10.

[0058] First, the variation amount learning data generation unit 111 (see FIG. 4) acquires past handling histories from the handling history storage unit 17 (S11). The variation amount learning data generation unit 111 generates an empty list for storing variation amount learning data (S12). The variation amount learning data generation unit 111 generates variation amount learning data (S13). The variation amount learning data generation unit 111 can generate a number of pieces of variation amount learning data corresponding to the number of handling histories within the business operator.

[0059] Next, the change amount learning data generation unit 111 collects the handling details and values ​​such as weight after handling from the handling history related to processing within the business, values ​​such as weight before handling from the handling history added just before, and production information from the handling history in production processing (S14).

[0060] Specifically, the change amount learning data generation unit 111 first acquires handling histories stored in chronological order from the handling history storage unit 17. Next, the change amount learning data generation unit 111 focuses on one handling history and acquires the handling details and post-handling weight in that handling history. Next, the change amount learning data generation unit 111 references the handling history of the birth registration of a product with the same food ID as the handling history of interest and acquires production information. The production information may include information such as the place of origin, brand, and gender of the product. Then, the change amount learning data generation unit 111 references the handling history with the same food ID that was registered immediately before the handling history of interest and acquires the pre-handling weight of the handling history of interest.

[0061] The variation learning data generating unit 111 adds the handling details, values ​​such as weight before and after handling, and production information to the list (S15). The variation learning data generating unit 111 returns to step S13 and repeats the generation of variation learning data (S16).

[0062] Next, the food feature vector generation unit 112 performs preprocessing on the explanatory variables in the list of change amount training data to generate a food feature vector (S17). The prediction model training unit 12 trains the food feature vector and the objective variables in the list of change amount training data to generate a change amount prediction model that predicts the weight after handling (S18).

[0063] The anomaly detection unit 14 will be described with reference to Fig. 10. Fig. 10 is a block diagram showing the configuration of the anomaly detection unit 14. The anomaly detection unit 14 uses the above-mentioned change amount prediction model to detect anomalies in the handling history of the target product that is the inspection target. The anomaly detection unit 14 includes an explanatory variable collection unit 141, a food feature vector generation unit 142, a change amount prediction unit 143, and a prediction error threshold determination unit 144.

[0064] The internal processing of the explanatory variable collection unit 141 will be described with reference to Fig. 11. Fig. 11 is a diagram showing input and output data in the explanatory variable collection unit 141.

[0065] The explanatory variable collection unit 141 is an example of the acquisition unit 101. The explanatory variable collection unit 141 acquires a target product handling history that indicates the handling history of the target product that is the subject of the audit. Here, the explanatory variable collection unit 141 receives a new handling history P101 from the handling history creation unit 15 as the target product handling history.

[0066] Next, the explanatory variable collection unit 141 acquires past handling histories related to the same food as the target food related to the handling history P101 from the handling history storage unit 17. The explanatory variable collection unit 141 acquires the handling histories by, for example, using the food ID stored in the handling history storage unit 17 to refer to the handling histories related to the same food.

[0067] The internal processing of the explanatory variable collection unit 141 will be described with reference to Fig. 12. Fig. 12 is a diagram illustrating a method for generating collected data 1005 from handling histories. As an example, it is assumed here that handling histories P31, P32, and P33 are stored in the handling history storage unit 17. The handling histories P31, P32, and P33 indicate the handling history for a product with food ID "100."

[0068] First, the explanatory variable collection unit 141 acquires handling histories P31, P32, and P33 for the product with food ID "100" from the handling history storage unit 17. Next, the explanatory variable collection unit 141 acquires the place of origin, brand, sex, etc. from the birth registration handling history P31. The explanatory variable collection unit 141 acquires the handling content "disassembly" from the new handling history P101.

[0069] Furthermore, the explanatory variable collection unit 141 refers to the most recently added handling history P33 from among the handling histories acquired from the handling history storage unit 17. The explanatory variable collection unit 141 acquires the weight "100 kg" from the handling history P33 as the weight before handling of the handling history P101. In this way, the explanatory variable collection unit 141 collects data from the handling history storage unit 17 and acquires the collected data 1005.

[0070] The explanatory variable collection unit 141 transmits the collected data 1005 to the food feature vector generation unit 142. The explanatory variable collection unit 141 also transmits the handling history P101 to the prediction error threshold determination unit 144.

[0071] The internal processing of the change amount prediction unit 143 will be described with reference to Fig. 13. Fig. 13 is a diagram showing input and output data in the change amount prediction unit 143.

[0072] The change amount prediction unit 143 is an example of the prediction unit 102. The change amount prediction unit 143 uses a change amount prediction model based on the target product handling history to predict the weight of the target product after handling for the handling included in the target product handling history, and calculates the predicted weight of the target product. The change amount prediction model is trained using production information related to the production of the product, handling content information indicating the handling content of the product, and weight information indicating the weight of the product before and after handling as training data.

[0073] Specifically, first, when the change amount prediction unit 143 receives the food feature vector 1006 from the food feature vector generation unit 112, it acquires the change amount prediction model from the prediction model storage unit 13. Next, the change amount prediction unit 143 calculates a predicted value 1007 of the post-handling weight using the acquired change amount prediction model and the food feature vector 1006. In the example shown in the figure, weight W1 is shown as the predicted value 1007. Then, the change amount prediction unit 143 transmits the predicted value 1007 of the post-handling weight to the prediction error threshold determination unit 144.

[0074] The internal processing of the prediction error threshold determination unit 144 will be described with reference to Fig. 14. Fig. 14 is a diagram showing input and output data in the prediction error threshold determination unit 144.

[0075] The prediction error threshold determination unit 144 is an example of the above-mentioned abnormality determination unit 103. The prediction error threshold determination unit 144 determines whether or not there is an abnormality in the weight change indicated by the target product handling history based on the difference between the weight of the target product after handling in the target product handling history and the predicted weight of the target product calculated by the prediction unit. Furthermore, if the difference is less than a threshold, the prediction error threshold determination unit 144 determines that the weight change indicated by the target product handling history is normal, and if the difference is equal to or greater than the threshold, determines that the weight change is abnormal.

[0076] Specifically, first, the prediction error threshold determination unit 144 receives the predicted value 1007 of the post-handling weight from the change amount prediction unit 143. The prediction error threshold determination unit 144 also receives new handling history P101 from the explanatory variable collection unit 141. Next, the prediction error threshold determination unit 144 calculates the difference (prediction error) between the input value of the post-handling weight acquired from the handling history P101 and the predicted value 1007.

[0077] Here, the "input value" indicates the weight after handling in the handling indicated by the handling history P101. In the example shown in the figure, the input value is "70 kg." Therefore, the prediction error threshold determination unit 144 calculates the difference between the input value "70 kg" and the predicted value 1007, "W1," as the prediction error.

[0078] The prediction error threshold determination unit 144 compares the prediction error with a predetermined threshold to determine whether the weight change indicated by the handling history P101 is normal or abnormal. The threshold may be set in advance by a system administrator or the like. The threshold may be fixed or may be changed as appropriate.

[0079] The prediction error threshold determination unit 144 outputs different information depending on whether the determination result is normal or abnormal, as shown on the left and right sides of the lower part of Fig. 14. The prediction error threshold determination unit 144 determines that the handling history P101 is normal if the prediction error is less than the threshold. In this case, the prediction error threshold determination unit 144 transmits the handling history P101 to the handling history distribution management unit 16 and ends the processing.

