Information processing device, inference device, machine learning device, information processing method, inference method, and machine learning method

Abstract

[Problem] To provide an information processing device which makes it possible to easily predict the possibility of occurrence of food accident damage by fishery products. [Solution] An information processing device comprising: an information acquisition unit 400 which acquires aquaculture condition information that includes water area environment information indicating the environment of an aquaculture water area in which a fishery product subject to prediction is cultivated; and a generation processing unit 401 which generates fishery product hazard information in the fishery product subject to prediction by inputting the aquaculture condition information that has been acquired by the information acquisition unit 400 into a trained model 12A which has learned, by machine learning, the correlation between the aquaculture condition information in the fishery product subject to learning and fishery product hazard information that indicates the abundance of one or more types of hazardous substances contained in the fishery product.

Description

Information processing device, inference device, machine learning device, information processing method, inference method, and machine learning method

[0001] The present invention relates to an information processing device, an inference device, a machine learning device, an information processing method, an inference method, and a machine learning method.

[0002] Conventionally, various devices and implements have been developed for cultivating marine products such as shellfish, seaweed, fish, and crustaceans in water areas such as the sea, rivers, lakes, and marshes (see, for example, Patent Document 1 and Patent Document 2).

[0003] JP 2005-333910 A JP 2006-020570 A

[0004] When seafood is eaten raw, if it contains or is contaminated with harmful substances such as bacteria or viruses, there is a possibility that a food accident (food poisoning) that has adverse effects on the human body may occur. However, because seafood is farmed in a variety of waters and environments, it has been difficult to determine the extent to which harmful substances are contained in or attached to each farmed seafood product when it is caught.

[0005] In view of the above-mentioned problems, the present invention aims to provide an information processing device, an inference device, a machine learning device, an information processing method, an inference method, and a machine learning method that make it possible to easily predict the possibility of food-related harm caused by seafood products.

[0006] In order to achieve the above-mentioned object, an information processing device according to one aspect of the present invention comprises: an information acquisition unit that acquires aquaculture condition information including water area environmental information that indicates the environment of the aquaculture waters in which the aquatic product to be predicted is cultivated; and a generation processing unit that generates the aquaculture condition information for the aquatic product to be predicted by inputting the aquaculture condition information acquired by the information acquisition unit into a learning model that has been trained by machine learning to determine the correlation between the aquaculture condition information for the aquatic product to be learned and aquatic product harm information that indicates the presence of one or more types of hazard-causing substances contained in the aquatic product.

[0007] According to an information processing device of one aspect of the present invention, the presence of a hazardous substance in a fishery product to be predicted is predicted based on fishery product hazard information output by inputting culture condition information for the fishery product to be predicted into a learning model. Therefore, it is possible to easily predict the possibility of a fishery product causing food harm.

[0008] Problems, configurations, and effects other than those described above will become apparent from the detailed description of the invention that follows.

[0009] 1 is an overall configuration diagram showing an example of an aquaculture management system 1A according to a first embodiment. FIG. 2 is a data configuration diagram showing an example of an aquaculture management database 22A according to the first embodiment. FIG. 3 is a hardware configuration diagram showing an example of a computer 900. FIG. 4 is a block diagram showing an example of a machine learning device 3A according to the first embodiment. FIG. 5 is a diagram showing examples of a learning model 12A and learning data 13A according to the first embodiment. FIG. 6 is a flowchart showing an example of a machine learning method by the machine learning device 3A. FIG. 7 is a block diagram showing an example of an information processing device 4A according to the first embodiment. FIG. 8 is a functional explanatory diagram showing an example of an information processing method by the information processing device 4A. FIG. 9 is an overall configuration diagram showing an example of an aquaculture management system 1B according to a second embodiment. FIG. 10 is a data configuration diagram showing an example of an aquaculture management database 22B according to the second embodiment. FIG. 11 is a block diagram showing an example of a machine learning device 3B according to the second embodiment. FIG. 12 is a diagram showing examples of a learning model 12B and learning data 13B according to the second embodiment. FIG. 13 is a block diagram showing an example of an information processing device 4B according to the second embodiment. FIG. 14 is a functional explanatory diagram showing an example of an information processing device 4B according to the second embodiment. FIG. 15 is an overall configuration diagram showing an example of an aquaculture management system 1C according to a third embodiment. 10 is a data structure diagram showing an example of an aquaculture management database 22C according to the third embodiment. FIG. 11 is a data structure diagram showing an example of an aquaculture management database 22C according to the third embodiment. FIG. 12 is a block diagram showing an example of a machine learning device 3C according to the third embodiment. FIG. 13 is a diagram showing an example of a learning model 12C and learning data 13C according to the third embodiment. FIG. 14 is a block diagram showing an example of an information processing device 4C according to the third embodiment. FIG. 15 is a functional explanatory diagram showing an example of an information processing device 4C according to the third embodiment.