[0080] Furthermore, if the prediction error is equal to or greater than the threshold, the prediction error threshold determination unit 144 determines that the weight change indicated by the handling history P101 is abnormal. In this case, the prediction error threshold determination unit 144 notifies the user of the handling history P101 as well as the predicted value 1007, and terminates the process. The user may be, for example, an auditor. In this way, the auditor can compare the weight stored in the handling history storage unit 17 with the weight predicted using the change amount prediction model. Therefore, the auditor can quickly grasp not only that an abnormality has been detected in the weight change in the handling history, but also the details of the abnormality.

[0081] Next, the processing of the abnormality detection unit 14 will be described with reference to Fig. 15. Fig. 15 is a flowchart showing the processing performed by the abnormality detection unit 14 (see Fig. 10).

[0082] First, the explanatory variable collection unit 141 inputs a handling history from the handling history creation unit 15 (S21). The input handling history indicates the handling history of the audit target.

[0083] Next, the explanatory variable collection unit 141 acquires past handling histories related to the same food as the input handling history from the handling history storage unit 17 (S22).

[0084] Next, the explanatory variable collection unit 141 collects the handling details from the input handling history, values ​​such as weight before handling from the most recently added past handling history, and production information from past handling history in the production process (S23).

[0085] Next, the food feature vector generation unit 142 performs preprocessing on the collected data to generate a food feature vector (S24). The change amount prediction unit 143 predicts the weight after handling using the change amount prediction model stored in the prediction model storage unit 13 and the food feature vector (S25).

[0086] Next, the prediction error threshold determination unit 144 calculates a prediction error from the prediction value and the input handling history (S26).The prediction error threshold determination unit 144 determines whether the prediction error is less than a threshold (S27).

[0087] If the prediction error is less than the threshold (YES in S27), the prediction error threshold determination unit 144 determines that the weight change indicated by the input handling history is normal (S28). In this case, the prediction error threshold determination unit 144 sends the input handling history to the handling history distributed management unit 16 (S29). On the other hand, if the prediction error is not less than the threshold (NO in S27), the prediction error threshold determination unit 144 determines that the weight change indicated by the input handling history is abnormal (S30). In this case, the prediction error threshold determination unit 144 notifies the user of the input handling history and the predicted value (S31).

[0088] As described above, the information processing system 10 uses a prediction model trained using production information, handling content information, and weight information as training data to predict the weight of the target product after handling for the handling included in the target product handling history and calculates the predicted weight of the target product. The information processing system 10 also determines whether there is an abnormality in the weight change indicated by the target product handling history based on the difference between the weight of the target product after handling in the target product handling history and the predicted target product weight calculated by the prediction unit. In this way, the information processing system 10 can appropriately detect abnormalities included in the product handling history.

[0089] <Embodiment 3> Next, embodiment 3 will be described. Embodiment 3 is a modification of embodiment 2. The information processing system 10 described above detects anomalies contained in the handling history based on changes in weight, etc. before and after handling of a product. The information processing system 10a according to the present disclosure is capable of dealing with falsification of qualitative data such as origin in addition to changes in weight, etc. As with embodiment 2, this embodiment will be described using product weight as an example of product quantity. Below, differences from the information processing system described above will be mainly described, and overlapping parts will be omitted as appropriate. The same applies to subsequent embodiments.

[0090] (Configuration of information processing system 10a) First, the information processing system 10a will be described with reference to Fig. 16. Fig. 16 is a block diagram showing the configuration of the information processing system 10a. In addition to the configuration of the information processing system 10 (see Fig. 3) described above, the information processing system 10a includes a production information combination creation unit 21 and a production information combination storage unit 22.

[0091] Furthermore, the abnormality detection unit 14a has a different configuration from the above-described abnormality detection unit 14. Specifically, the abnormality detection unit 14a generates verification data including the weight of the target product before handling, handling information of the target product, and possible combinations of production information, and detects an abnormality using the verification data.

[0092] The anomaly detection unit 14a also calculates, using the change amount prediction model, a predicted value of the post-handling weight corresponding to each of the plurality of verification data as a predicted verification weight. If, of the predicted target product weight and the plurality of predicted verification weights, the predicted target product weight is close to the post-handling weight in the target product handling history, the anomaly detection unit 14a determines that the weight change indicated in the target product handling history is normal; otherwise, it determines that the weight change indicated in the target product handling history is abnormal.

[0093] For example, the anomaly detection unit 14a calculates the difference (herein referred to as the "first predicted error") between the predicted weight of the target product and the weight after handling in the target product handling history. The first predicted error is a predicted error calculated using the collected data. The anomaly detection unit 14a also calculates the differences (herein referred to as the "second predicted error") between multiple predicted weights for verification and the weight after handling in the target product handling history. The second predicted error is a predicted error calculated using the verification data. The anomaly detection unit 14a calculates multiple second predicted errors corresponding to each of the multiple predicted weights for verification.

[0094] The anomaly detection unit 14a sorts the first prediction error and the plurality of second prediction errors in ascending order of value. The anomaly detection unit 14a identifies the rank of the first prediction error within the entire set (the first prediction error and the plurality of second prediction errors). For example, the anomaly detection unit 14a determines that the weight change is normal if the rank of the first prediction error is higher than a predetermined rank, and determines that the weight change is abnormal if the rank of the first prediction error is not higher than a predetermined rank.

[0095] Alternatively, the anomaly detection unit 14a may determine that the weight change is normal if the first prediction error is ranked highest, and determine that the weight change is abnormal if this is not the case. In other words, in this case, the anomaly detection unit 14a determines that the weight change is normal if the predicted weight of the target product is closest to the weight after handling in the target product handling history, and determines that the weight change is abnormal if this is not the case. In this embodiment, the explanation will mainly use this determination method.

[0096] Furthermore, if the abnormality detection unit 14a determines that the weight change indicated by the handling history of the target product is abnormal, it identifies, from the multiple verification data, the verification data that indicates the predicted verification weight that is closest to the weight after handling in the handling history of the target product, and notifies the user of information about the verification data.

[0097] The production information combination creation unit 21 creates possible combinations of production information (e.g., origin, brand, and producer) based on the variation learning data. The anomaly detection unit 14 creates verification data to which the combinations of production information are applied, and identifies the production information with the smallest prediction error (matching), thereby detecting anomalies in weight changes indicated by the handling history.

[0098] The internal processing of the production information combination creation unit 21 will be described with reference to Fig. 17. Fig. 17 is a diagram showing input and output data in the production information combination creation unit 21. First, the production information combination creation unit 21 receives explanatory variables 2001 from the variation learning data generation unit 111. Next, the production information combination creation unit 21 collects possible combinations of the production areas, brands, and producers included in the explanatory variables 2001. Then, the production information combination creation unit 21 transmits the production information combination 2002 to the production information combination storage unit 22.

[0099] Here, we will explain in detail the possible combinations of production information created by the production information combination creation unit 21. In the example shown in the figure, three combinations of production area, brand, and producer are shown, but here we will explain using two combinations of production area and brand.

[0100] For example, an example of a possible combination of place of origin and brand is the combination of place of origin "Kagawa Prefecture" and brand "Sanuki Beef" (registered trademark). In this way, the production information combination creation unit 21 creates a possible combination of production information, combining a place of origin and a brand corresponding to the place of origin. The production information combination creation unit 21 also collects such combinations. On the other hand, for example, place of origin "Kagawa Prefecture" and brand "Matsusaka Beef" (registered trademark) do not correspond to each other. The production information combination creation unit 21 does not create combinations of production information with contradictory correspondences like this.

[0101] The configuration of the abnormality detection unit 14a will be described with reference to Fig. 18. Fig. 18 is a block diagram showing the configuration of the abnormality detection unit 14a. In addition to the configuration of the abnormality detection unit 14 described above, the abnormality detection unit 14a includes a production information combination comprehensive unit 145 and a handling history consistency determination unit 146.

[0102] The internal processing of the production information combination comprehension unit 145 will be described with reference to Fig. 19. Fig. 19 is a diagram showing input and output data in the production information combination comprehension unit 145. First, when the production information combination comprehension unit 145 receives collected data 2003 from the explanatory variable collection unit 141, it acquires a combination of production information from the production information combination storage unit 22.

[0103] Next, the production information combination comprehensive unit 145 generates verification data 2004 based on the combinations of production information acquired from the production information combination storage unit 22. The production information combination comprehensive unit 145 generates verification data 2004 to which all possible production information has been applied. For example, in the example shown in the figure, the production information includes place of origin, brand, gender, and producer. The production information combination comprehensive unit 145 generates verification data 2004 consisting of the weight of the target product before handling, the handling details of the target product, and possible combinations of this production information. In this way, the production information combination comprehensive unit 145 can generate verification data 2004 that comprehensively covers all combinations of information included in the production information.

[0104] The weight before handling, which is quantitative data, is fixed. For example, in the example shown in the figure, the production information combination compiling unit 145 fixes the weight before handling to “100 kg” and generates the verification data 2004.

[0105] Furthermore, the production information combination comprehensive unit 145 is not limited to using all possible combinations, and may use a number of combinations less than all possible combinations to generate the verification data 2004. The production information combination comprehensive unit 145 transmits the collected data 2003 and the verification data 2004 to the food feature vector generation unit 142. The production information combination comprehensive unit 145 also transmits the collected data 2003 and the verification data 2004 to the handling history consistency determination unit 146.

[0106] The handling history consistency determination unit 146 will be described with reference to Fig. 20. Fig. 20 is a diagram showing input and output data in the handling history consistency determination unit 146. First, the handling history consistency determination unit 146 receives a list 2005 of predicted values ​​of post-handling weight from the change amount prediction unit 143. List 2005 contains predicted values ​​of post-handling weight corresponding to each of all combinations in the verification data. In the example shown, list 2005 includes W1, W2, ..., W20.

[0107] The handling history consistency determination unit 146 also receives a new handling history P201 from the explanatory variable collection unit 141. Furthermore, the handling history consistency determination unit 146 receives collected data 2003 and verification data 2004 from the production information combination collection unit 145.

[0108] Next, the handling history consistency determination unit 146 calculates the difference (prediction error) between the input value of the post-handling weight obtained from the handling history P201 and the predicted value. Then, if the prediction error in the collected data 2003 generated by the explanatory variable collection unit 141 is minimum, the handling history consistency determination unit 146 determines that the weight change indicated by the handling history P201 is normal. In this case, the handling history consistency determination unit 146 sends the handling history P201 to the handling history distribution management unit 16 and ends the processing.

[0109] Furthermore, if the prediction error in the collected data 2003 is not the smallest, the handling history consistency determination unit 146 determines that the weight change indicated by the handling history P101 is abnormal. In this case, the handling history consistency determination unit 146 identifies the verification data 2004a with the smallest prediction error from among the verification data 2004 generated by the production information combination comprehensive unit 145. This allows the handling history consistency determination unit 146 to identify information that is assumed to be true production information or information that is assumed to be true handling content when the production information or handling content included in the handling history P201 has been falsified.

[0110] 20, the handling history consistency determination unit 146 identifies the data with the place of origin "Tokushima Prefecture" as the verification data 2004a with the smallest prediction error. In this case, it is assumed that there is a high possibility that a product registered as being produced in "Kagawa Prefecture" is actually produced in "Tokushima Prefecture."

[0111] The handling history consistency determination unit 146 notifies the user of the handling history P201, the verification data 2004a with the smallest prediction error, and the predicted value 2006 for the collected data 2003. The user may be, for example, an auditor. In this manner, the auditor can determine that the weight change indicated by the handling history is abnormal, as well as grasp the true production information and the weight after handling that corresponds to the collected data 2003. After sending this data, the handling history consistency determination unit 146 ends the process.

[0112] The process performed by the abnormality detection unit 14a will be described with reference to Fig. 21 and Fig. 22. Fig. 21 and Fig. 22 are flowcharts showing the process performed by the abnormality detection unit 14a.

[0113] First, the explanatory variable collection unit 141 (see FIG. 18) inputs a handling history (S31). Next, the explanatory variable collection unit 141 acquires the input handling history and past handling history related to the same food item stored in the handling history storage unit 17 (S32).

[0114] Next, the explanatory variable collection unit 141 collects the handling details from the input handling history, values ​​such as weight before handling from the most recently added past handling history, and production information from past handling history in the production process (S33).

[0115] Next, the production information combination comprehensive unit 145 generates verification data that applies all possible production information based on the handling details, values ​​such as weight before handling, and combinations of production information stored in the production information combination storage unit 22 (S34). The food feature vector generation unit 142 preprocesses the collected data to generate a food feature vector (S35). The food feature vector generation unit 142 also preprocesses the verification data to generate a food feature vector (S36).

[0116] 22, the explanation continues. The change amount prediction unit 143 predicts the weight after handling using the change amount prediction model stored in the prediction model storage unit 13 and a food feature vector based on the collected data (S37). The change amount prediction unit 143 also predicts the weight after handling using the change amount prediction model stored in the prediction model storage unit 13 and a food feature vector based on the verification data (S38). The handling history consistency determination unit 146 calculates a prediction error from the predicted value and the input handling history (S39).

[0117] Next, the handling history consistency determination unit 146 determines whether the prediction error in the collected data is minimum (S40). If it is determined that the prediction error is minimum (YES in S40), the handling history consistency determination unit 146 determines that the weight change indicated by the handling history is normal (S41). In this case, the handling history consistency determination unit 146 transmits the input handling history to the handling history distribution management unit 16 (S42).

[0118] If it is determined that the prediction error in the collected data is not the smallest (NO in S40), the handling history consistency determination unit 146 determines that the weight change indicated by the handling history is abnormal (S43). In this case, the handling history consistency determination unit 146 notifies the user of the input handling history, the verification data with the smallest prediction error, and the prediction value in the collected data (S44).

[0119] As described above, in the information processing system 10a, the abnormality detection unit 14a generates verification data based on a combination of the handling details, the weight before handling, and the production information included in the collected data.

[0120] The anomaly detection unit 14a uses a change amount prediction model to predict the post-handling weight corresponding to the collected data and the post-handling weight corresponding to the verification data. Using these prediction results, the anomaly detection unit 14a determines whether the weight change indicated by the handling history is normal or abnormal based on the prediction error between the input value and the predicted value. If the weight change indicated by the target product handling history is determined to be abnormal, there is a risk that the target product has been counterfeited. The anomaly detection unit 14a uses the verification data with the smallest prediction error to identify information that is assumed to be the target product's true production information or true handling content.