[0010] Hereinafter, an embodiment for carrying out the present invention will be described with reference to the drawings. The scope necessary for the explanation to achieve the object of the present invention will be schematically shown, and the scope necessary for explaining the relevant part of the present invention will be mainly explained, and the parts that are omitted from the explanation will be based on publicly known techniques.

[0011] 1 is an overall configuration diagram showing an example of a seafood farming management system 1A according to a first embodiment. The seafood farming management system 1A according to this embodiment functions as a system that manages seafood 11 farmed in a culture area 10 and predicts the possibility of a food-related accident caused by the seafood 11 based on the culture conditions under which the seafood 11 was farmed in the culture area 10.

[0012] The aquaculture water area 10 is a water area for cultivating various marine products 11. The aquaculture water area 10 may be, for example, a seawater area or a freshwater area. The aquaculture water area 10 is not limited to a natural water area, and may also be an artificial water area. The aquaculture water area 10 may be managed, for example, in sections separated by a predetermined range, or the entire area may be managed collectively.

[0013] The marine product 11 is a shellfish, and examples thereof include bivalve shells such as oysters, clams, and scallops, and gastropods such as abalone. In addition to the shellfish, the marine product 11 may also be seaweed such as wakame seaweed, fish such as yellowtail, or crustaceans such as kuruma shrimp.

[0014] Hazardous substances are substances that can cause food accidents that have adverse effects on the human body. Food accidents occur, for example, when seafood 11 that contains or is attached to a hazardous substance is eaten raw. Hazardous substances can be evaluated by detecting the presence of specific factors in seafood 11, and specific factors include, for example, bacteria and viruses. Specific examples of bacteria include general viable bacteria, coliform bacteria, Escherichia coli (E. coli), and Vibrio parahaemolyticus. Specific examples of viruses include norovirus. Note that hazardous substances are not limited to the above examples, as long as they can cause food accidents.

[0015] The seafood farming management system 1A predicts the possibility of a food accident occurring when the seafood 11 to be predicted is eaten raw, based on the farming condition information when the seafood 11 to be predicted is farmed in the farming waters 10.

[0016] As shown in Fig. 1, the aquaculture management system 1A mainly comprises an aquaculture management device 2A, a machine learning device 3A, an information processing device 4A, and an operator terminal device 5. Each of the devices 2A to 5 is configured, for example, as a general-purpose or dedicated computer (see Fig. 3 described below), and is connected to a wired or wireless network 6 so as to be able to transmit and receive various data to and from each other. Note that the number of the devices 2A to 5 and the connection configuration of the network 6 are not limited to the example shown in Fig. 1 and may be changed as appropriate.

[0017] The aquaculture management device 2A includes an environmental sensor group 20 used to monitor the environment of the aquaculture water area 10, a hazardous substance inspection machine 21 used to inspect the marine products 11 as samples for hazardous substances contained in the samples, and an aquaculture management database 22A for managing aquaculture condition information indicating the aquaculture conditions for the marine products 11 in the aquaculture water area 10. The aquaculture management device 2A may also be one that manages aquaculture water areas 10 in multiple locations.

[0018] The environmental sensor group 20 consists of a meteorological environment sensor that detects the meteorological environment of the aquaculture water area 10, a water quality environment sensor that detects the water quality environment of the aquaculture water area 10, and a wave environment sensor that detects the wave environment of the aquaculture water area 10. The detection results by the environmental sensor group 20 are acquired as water area environment information. The detection results by the meteorological environment sensor are acquired as meteorological environment information. The detection results by the water quality environment sensor are acquired as water quality environment information. The detection results by the wave environment sensor are acquired as wave environment information.

[0019] The meteorological environment sensor detects meteorological factors that represent the meteorological environment, such as weather, temperature, humidity, wind direction, wind speed, precipitation, and amount of sunlight. The water quality environment sensor detects water quality factors that represent the water quality environment, such as water surface temperature, underwater water temperature, dissolved oxygen content, chlorophyll-a concentration, pH value, salinity concentration, water level, and river flow rate. The wave environment sensor detects wave height, current direction, current speed, and other wave factors that represent the wave environment.