[0121] In this way, the information processing system 10a can properly detect abnormalities contained in the product handling history, and can properly detect falsified production information. Furthermore, information regarding true production information, etc. can be provided to the user.

[0122] <Fourth Embodiment> Next, a fourth embodiment will be described. The fourth embodiment is a modified example of the third embodiment. In the third embodiment, in addition to changes in weight and the like, it is possible to deal with falsification of qualitative data such as origin. The information processing system 10b according to the present disclosure is capable of dealing with new production information such as the addition of new varieties. As with the second embodiment, in this embodiment, the weight of a product will be used as an example of the quantity of the product.

[0123] (Configuration of Information Processing System 10b) The information processing system 10b will be described with reference to Fig. 23. Fig. 23 is a block diagram showing the configuration of the information processing system 10b.

[0124] The information processing system 10b has the same configuration as the information processing system 10a (see FIG. 16) as above, but also includes a classification processing function. As a result, in the information processing system 10b, the change amount prediction model is trained using training data that has been grouped based on feature information that indicates the characteristics of the production information. In this way, the information processing system 10b can group similar items by production area, brand, and producer, and generate a change amount prediction model that has learned the change trends of values ​​such as weight before and after handling for each group.

[0125] Using the classification processing function, the information processing system 10b groups feature vectors corresponding to production information based on the similarity between the vectors and assigns group IDs. Furthermore, when a new production area, brand, or producer is added after the generation of the change amount prediction model, the information processing system 10b calculates the distance between the feature vector and the center of gravity vector of each group and associates it with the group ID of the group with the closest distance.

[0126] The information processing system 10b includes a production information feature vector storage unit 33, a production information association unit 34, and a centroid vector storage unit 35, which correspond to the classification processing function. In addition, the information processing system 10b also includes a production information grouping unit 31 as part of the learning processing function. Furthermore, the information processing system 10b also includes a classification result storage unit 32 as part of the traceability system function. The handling history shaping unit 11b and the anomaly detection unit 14b each have a different configuration from the configuration exemplified in the third embodiment. The anomaly detection unit 14b identifies a group to which production information corresponding to the target product handling history belongs, and detects an anomaly based on the identified group.

[0127] For the sake of explanation, the information processing system 10b is configured by adding these functional units to the information processing system 10a, but the configuration of the information processing system 10b may be changed as appropriate. For example, the information processing system 10b may not include the production information combination creation unit 21.

[0128] 24 and 25, the internal processing of the production information grouping unit 31 will be described. Fig. 24 is a diagram showing input data in the production information grouping unit 31. Fig. 25 is a diagram showing output data in the production information grouping unit 31.

[0129] First, input to the production information grouping unit 31 will be described with reference to Fig. 24. The production information grouping unit 31 acquires a production area characteristic vector 3001, a brand characteristic vector 3002, and a producer characteristic vector 3003 from the production information characteristic vector storage unit 33. The characteristic vectors 3001 to 3003 are information indicating the characteristics of each piece of production information. For example, the production area characteristic vector 3001 indicates the annual average temperature, the monthly maximum temperature, and the number of rainy days per year, which are characteristics of a production area. As shown in the figure, the characteristic vectors 3001 to 3003 are merely examples, and each may include other information.

[0130] Next, the production information grouping unit 31 groups the production area characteristic vectors 3001 based on the similarity between the production area characteristic vectors and assigns a production area group ID to each group. Similarly, the production information grouping unit 31 groups the brand characteristic vectors 3002 based on the similarity between the brand characteristic vectors and assigns a brand group ID to each group. Furthermore, the production information grouping unit 31 groups the producer characteristic vectors 3003 based on the similarity between the producer characteristic vectors and assigns a producer group ID to each group.

[0131] Next, the output of the production information grouping unit 31 will be described with reference to Fig. 25. The production information grouping unit 31 saves a pair 3004 of a production area and a production area group ID, a pair 3005 of a brand and a brand group ID, and a pair 3006 of a producer and a producer group ID in the classification result storage unit 32 and transmits them to the handling history formatting unit 11b.

[0132] Next, the production information grouping unit 31 stores a pair 3007 of the production area group ID and the centroid vector of the production area group, a pair 3008 of the brand group ID and the centroid vector of the brand group, and a pair 3009 of the production company group ID and the centroid vector of the production company group in the centroid vector storage unit 35. The production information grouping unit 31 can calculate the centroid vector using the average value or median value of each feature vector.

[0133] Next, the handling history shaping unit 11b will be described with reference to Fig. 26. Fig. 26 is a block diagram showing the configuration of the handling history shaping unit 11b. The handling history shaping unit 11b includes a variation amount learning data generation unit 111 and a food feature vector generation unit 112b.

[0134] The internal processing of the food feature vector generation unit 112b will be described with reference to Fig. 27. Fig. 27 is a diagram showing input and output data in the food feature vector generation unit 112b.

[0135] First, when the food feature vector generation unit 112b receives the explanatory variables 3010 from the change amount learning data generation unit 111, it acquires from the production information grouping unit 31 a pair 3004 of a production area and a production area group ID, a pair 3005 of a brand and a brand group ID, and a pair 3006 of a producer and a producer group ID.

[0136] Next, the food feature vector generation unit 112b replaces the values ​​of the origin, brand, and producer included in the explanatory variables 3010 with the origin group ID, brand group ID, and producer group ID to which they belong, respectively.

[0137] Next, the food feature vector generation unit 112b performs preprocessing such as one-hot encoding on the qualitative variables included in the explanatory variables 3010. The food feature vector generation unit 112b also performs preprocessing such as standardization on the quantitative variables included in the explanatory variables 3010. The food feature vector generation unit 112b transmits the food feature vector 3011 generated by the preprocessing to the prediction model learning unit 12.

[0138] Next, the processing performed by the information processing system 10b will be described with reference to Fig. 28. Fig. 28 is a flowchart showing the processing performed by the information processing system 10b.

[0139] First, the variation amount learning data generation unit 111 acquires past handling histories stored in the handling history storage unit 17 (S51). The variation amount learning data generation unit 111 generates an empty list for storing variation amount learning data (S52). The variation amount learning data generation unit 111 generates variation amount learning data (S53). The variation amount learning data generation unit 111 can generate a number of pieces of variation amount learning data corresponding to the number of handling histories within the business operator.

[0140] Next, the variation learning data generation unit 111 collects handling details and values ​​such as weight after handling from handling history related to processing within the business, values ​​such as weight before handling from handling history added immediately before the handling history related to processing within the business, and production information from handling history in production processing (S54). The variation learning data generation unit 111 adds the handling details, values ​​such as weight before and after handling, and production information to the list (S55). The variation learning data generation unit 111 returns to step S53 and repeats the generation of variation learning data (S56).

[0141] Next, the production information grouping unit 31 groups the origin feature vectors, brand feature vectors, and producer feature vectors stored in the production information feature vector storage unit 33 based on the similarity between the vectors, and assigns an origin group ID, brand group ID, and producer group ID to each feature vector (S57). Furthermore, the food feature vector generation unit 112 replaces the origin, brand, and producer values ​​in the list of variation learning data with the origin group ID, brand group ID, and producer group ID to which they belong, respectively (S58).