[0020] The environmental sensor group 20 may be installed at various locations (or multiple locations) and using various detection methods depending on the object to be detected. For example, an image sensor (camera) or a weather satellite may be used. The environmental sensor group 20 may be installed not only in the aquaculture water area 10, but also in the surrounding environment at a predetermined distance from the aquaculture water area 10 (or in multiple locations at different distances or in different directions). In this case, the environmental condition may be detected by the environmental sensor group 20 installed in the surrounding environment. Furthermore, instead of or in addition to being acquired by the environmental sensor group 20, the aquatic environment information may be provided and acquired from an information provider managed by an external business operator.

[0021] The hazardous substance inspection machine 21 is a device that is composed of, for example, a measuring device, an inspection kit, etc., and is capable of inspecting the presence of one or more types of hazardous substances. If the hazardous substance inspection machine 21 is configured to be able to output the inspection results of hazardous substances as data, seafood hazard information including the inspection results of the hazardous substances is registered in the aquaculture management database 22A. If the hazardous substance inspection machine 21 does not have the function of outputting the above-mentioned seafood hazard information, the inspection worker performs an input operation to input the inspection results of hazardous substances into the worker terminal device 5, and the seafood hazard information based on the input operation is registered in the aquaculture management database 22A.

[0022] The hazardous substance testing machine 21 may perform any of the testing methods of morphological testing, biochemical property testing, serological testing, and genetic testing, or may perform a combination of multiple testing methods. For example, the hazardous substance testing machine 21 may use tissue collected from the marine product 11 as a sample and perform flora analysis or genetic analysis of hazardous substances contained in the sample. For example, next-generation sequencers (NGS) or real-time PCR (RT-PCR) are used for flora analysis. Furthermore, although the hazardous substance testing machine 21 is designed to test for a specific type of hazardous substance, multiple types of hazardous substance testing machines 21 with different test targets and testing methods may be used.

[0023] The aquaculture management database 22A stores the detection results of the environmental sensor group 20 (aquatic environment information), the inspection results of hazardous substances by the hazardous substance inspection device 21 (harmful product information), etc. Details of the aquaculture management database 22A will be described later.

[0024] The machine learning device 3A is a device that operates as the main subject of the learning phase of machine learning. The machine learning device 3A acquires training data 13A from, for example, the aquaculture management database 22A, and generates a training model 12A to be used in the information processing device 4A by machine learning based on the training data 13A. The trained training model 12A is provided to the information processing device 4A via the network 6, a recording medium, or the like.

[0025] The information processing device 4A is a device that operates as the main subject of the inference phase of machine learning. The information processing device 4A uses the learning model 12A generated by the machine learning device 3A to predict the possibility of a food accident occurring when the prediction target seafood product 11 is eaten raw, based on the culture condition information of the prediction target seafood product 11. The seafood damage information as the prediction result is provided to, for example, the operator terminal device 5 and presented to an inspection operator or aquaculture farm manager. The seafood damage information may also be provided to the aquaculture management device 2A and accumulated in the aquaculture management database 22A.

[0026] The worker terminal device 5 is a terminal device used by workers (aquaculture workers, inspection workers, aquaculture farm managers, test workers, etc.) working in the aquaculture waters 10, and may be a stationary device or a portable device. The worker terminal device 5 accepts various input operations via a display screen or reader / writer such as an application program or a web browser, and displays various information on the display screen.

[0027] Fig. 2 is a data structure diagram showing an example of the aquaculture management database 22A according to the first embodiment. Note that each table also registers the time when each piece of information was acquired, but the time-related fields are omitted in Fig. 2. Fig. 2 also shows examples of registration in which a value normalized to the range of 0 to 1 is registered in each table field, or a class classified as C1, C2, ..., Cn, etc. is registered.

[0028] The aquaculture management database 22A is a database that manages the aquaculture conditions for the aquaculture products 11 cultivated in the aquaculture waters 10 by type and management unit of the aquaculture products 11. The aquaculture management database 22A has, for example, a aquaculture product table 220, an aquaculture area environment table 221, and a aquaculture product damage table 222, and is managed based on a aquaculture product ID that identifies the aquaculture product 11.