[0142] Next, the food feature vector generation unit 112 performs preprocessing on the explanatory variables in the list of change amount training data to generate a food feature vector (S59).Then, the prediction model training unit 12 trains the food feature vector and the objective variables in the list of change amount training data to generate a change amount prediction model that predicts values ​​such as weight after handling (S60).

[0143] Next, the processing performed by the production information grouping unit 31 will be described with reference to Fig. 29. Fig. 29 is a flowchart showing the processing performed by the production information grouping unit 31.

[0144] First, the production information grouping unit 31 acquires the origin characteristic vector, brand characteristic vector, and producer characteristic vector from the production information characteristic vector storage unit 33 (S71). The production information grouping unit 31 performs grouping processing (S72). The production information grouping unit 31 can perform grouping processing using groups whose number corresponds to the type of characteristic vector.

[0145] Next, the production information grouping unit 31 groups the feature vectors based on the similarity between the feature vectors and assigns group IDs to the feature vectors (S73). The production information grouping unit 31 stores a pair of the qualitative data associated with the feature vector and the group ID in the classification result storage unit 32 and transmits the pair to the handling history formatting unit 11b (S74). The production information grouping unit 31 also stores a pair of the group ID and the centroid vector in the centroid vector storage unit 35 (S75). The production information grouping unit 31 returns to step S72 and repeats the grouping process (S76).

[0146] The internal processing of the production information associating unit 34 will be described with reference to Fig. 30 and Fig. 31. Fig. 30 is a diagram showing input data in the production information associating unit 34. Fig. 31 is a diagram showing output data in the production information associating unit 34.

[0147] First, referring to Fig. 30, the input to the production information association unit 34 will be described. The production information association unit 34 acquires, from the production information characteristic vector storage unit 33, a production area characteristic vector 3012 corresponding to the new production area, a brand characteristic vector 3013 corresponding to the new brand, and a producer characteristic vector 3014 corresponding to the new producer.

[0148] The production information association unit 34 acquires from the centroid vector storage unit 35 a pair 3007 of a production area group ID and a centroid vector, a pair 3008 of a brand group ID and a centroid vector, and a pair 3009 of a producer group ID and a centroid vector.

[0149] Next, the output of the production information associating unit 34 will be described with reference to Figure 31. The production information associating unit 34 calculates the distance between the production area characteristic vector 3012 corresponding to the new production area and the centroid vector of the production area group. The production information associating unit 34 saves a pair 3015 of the production area group ID of the closest production area group and the new production area in the classification result storage unit 32.

[0150] Similarly, the production information associating unit 34 calculates the distance between each brand characteristic vector and the center of gravity vector of the brand group. The production information associating unit 34 stores a pair 3016 of the brand group ID of the brand group with the shortest distance and the new brand in the classification result storage unit 32.

[0151] The production information association unit 34 also calculates the distance between the producer characteristic vector and the centroid vector of the producer group. The production information association unit 34 stores a pair 3017 of the producer group ID of the closest producer group and the new producer in the classification result storage unit 32.

[0152] The classification process performed by the information processing system 10b will be described with reference to Fig. 32. Fig. 32 is a flowchart showing the classification process performed by the information processing system 10b.

[0153] The production information grouping unit 31 acquires the production area characteristic vector, brand characteristic vector, and producer characteristic vector corresponding to the new production area, brand, and producer from the production information characteristic vector storage unit 33 (S81).

[0154] The production information grouping unit 31 acquires a pair of a production area group ID and a centroid vector, a pair of a brand group ID and a centroid vector, and a pair of a producer group ID and a centroid vector from the centroid vector storage unit 35 (S82).

[0155] The production information grouping unit 31 calculates the distance between each of the production area characteristic vectors and the centroid vectors of the production area groups (S83). The production information grouping unit 31 stores the pair of the production area group ID of the closest production area group and the new production area in the classification result storage unit 32 (S84).

[0156] The production information grouping unit 31 calculates the distance between the brand specific vector and the center of gravity vector of each brand group (S85). The production information grouping unit 31 stores the pair of the brand group ID of the brand group with the closest distance and the new brand in the classification result storage unit 32 (S86).

[0157] The production information grouping unit 31 calculates the distance between each of the producer characteristic vectors and the centroid vectors of the producer groups (S87).The production information grouping unit 31 stores the pair of the producer group ID of the producer group with the closest distance and the new producer in the classification result storage unit 32 (S88).

[0158] The configuration of the anomaly detection unit 14b will be described with reference to Fig. 33. Fig. 33 is a block diagram showing the configuration of the anomaly detection unit 14b. The anomaly detection unit 14b includes a food feature vector generation unit 142b instead of the food feature vector generation unit 142 of the anomaly detection unit 14a (see Fig. 18) described above.

[0159] The processing of the abnormality detection unit 14b will be described with reference to Fig. 34 and Fig. 35. Fig. 34 and Fig. 35 are flowcharts showing the processing performed by the abnormality detection unit 14b.

[0160] First, the explanatory variable collection unit 141 (see FIG. 18 ) inputs a handling history (S91). Next, the explanatory variable collection unit 141 refers to the handling history storage unit 17 and collects handling details, values ​​such as weight before handling, and production information from past handling histories related to the same food as the input handling history (S92).

[0161] Next, the explanatory variable collection unit 141 generates verification data that applies all possible production information based on a combination of the handling content, values ​​such as weight before handling, and production information stored in the production information combination storage unit 22 (S93).

[0162] The food feature vector generation unit 142b replaces the production information contained in the collected data and the production information contained in the verification data with a production area group ID, brand group ID, and producer group ID (S94). The food feature vector generation unit 142b preprocesses the collected data to generate a food feature vector (S95). The food feature vector generation unit 142b also preprocesses the verification data to generate a food feature vector (S96). The change amount prediction unit 143 predicts the weight after handling based on the change amount prediction model stored in the prediction model storage unit 13 and the food feature vector based on the collected data (S97).

[0163] 35, the explanation continues. The change amount prediction unit 143 predicts the weight after handling based on the change amount prediction model stored in the prediction model storage unit 13 and the food feature vector based on the verification data (S98). The handling history consistency determination unit 146 calculates the prediction error based on the predicted value and the input handling history (S99).

[0164] Next, the handling history consistency determination unit 146 determines whether the prediction error in the collected data is minimum (S100). If it is determined that the prediction error is minimum (YES in S100), the handling history consistency determination unit 146 determines that the weight change indicated by the handling history is normal (S101). In this case, the handling history consistency determination unit 146 transmits the input handling history to the handling history distribution management unit 16 (S102).

[0165] If it is determined that the prediction error in the collected data is not the smallest (NO in S100), the handling history consistency determination unit 146 determines that the weight change indicated by the handling history is abnormal (S103). In this case, the handling history consistency determination unit 146 notifies the user of the input handling history, the verification data with the smallest prediction error, and the prediction value in the collected data (S104).

[0166] As described above, the classification processing function of the information processing system 10b can group production information according to its characteristics. In this way, the information processing system 10b can accurately detect anomalies even when it is not possible to learn using a sufficient amount of learning data, for example, in the period immediately after the system is introduced.

[0167] <Fifth Embodiment> Next, a fifth embodiment will be described. The fifth embodiment is a modified example of the fourth embodiment. The fourth embodiment is capable of handling new production information, such as the addition of new varieties. The information processing system 10c according to the present disclosure further makes it possible to easily analyze data stored in a blockchain or the like.