[0029] The seafood table 220 is configured, for example, by aquaculture workers or aquaculture farm managers using the worker terminal device 5 to register the types of seafood 11 and the divisions (management units) of the aquaculture waters 10 .

[0030] The water area environment table 221 registers meteorological environment information, water quality environment information, and wave environment information as water area environment information that is the detection result of the environmental sensor group 20. Note that the water area environment table 221 may also register water area environment information provided by an external information providing device.

[0031] The fishery product hazard table 222 registers the presence of one or more types of hazardous substances (three types in the example of FIG. 2) as the inspection results of the hazardous substance inspection device 21.

[0032] (Hardware Configuration of Each Device) Fig. 3 is a hardware configuration diagram showing an example of the computer 900. Each of the devices 2A to 5 of the aquaculture management system 1A is configured by a general-purpose or dedicated computer 900.

[0033] 3, the computer 900 includes, as its main components, a bus 910, a processor 912, a memory 914, an input device 916, an output device 917, a display device 918, a storage device 920, a communication I / F (interface) unit 922, an external device I / F unit 924, an I / O (input / output) device I / F unit 926, and a media input / output unit 928. Note that the above components may be omitted as appropriate depending on the application of the computer 900.

[0034] The processor 912 is composed of one or more arithmetic processing devices (such as a CPU (Central Processing Unit), an MPU (Micro-processing unit), a DSP (Digital Signal Processor), or a GPU (Graphics Processing Unit)), and operates as a control unit that controls the entire computer 900. The memory 914 stores various data and programs 930, and is composed of, for example, a volatile memory (DRAM, SRAM, etc.) that functions as a main memory, a non-volatile memory (ROM), a flash memory, etc.

[0035] The input device 916 is composed of, for example, a keyboard, a mouse, a numeric keypad, an electronic pen, a microphone, etc., and functions as an input unit. The output device 917 is composed of, for example, a sound (audio) output device, a vibration device, etc., and functions as an output unit. The display device 918 is composed of, for example, a liquid crystal display, an organic EL display, electronic paper, a projector, etc., and functions as an output unit. The input device 916 and the display device 918 may be integrated, such as a touch panel display. The storage device 920 is composed of, for example, an HDD, an SSD, etc., and functions as a storage unit. The storage device 920 stores various data necessary for executing the operating system and the program 930.

[0036] The communication I / F unit 922 is connected to a network 940 such as the Internet or an intranet (which may be the same as the network 6 in FIG. 1 ) via a wired or wireless connection and functions as a communication unit that transmits and receives data to and from other computers in accordance with a predetermined communication standard. The external device I / F unit 924 is connected to an external device 950 such as a camera, printer, scanner, or reader / writer via a wired or wireless connection and functions as a communication unit that transmits and receives data to and from the external device 950 in accordance with a predetermined communication standard. The I / O device I / F unit 926 is connected to I / O devices 960 such as various sensors and actuators and functions as a communication unit that transmits and receives various signals and data, such as detection signals from sensors and control signals to actuators, to and from the I / O devices 960. The media input / output unit 928 is composed of a drive device such as a DVD drive or a CD drive and reads and writes data from and to media (non-transitory storage media) 970 such as DVDs and CDs.

[0037] In the computer 900 having the above configuration, the processor 912 loads a program 930 stored in the storage device 920 into the memory 914, executes the program, and controls each unit of the computer 900 via the bus 910. The program 930 may be stored in the memory 914 instead of the storage device 920. The program 930 may be recorded on the medium 970 in an installable file format or an executable file format and provided to the computer 900 via the media input / output unit 928. The program 930 may be provided to the computer 900 by downloading it via the communication I / F unit 922 over the network 940. Furthermore, the computer 900 may implement various functions realized by the processor 912 executing the program 930 using hardware such as an FPGA or an ASIC.

[0038] The computer 900 is an electronic device of any type, such as a desktop computer or a portable computer. The computer 900 may be a client computer, a server computer, a cloud computer, or an embedded computer such as a control panel or a controller (including a microcomputer, a programmable logic controller, or a sequencer).

[0039] 4 is a block diagram showing an example of a machine learning device 3A according to embodiment 1. The machine learning device 3A includes a control unit 30, a communication unit 31, a learning data storage unit 32, and a trained model storage unit 33.

[0040] The control unit 30 functions as a learning data acquisition unit 300 and a machine learning unit 301. The communication unit 31 is connected to an external device via the network 6 and functions as a communication interface for transmitting and receiving various types of data.

[0041] The learning data acquisition unit 300 is connected to an external device (e.g., the aquaculture management device 2A, the operator terminal device 5, etc.) via the communication unit 31 and the network 6, and acquires learning data 13A consisting of aquaculture condition information as input data and marine product damage information as output data. The learning data 13A is data used as teacher data (training data), verification data, and test data in supervised learning. The marine product damage information is data used as a correct answer label in supervised learning.

[0042] The learning data storage unit 32 is a database that stores a plurality of sets of learning data 13A acquired by the learning data acquisition unit 300. The specific configuration of the database that constitutes the learning data storage unit 32 may be designed as appropriate.

[0043] The machine learning unit 301 performs machine learning using multiple sets of training data 13A stored in the training data storage unit 32. That is, the machine learning unit 301 inputs multiple sets of training data 13A to the training model 12A and causes the training model 12A to learn the correlation between the input data (aquaculture condition information) and output data (fishery product damage information) that constitute the training data 13A, thereby generating a trained training model 12A. Note that the machine learning unit 301 may perform predetermined preprocessing on the input data (aquaculture condition information) to be input to the training model 12A, or may perform predetermined postprocessing on the output data (fishery product damage information) output from the training model 12A.

[0044] The trained model storage unit 33 is a database that stores the trained learning model 12A (specifically, the adjusted weight parameter group) generated by the machine learning unit 301. The trained learning model 12A stored in the trained model storage unit 33 is provided to an actual system (e.g., an information processing device 4A) via the network 6, a recording medium, or the like. Note that although the training data storage unit 32 and the trained model storage unit 33 are shown as separate storage units in FIG. 4, they may be configured as a single storage unit.

[0045] 5 is a diagram showing an example of the learning model 12A and the learning data 13A according to the first embodiment. The learning data 13A used for machine learning of the learning model 12A is composed of input data and output data.

[0046] The input data constituting the learning data 13A is aquaculture condition information including water area environmental information that indicates the environment of the aquaculture water area 10 where the aquatic product 11 to be learned is cultivated. The water area environmental information includes at least one of meteorological environment information, water quality environment information, and wave environment information of the aquaculture water area 10. It is preferable that the water area environmental information include at least meteorological environment information, next preferably at least water quality environment information, next preferably at least wave environment information, and next most preferably all of them.

[0047] The meteorological environment information includes at least one meteorological factor among weather, temperature, humidity, wind direction, wind speed, precipitation, and amount of sunlight. The water quality environment information includes at least one water quality factor among water surface temperature, underwater water temperature, dissolved oxygen content, chlorophyll a concentration, pH value, salinity concentration, water level, and river flow rate. The river flow rate is the flow rate of the river flowing into the aquaculture water area 10. The wave environment information includes at least one wave factor among wave height, current direction, and current speed. The underwater water temperature is the water temperature at a predetermined depth, and may be multiple water temperatures at multiple depths.

[0048] The weather factors included in the weather environment information, the water quality factors included in the water quality environment information, and the wave factors included in the wave environment information may be point-in-time data indicating the state at a predetermined point in time, or may be time-series data consisting of data at multiple points in time at predetermined time intervals (every hour, every day, every week, every month, etc.). The weather factors included in the weather environment information, the water quality factors included in the water quality environment information, and the wave factors included in the wave environment information may be statistical values ​​such as average values, maximum values, and minimum values ​​over a predetermined period of time, or may be statistical values ​​such as average values, maximum values, and minimum values ​​at multiple detection points within the aquaculture water area 10.

[0049] The output data constituting the learning data 13A is seafood hazard information indicating the presence of one or more types of hazardous substances contained in the target seafood product 11. The seafood hazard information is at least one of the presence of general viable bacteria, the presence of Escherichia coli, the presence of coliform bacteria, the presence of Vibrio parahaemolyticus, and the presence of norovirus contained in the seafood product 11.

[0050] The presence of a hazardous substance is expressed, for example, by the presence or absence of the hazardous substance, its presence level, the number of individuals, etc. The presence level is a graded value or a continuous value, and in the case of a continuous value, it may be a value normalized to a predetermined range (for example, 0 to 1).

[0051] The learning data acquisition unit 300 acquires the learning data 13A by referring to various information registered in the aquaculture management database 22A when the marine product 11 was cultivated in the past, or by accepting input operations from the operator via the operator terminal device 5.