[0168] (Configuration of information processing system 10c) The information processing system 10c will be described with reference to Fig. 36. Fig. 36 is a block diagram showing the configuration of the information processing system 10c. In addition to the configuration of the information processing system 10b described above (see Fig. 23), the information processing system 10c further includes a handling history food-based classification unit 41.

[0169] The handling history by food categorization unit 41 acquires time-series handling history information in which the handling histories of multiple products are stored in chronological order, and extracts the handling history for each product by performing a classification process on the time-series handling history information. The time-series handling history information is, for example, the handling history of multiple products stored in a blockchain. By performing the classification process, the handling history by food categorization unit 41 can classify the handling history, which is stored in a state where it can only be traced in chronological order, by food.

[0170] For the sake of explanation, the following description will be given using an example in which the handling history food classification unit 41 is added to the information processing system 10b described above, but the present invention is not limited to this. For example, the handling history food classification unit 41 may be added to any of the above-described embodiments 1 to 4.

[0171] The characteristics of the handling history stored in the blockchain will be described with reference to Figure 37. Figure 37 is a diagram illustrating the characteristics of the handling history stored in the blockchain. The handling history is stored in the handling history distributed management unit 16 and the handling history storage unit 17. Here, it is assumed that the handling history distributed management unit 16 and the handling history storage unit 17 are configured using a blockchain. Therefore, it is assumed that the handling histories P41, P42, and P43 shown in the figure are stored in the blockchain.

[0172] The handling history P41 indicates the handling history of production processing within a business. The handling history P42 indicates the handling history of processing within a business. The handling history P43 indicates the handling history of transactions between businesses.

[0173] On a blockchain, there is a limit to the amount of data that can be stored in one block. Therefore, the amount of transaction history data stored on the blockchain must be kept small enough to not exceed the limit. This means that redundant information required for analysis cannot be stored.

[0174] On a blockchain, multiple handling histories are stored in chronological order, making it difficult to extract only products with a specific food ID. For example, suppose handling histories P41 to P43 relate to one pig A, which is managed with the same food ID. Handling histories P41 to P43 are stored in a chain with handling histories for other pigs B, C, and so on, which are managed with other food IDs. Therefore, it is difficult to extract only information about pig A from the blockchain.

[0175] The internal processing of the handling history food categorization unit 41 will be described with reference to Fig. 38. Fig. 38 is a diagram showing input and output data in the handling history food categorization unit 41.

[0176] First, the handling history by food categorization unit 41 acquires handling histories in chronological order from the handling history storage unit 17. For example, in the example shown in the figure, the handling history by food categorization unit 41 acquires handling histories P1, P2, P3, ..., P9, and P10. The handling histories P1, P2, P3, ..., P9, and P10 store information on a product with a food ID of "001" and information on a product with a food ID of "002" in a chain.

[0177] The handling history by food categorization unit 41 categorizes the handling history data so that handling histories related to the same food are listed in the same list. In the example shown in the figure, the handling history by food categorization unit 41 categorizes the handling history data so that only handling histories related to the product with food ID "001" are extracted. The handling history by food categorization unit 41 sends the handling history lists P51, P52, P53, P54, ..., P60, categorized by food, to the handling history formatting unit 11b.

[0178] The processing of the handling history food classification unit 41 will be described with reference to Fig. 39. Fig. 39 is a flowchart showing the processing of the handling history food classification unit 41.

[0179] First, the handling history food categorization unit 41 performs food categorization processing (S111). The handling history food categorization unit 41 can perform food categorization processing for the number of food items corresponding to the number of handling histories. Next, the handling history food categorization unit 41 inputs past handling histories (S112).

[0180] Next, the handling history food categorization unit 41 determines whether or not there is a handling history list related to the same food (S113). If it is determined that there is a handling history list related to the same food (YES in S113), the handling history food categorization unit 41 proceeds to the processing of step S115. If it is determined that there is no handling history list related to the same food (NO in S113), the handling history food categorization unit 41 creates a new list storing handling history related to the same food (S114).

[0181] Next, the handling history by food categorization unit 41 adds the handling history to the list of handling histories related to the same food (S115). The handling history by food categorization unit 41 returns to step S111 and repeats the food categorization process (S116). The handling history by food categorization unit 41 then sends the list of handling histories categorized by food to the handling history formatting unit 11b (S117).

[0182] As described above, the information processing system 10c classifies multiple transaction histories, which are stored in a state where they can only be traced in chronological order, by product. In this way, classification can be performed by product, not just chronological order. Therefore, the information processing system 10c can easily manage information for each product while appropriately detecting counterfeit products.

[0183] The information processing systems 10, 10a, 10b, and 10c (hereinafter referred to as "information processing systems 10, etc.") have been described above as specific examples of the first embodiment. The configuration of the information processing system 10, etc. described above is merely an example and may be modified as appropriate. For example, when some or all of the components of the information processing system 10, etc. are realized by multiple information processing devices, circuits, etc., the multiple information processing devices, circuits, etc. may be centrally or decentralized. For example, the information processing devices, circuits, etc. may be realized as a client-server system, a cloud computing system, or the like, in a form in which each is connected via a communication network. Furthermore, the functions of the information processing system 10, etc. may be provided in a SaaS (Software as a Service) format.

[0184] The above-described embodiments can be implemented in any combination. For example, embodiment 2 and embodiment 3 may be combined. Embodiment 2 and embodiment 4 may be combined. Furthermore, embodiments 2, 3, and 4 may be combined. Furthermore, embodiment 5 may be combined with each of embodiments 2 to 4.

[0185] <Example of Hardware Configuration> Each functional component of the information processing system 10, etc. may be realized by hardware that realizes the functional component (e.g., a hardwired electronic circuit, etc.), or may be realized by a combination of hardware and software (e.g., a combination of an electronic circuit and a program that controls it, etc.). Below, a case where each functional component of the information processing system 10, etc. is realized by a combination of hardware and software will be further described.

[0186] 40 is a block diagram illustrating an example of the hardware configuration of a computer 900 that realizes the information processing system 10, etc. The computer 900 may be a dedicated computer designed to realize the information processing system 10, etc., or may be a general-purpose computer. The computer 900 may also be a portable computer such as a smartphone or a tablet terminal.

[0187] For example, by installing a predetermined application on the computer 900, the computer 900 realizes each function of the information processing system 10, etc. The application is configured by a program for realizing the functional components of the information processing system 10, etc.

[0188] The computer 900 has a bus 902, a processor 904, a memory 906, a storage device 908, an input / output interface 910, and a network interface 912. The bus 902 is a data transmission path for the processor 904, the memory 906, the storage device 908, the input / output interface 910, and the network interface 912 to transmit and receive data to and from each other. However, the method of connecting the processor 904 and other components to each other is not limited to a bus connection.

[0189] The processor 904 is one of various processors such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an FPGA (Field-Programmable Gate Array), or a quantum processor (quantum computer control chip). The memory 906 is a main storage device realized using a RAM (Random Access Memory) or the like. The storage device 908 is an auxiliary storage device realized using a hard disk, an SSD (Solid State Drive), a memory card, a ROM (Read Only Memory), or the like.

[0190] The input / output interface 910 is an interface for connecting the computer 900 with input / output devices. For example, the input / output interface 910 is connected to an input device such as a keyboard and an output device such as a display device.