[0052] When the learning data acquisition unit 300 refers to the aquaculture management database 22A, it acquires the learning data 13A by, for example, acquiring the aquaculture condition information and the aquaculture product damage information associated with a specific aquaculture product ID in the water area environment table 221 and the aquaculture product damage table 222. Note that if there are multiple aquaculture water areas 10, the learning data 13A may be information acquired when the aquaculture products 11 are cultivated under various aquaculture conditions in each aquaculture water area 10. Furthermore, the learning data 13A may be information acquired when the aquaculture products 11 are cultivated under various aquaculture conditions in a test water area simulating the aquaculture water area 10.

[0053] The learning model 12A employs, for example, a neural network structure and includes an input layer 120, an intermediate layer 121, and an output layer 122. Synapses (not shown) that connect each neuron are established between each layer, and each synapse is associated with a weight. A group of weight parameters consisting of the weights of each synapse is adjusted by machine learning.

[0054] The input layer 120 has neurons whose number corresponds to the aquaculture condition information as input data, and each value of the aquaculture condition information is input to each neuron. The output layer 122 has neurons whose number corresponds to the seafood hazard information as output data, and prediction results (inference results) of the seafood hazard information for the aquaculture condition information are output as output data. When the learning model 12A is configured as a regression model, the seafood hazard information is output as a numerical value normalized to a predetermined range (e.g., 0 to 1). When the learning model 12A is configured as a classification model, the seafood hazard information is output as a score (accuracy) for each class as a numerical value normalized to a predetermined range (e.g., 0 to 1).

[0055] 5, a plurality of data configurations with different conditions may be employed, such as the machine learning method, the type of marine product 11, the type of hazardous substance inspection device 21 (inspection target and inspection method), the type of data included in the aquaculture condition information, the type of data included in the marine product hazard information, etc. In this case, the learning data acquisition unit 300 acquires a plurality of types of learning data 13A corresponding to the plurality of data configurations with different conditions, and the machine learning unit 301 performs machine learning using each of these pieces of learning data 13A.

[0056] (Machine Learning Method) FIG. 6 is a flowchart showing an example of a machine learning method performed by the machine learning device 3A.

[0057] First, in step S100, the learning data acquisition unit 300 acquires a desired number of pieces of learning data 13A as advance preparation for starting machine learning, and stores the acquired learning data 13A in the learning data storage unit 32. The number of pieces of learning data 13A to be prepared here may be set in consideration of the inference accuracy required for the learning model 12A to be finally obtained.

[0058] Next, in step S110, the machine learning unit 301 prepares a pre-learning learning model 12A to start machine learning. The pre-learning learning model 12A prepared here is configured with the neural network model exemplified in Fig. 5, and the weights of each synapse are set to initial values.

[0059] Next, in step S120 , the machine learning unit 301 randomly acquires, for example, one set of learning data 13A from the multiple sets of learning data 13A stored in the learning data storage unit 32 .

[0060] Next, in step S130, the machine learning unit 301 inputs the aquaculture condition information (input data) included in one set of training data 13A to the input layer 120 of the prepared pre-training (or training) training model 12A. As a result, seafood hazard information (output data) is output as an inference result from the output layer 122 of the training model 12A, and this output data was generated by the pre-training (or training) training model 12A. Therefore, in the pre-training (or training) state, the output data output as an inference result indicates information that differs from the seafood hazard information (correct label) included in the training data 13A.

[0061] Next, in step S140, the machine learning unit 301 performs machine learning by comparing the seafood damage information (correct label) included in the set of learning data 13A acquired in step S120 with the seafood damage information (output data) output as an inference result from the output layer 122 in step S130, and performing a process of adjusting the weight of each synapse (backpropagation). In this way, the machine learning unit 301 causes the learning model 12A to learn the correlation between the aquaculture condition information and the seafood damage information.

[0062] Next, in step S150, the machine learning unit 301 determines whether a specified learning termination condition has been met, for example, based on the evaluation value of an error function based on the seafood hazard information (correct label) contained in the learning data 13A and the seafood hazard information (output data) output as an inference result, or the remaining number of unlearned learning data 13A stored in the learning data storage unit 32.

[0063] In step S150, if the machine learning unit 301 determines that the learning termination condition is not satisfied and that machine learning should be continued (No in step S150), the process returns to step S120, and the processes of steps S120 to S140 are performed multiple times on the learning model 12A under training using unlearned learning data 13A. On the other hand, in step S150, if the machine learning unit 301 determines that the learning termination condition is satisfied and that machine learning should be terminated (Yes in step S150), the process proceeds to step S160.

[0064] Then, in step S160, the machine learning unit 301 stores the trained learning model 12A (adjusted weight parameter group) generated by adjusting the weights associated with each synapse in the trained model storage unit 33, thereby completing the series of machine learning methods shown in Fig. 6. In the machine learning method, step S100 corresponds to a training data storage step, steps S110 to S150 correspond to a machine learning step, and step S160 corresponds to a trained model storage step.

[0065] As described above, the machine learning device 3A and machine learning method of this embodiment can provide a learning model 12A that can predict (infer) seafood hazard information for the seafood 11 from the cultivation condition information for the seafood 11.

[0066] (Information processing device 4A) Fig. 7 is a block diagram showing an example of the information processing device 4A according to the first embodiment. Fig. 8 is a functional explanatory diagram showing an example of the information processing device 4A according to the first embodiment. The information processing device 4A includes a control unit 40, a communication unit 41, and a trained model storage unit 42.

[0067] The control unit 40 functions as an information acquisition unit 400, a generation processing unit 401, and an output processing unit 402. The communication unit 41 is connected to an external device via the network 6 and functions as a communication interface for transmitting and receiving various types of data.

[0068] The information acquisition unit 400 is connected to an external device (e.g., an aquaculture management device 2A, an operator terminal device 5, etc.) via the communication unit 41 and the network 6, and acquires aquaculture condition information including water area environmental information indicating the environment of the aquaculture water area 10 in which the predicted aquatic product 11 is cultivated.

[0069] For example, the information acquisition unit 400 acquires the aquaculture condition information by receiving, from the environmental sensor group 20, aquatic environment information at the time when the aquatic product 11 to be predicted was cultivated as a detection result by the environmental sensor group 20. Furthermore, if the aquaculture condition information for the aquatic product 11 to be predicted is registered in the aquaculture management database 22A, the information acquisition unit 400 acquires the aquaculture condition information by referring to the aquaculture management database 22A. In this case, the information acquisition unit 400 may acquire the aquaculture condition information in accordance with the data structure of the input data in the learning model 12A. Note that, when acquiring time-series data of the aquatic environment information, the information acquisition unit 400 may acquire the current aquatic environment information from the environmental sensor group 20 and acquire the past aquatic environment information from the aquaculture management database 22A.

[0070] As described above, the generation processing unit 401 generates seafood damage information for the seafood 11 based on the seafood damage information output by inputting the culture condition information for the seafood 11 to be predicted, acquired by the information acquisition unit 400, into the learning model 12A as input data. Note that the generation processing unit 401 may perform predetermined pre-processing on the input data (culture condition information) to be input into the learning model 12A, or may perform predetermined post-processing on the output data (fish damage information) output from the learning model 12A.

[0071] The trained model storage unit 42 is a database that stores trained learning models 12A used by the generation processing unit 401. Note that the number of learning models 12A stored in the trained model storage unit 42 is not limited to the above example, and multiple trained models with different conditions, such as machine learning method, type of marine product 11, type of hazardous substance inspection device 21 (inspection target and inspection method), type of data included in the aquaculture condition information, type of data included in the marine product hazard information, etc., may be stored and used selectively or in parallel. Furthermore, the trained model storage unit 42 may be substituted by a storage unit of an external computer (e.g., a server-type computer or a cloud-type computer), in which case the generation processing unit 401 simply needs to access the external computer.

[0072] The output processing unit 402 performs output processing for outputting the seafood hazard information generated by the generation processing unit 401. For example, the output processing unit 402 may transmit the seafood hazard information to the operator terminal device 5, thereby causing a display screen based on the seafood hazard information to be displayed on the operator terminal device 5, or may transmit the seafood hazard information to the aquaculture management device 2A, thereby registering the seafood hazard information in the aquaculture management database 22A.

[0073] 9 is a flowchart showing an example of an information processing method by the information processing device 4 A. The following describes an example of the operation of the information processing device 4 A when an inspector checks the possibility of a food accident occurring for a fishery product 11 at the catch stage as the target fishery product 11 without using a hazard-causing substance inspection device 21.