[0191] The network interface 912 is an interface for connecting the computer 900 to a network. This network may be a LAN (Local Area Network) or a WAN (Wide Area Network).

[0192] The storage device 908 stores programs (programs that realize the above-mentioned applications) that realize the various functional components of the information processing system 10, etc. The processor 904 reads these programs into the memory 906 and executes them to realize the various functional components of the information processing system 10, etc.

[0193] Each processor executes one or more programs containing instructions for causing a computer to perform an algorithm. The programs contain instructions (or software code) that, when loaded into a computer, cause the computer to perform one or more functions described in the embodiments. The programs may be stored on various types of non-transitory computer-readable or tangible storage media. By way of example and not limitation, non-transitory computer-readable or tangible storage media include random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technologies, CD-ROM, digital versatile disc (DVD), Blu-ray disc or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices. The programs may also be transmitted over various types of transitory computer-readable or communication media. By way of example and not limitation, transitory computer-readable or communication media include electrical, optical, acoustic, or other forms of propagated signals.

[0194] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. Furthermore, each embodiment can be combined with other embodiments as appropriate.

[0195] For example, in the above embodiment, weight information indicating the weight of a product is used as the quantity information, but as described above, other quantitative data may be used as the quantity information. For example, information such as the volume, length, or number of products may be used as the quantity of products used as learning data, the quantity of products to be predicted, and the quantity of products to be subject to anomaly detection.

[0196] Each drawing is merely an example for describing one or more embodiments. Each drawing may not relate only to one particular embodiment, but may also relate to one or more other embodiments. As will be understood by those skilled in the art, various features or steps described with reference to any one drawing can be combined with features or steps shown in one or more other drawings to create, for example, an embodiment not explicitly shown or described. Not all features or steps shown in any one drawing are necessary to describe an exemplary embodiment, and some features or steps may be omitted. The order of steps described in any drawing may be changed as appropriate.

[0197] Some or all of the above embodiments can be described as, but are not limited to, the following supplementary notes. (Supplementary Note 1) An information processing device comprising: an acquisition unit that acquires a target product handling history indicating the handling history of a target product that is the target of an audit; a prediction unit that calculates a target product forecast quantity by predicting a post-handling quantity for the handling included in the target product handling history using a prediction model trained using, as learning data, production information related to the production of the product, handling content information indicating the handling content of the product, and quantity information indicating the quantity of the product before and after handling, based on the target product handling history; and an abnormality determination unit that determines whether or not there is an abnormality in the quantity change indicated by the target product handling history, based on the difference between the post-handling quantity of the target product in the target product handling history and the target product forecast quantity calculated by the prediction unit. (Supplementary Note 2) The information processing device according to Supplementary Note 1, wherein the prediction model is trained using, as the production information, information related to at least one of the place of origin, brand, and producer of the product. (Supplementary Note 3) The information processing device according to Supplementary Note 1 or 2, wherein the abnormality determination unit determines that the quantity change is normal if the difference is less than a threshold, and determines that the quantity change is abnormal if the difference is equal to or greater than a threshold. (Supplementary Note 4) The information processing device according to any one of Supplements 1 to 3, wherein the abnormality determination unit generates verification data including a quantity of the target product before handling, handling information of the target product, and possible combinations of production information, and determines whether or not there is an abnormality in the quantity change using the verification data. (Supplementary Note 5) The information processing device according to Supplementary Note 4, wherein the abnormality determination unit uses the prediction model to calculate, as a verification predicted quantity, a predicted value of the post-handling quantity corresponding to each of the plurality of verification data, and determines that the quantity change is normal if the target product predicted quantity and the plurality of verification predicted quantities are close to the post-handling quantity in the target product handling history, and otherwise determines that the quantity change is abnormal.(Supplementary Note 6) The information processing device according to Supplementary Note 5, wherein, when it is determined that the quantity change is abnormal, the abnormality determination unit identifies, from the plurality of verification data, verification data indicating the verification predicted quantity that is closest to the quantity after handling in the target product handling history, and notifies a user of information about the verification data. (Supplementary Note 7) The prediction model performs learning using the learning data grouped based on feature information that indicates features of the production information, and the abnormality determination unit identifies a group to which production information corresponding to the target product handling history belongs, and determines whether or not there is an abnormality in the quantity change based on the identified group. (Supplementary Note 8) The information processing device according to any one of Supplementary Notes 1 to 7, further comprising a handling history product classifier that acquires time-series handling history information in which handling histories of a plurality of products are stored in chronological order, and performs a classification process on the time-series handling history information to extract a handling history for each product, and the prediction model performs learning using the extracted handling history for each product. (Supplementary Note 9) A prediction model for causing a computer to function such that, when a target product handling history indicating the handling history of a target product that is the subject of an audit is input, the prediction model predicts the post-handling quantity for the handling included in the target product handling history and outputs a predicted target product quantity. (Supplementary Note 10) The prediction model according to Supplementary Note 9 is trained using information on at least one of the place of origin, brand, and producer of the product as the production information.(Supplementary Note 11) An information processing method comprising: an acquisition step of acquiring a target product handling history indicating the handling history of a target product that is the target of an audit; a prediction step of calculating a target product forecast quantity by predicting the post-handling quantity for the handling included in the target product handling history using a prediction model trained using, as training data, production information regarding the production of the product, handling content information indicating the handling content of the product, and quantity information indicating the quantity of the product before and after handling, based on the target product handling history; and an abnormality determination step of determining whether or not there is an abnormality in the quantity change indicated by the target product handling history, based on the difference between the post-handling quantity of the target product in the target product handling history and the target product forecast quantity calculated in the prediction step. (Supplementary Note 12) The information processing method according to Supplementary Note 11, wherein the prediction model is trained using information regarding at least one of the place of origin, brand, and producer of the product as the production information. (Supplementary Note 13) The information processing method according to Supplementary Note 11 or 12, wherein in the abnormality determination step, if the difference is less than a threshold, the amount change is determined to be normal, and if the difference is equal to or greater than a threshold, the amount change is determined to be abnormal. (Supplementary Note 14) The information processing method according to any one of Supplements 11 to 13, wherein in the abnormality determination step, verification data including a pre-handling quantity of the target product, handling information of the target product, and possible combinations of production information are generated, and whether or not there is an abnormality in the amount change is determined using the verification data. (Supplementary Note 15) The information processing method according to Supplementary Note 14, wherein in the abnormality determination step, using the prediction model, a predicted value of the post-handling quantity corresponding to each of the plurality of verification data is calculated as a verification predicted quantity, and if the target product predicted quantity, of the target product predicted quantity and the plurality of verification predicted quantities, is close to the post-handling quantity in the target product handling history, the amount change is determined to be normal, and otherwise the amount change is determined to be abnormal.(Supplementary Note 16) The information processing method according to Supplementary Note 15, wherein, in the anomaly determination step, if it is determined that the quantity change is abnormal, verification data indicating the verification predicted quantity closest to the quantity after handling in the target product handling history is identified from the plurality of verification data, and information about the verification data is notified to a user. (Supplementary Note 17) The information processing method according to any one of Supplements 11 to 16, wherein the prediction model performs learning using the learning data grouped based on feature information indicating features of the production information, and the anomaly determination step identifies a group to which production information corresponding to the target product handling history belongs, and determines whether or not there is an abnormality in the quantity change based on the identified group. (Supplementary Note 18) The information processing method according to any one of Supplements 11 to 17, further comprising a handling history product classifying step of acquiring time-series handling history information in which handling histories of a plurality of products are stored in chronological order, and performing a classification process on the time-series handling history information to extract handling histories for each product, and the prediction model performs learning using the extracted handling histories for each product. (Supplementary Note 19) A program that causes a computer to execute the following steps: an acquisition step of acquiring a target product handling history that indicates the handling history of a target product that is the target of an audit; a prediction step of calculating a target product forecast quantity by predicting a post-handling quantity for a handling included in the target product handling history using a prediction model trained using, as training data, production information related to the production of the product, handling content information that indicates the handling content of the product, and quantity information that indicates the quantity of the product before and after handling, based on the target product handling history; and an abnormality determination step of determining whether or not there is an abnormality in the quantity change indicated by the target product handling history, based on the difference between the post-handling quantity of the target product in the target product handling history and the target product forecast quantity calculated in the prediction step. (Supplementary Note 20) The program according to Supplementary Note 19, wherein the prediction model is trained using information on at least one of the place of origin, brand, and producer of the product as the production information.

[0198] Some or all of the elements (e.g., configurations and functions) described in Supplementary Notes 2 to 8 that are dependent on Supplementary Note 1 may also be dependent on Supplementary Notes 9, 11, and 19 in the same dependency relationship as Supplementary Notes 2 to 8. Some or all of the elements described in any Supplementary Note may be applied to various hardware, software, recording means for recording software, systems, and methods.

[0199] This application claims priority based on Japanese Patent Application No. 2023-021783, filed February 15, 2023, the disclosure of which is incorporated herein in its entirety.

[0200] 10, 10a to 10c Information processing system 11, 11b Handling history shaping unit 12 Prediction model learning unit 13 Prediction model storage unit 14, 14a, 14b Anomaly detection unit 15 Handling history creation unit 16 Handling history distributed management unit 17 Handling history storage unit 18 Handling history tracking unit 21 Production information combination creation unit 22 Production information combination storage unit 31 Production information grouping unit 32 Classification result storage unit 33 Production information feature vector storage unit 34 Production information association unit 35 Center of gravity vector storage unit 41 Handling history food classification unit 100 Information processing device 101 Acquisition unit 102 Prediction unit 103 Anomaly determination unit 111 Variation amount learning data generation unit 112, 112b Food feature vector generation unit 141 Explanatory variable collection unit 142, 142b Food feature vector generation unit 143 Change amount prediction unit 144 Prediction error threshold determination unit 145 Production information combination comprehensive unit 146 Handling history consistency determination unit 1001 Explanatory variable 1002 Objective variable 1003 Collected data 1004 Food feature vector 1005 Collected data 1006 Food feature vector 1007 Predicted value 2001 Explanatory variable 2003 Collected data 2004, 2004a Verification data 2005 List of predicted values ​​of weight after handling 2006 Predicted value 3001 Origin feature vector 3002 Brand feature vector 3003 Producer feature vector 3004 Pair of origin and origin group ID 3005 Pair of brand and brand group ID 3006 Pair of producer and producer group ID 3007 Pair of origin group ID and origin group centroid vector 3008 Pair of brand group ID and brand group centroid vector 3009 Pair of producer group ID and producer group centroid vector 3010 Explanatory variable 3011 Food feature vector 3012 Production area feature vector 3013 Brand feature vector 3014 Producer feature vector 3015 Pair of new production area and production area group ID 3016 Pair of new brand and brand group ID 3017 Pair of new producer and producer group ID P1 to P201 Handling history 900 Computer 902 Bus 904 Processor 906 Memory 908 Storage device 910 Input / output interface 912 Network interface

Claims

1. an acquisition means for acquiring a target product handling history indicating a handling history of the target product to be audited; a prediction means for calculating a predicted quantity of the target commodity by predicting the quantity after handling for the handling included in the target commodity handling history, using a prediction model trained using production information related to the production of the commodity, handling content information indicating the handling content of the commodity, and quantity information indicating the quantity of the commodity before and after handling as learning data based on the target commodity handling history; and an abnormality determination means for determining whether or not there is an abnormality in the quantity change indicated by the target product handling history based on the difference between the post-handling quantity of the target product in the target product handling history and the target product predicted quantity calculated by the prediction means. Information processing device.

2. The prediction model is trained using information on at least one of the origin, brand, and manufacturer of the product as the production information. The information processing device according to claim 1 .

3. The abnormality determination means determines that the change in amount is normal when the difference is less than a threshold value, and determines that the change in amount is abnormal when the difference is equal to or greater than a threshold value. The information processing device according to claim 1 .

4. The abnormality determination means generating verification data consisting of possible combinations of the quantity of the target product before handling, handling information of the target product, and production information; The verification data is used to determine whether or not there is an abnormality in the change in quantity. The information processing device according to claim 1 .

5. The abnormality determination means Using the prediction model, a predicted value of the post-handling amount corresponding to each of the plurality of verification data is calculated as a verification predicted amount; If the target product predicted quantity among the target product predicted quantity and the plurality of verification predicted quantities is close to the quantity after handling in the target product handling history, the quantity change is determined to be normal, and otherwise, the quantity change is determined to be abnormal. The information processing device according to claim 4 .

6. The abnormality determination means If it is determined that the change in the amount is abnormal, the verification data indicating the predicted verification amount that is closest to the amount after handling in the target product handling history is identified from the plurality of verification data, and information about the verification data is notified to the user. The information processing device according to claim 5 .

7. the prediction model is trained using the training data grouped based on feature information indicating features of the production information, The abnormality determination means identifies a group to which the production information corresponding to the target product handling history belongs, and determines whether or not there is an abnormality in the quantity change based on the identified group. The information processing device according to any one of claims 1 to 6.

8. The information is generated by learning based on learning data including production information related to the production of the product, handling content information indicating the handling content of the product, and quantity information indicating the quantity of the product before and after handling, A prediction model that causes a computer to function such that, when a target product handling history indicating the handling history of the target product being audited is input, the computer predicts the post-handling quantity for the handling included in the target product handling history and outputs the predicted target product quantity.

9. Obtain a product handling history showing the handling history of the product being audited, Based on the target commodity handling history, a prediction model is used that has been trained using production information related to the production of the commodity, handling content information indicating the handling content of the commodity, and quantity information indicating the quantity of the commodity before and after handling as training data to predict the quantity after handling for the handling included in the target commodity handling history, thereby calculating a target commodity predicted quantity; Based on the difference between the post-handling quantity of the target product in the target product handling history and the calculated predicted quantity of the target product, it is determined whether or not there is an abnormality in the quantity change indicated by the target product handling history. Information processing methods.

10. Obtain a product handling history showing the handling history of the product being audited, Based on the target commodity handling history, a prediction model is used that has been trained using production information related to the production of the commodity, handling content information indicating the handling content of the commodity, and quantity information indicating the quantity of the commodity before and after handling as training data to predict the quantity after handling for the handling included in the target commodity handling history, thereby calculating a target commodity predicted quantity; and determining whether or not there is an abnormality in the quantity change indicated by the target product handling history based on the difference between the post-handling quantity of the target product in the target product handling history and the calculated target product predicted quantity. program.