[0074] First, in step S200, the information acquisition unit 400 receives a seafood ID that identifies the seafood 11 that is the prediction target from the operator terminal device 5 used by the inspection operator. Then, the information acquisition unit 400 receives detection results from the environmental sensor group 20 that monitors the seafood 11 that is the prediction target identified by the seafood ID, and acquires aquaculture condition information, including water area environmental information, for the seafood 11 that is the prediction target, by referring to the aquaculture management database 22A.

[0075] Next, in step S210, the generation processing unit 401 inputs the aquaculture condition information acquired in step S200 into the learning model 12A, thereby generating fishery product hazard information for the aquaculture condition information as output data, and predicts the presence of hazard-causing substances contained in the fishery product 11 to be predicted.

[0076] Next, in step S220, the output processing unit 402 transmits the seafood hazard information generated in step S210 to the worker terminal device 5 used by the inspector, as output processing for outputting the seafood hazard information. The worker terminal device 5 then displays a display screen based on the seafood hazard information, thereby presenting the seafood hazard information for that seafood 11 to the inspector. The destination of the seafood hazard information may be the aquaculture management device 2A in addition to or instead of the worker terminal device 5. In the above information processing method, step S200 corresponds to an information acquisition step, step S210 corresponds to a generation processing step, and step S220 corresponds to an output processing step.

[0077] As described above, according to the information processing device 4A and information processing method of this embodiment, by inputting the cultivation condition information for the fishery product 11 to be predicted into the learning model 12A, the presence of hazard-causing substances contained in the fishery product 11 to be predicted is predicted as fishery product hazard information for the fishery product 11 to be predicted.

[0078] The growth conditions of the marine products 11 are affected by the meteorological environment, water quality environment, and wave environment of the aquaculture waters 10. For example, if the meteorological environment, water quality environment, and wave environment are suitable for the growth of the marine products 11, the growth conditions of the marine products 11 will be good and the presence of hazardous substances will be reduced. Therefore, the presence of hazardous substances in the marine products 11 will vary depending on fluctuations in the meteorological environment information, water quality environment information, and wave environment information included in the aquaculture condition information. Therefore, by using these parameters that affect the growth conditions of the marine products 11 as input data and inferring the presence of hazardous substances as output data, the presence of hazardous substances can be predicted with high accuracy.

[0079] Second Embodiment Fig. 10 is an overall configuration diagram showing an example of a fishery product aquaculture management system 1B according to a second embodiment. The second embodiment differs from the first embodiment in that an aquaculture management device 2B includes a pollution indicator substance tester 23 that uses water collected from an aquaculture water area 10 as a sample and tests for pollution indicator substances contained in the sample. The second embodiment also differs from the first embodiment in that the aquaculture condition information serving as input data constituting the learning data 13B includes water area pollution information that indicates the presence of pollution indicator substances in the aquaculture water area 10. The following describes a machine learning device 3B and an information processing device 4B according to the second embodiment, focusing on differences from the first embodiment.

[0080] The pollution indicator substance is a substance that can serve as an indicator for evaluating the pollution state of the aquaculture water area 10. The pollution indicator substance can be evaluated by detecting the presence of specific factors in the aquaculture water area 10, and specific factors include, for example, bacteria and viruses. Specific examples of bacteria include general viable bacteria, coliform bacteria, Escherichia coli (E. coli), and Vibrio parahaemolyticus. Specific examples of viruses include norovirus. Note that the pollution indicator substance is not limited to the above examples, as long as it serves as an indicator of the pollution state of the aquaculture water area 10. Furthermore, the pollution indicator substance may be the same substance as the hazard-causing substance, or may be a different substance.

[0081] The pollution indicator substance inspection machine 23, like the hazardous substance inspection machine 21, is composed of, for example, a measuring instrument, an inspection kit, etc., and is a device capable of inspecting the presence of one or more types of pollution indicator substances. If the pollution indicator substance inspection machine 23 is configured to be able to output the inspection results of pollution indicator substances as data, water area pollution information including the inspection results of the pollution indicator substances is registered in the aquaculture management database 22B. If the pollution indicator substance inspection machine 23 does not have the function of outputting the water area pollution information as described above, the inspection operator performs an input operation to input the inspection results of the pollution indicator substances into the operator terminal device 5, and the water area pollution information based on the input operation is registered in the aquaculture management database 22B.

[0082] The pollution indicator substance testing machine 23 may perform any of the following testing methods: morphological testing, biochemical property testing, serological testing, and genetic testing, or may perform a combination of multiple testing methods. For example, the pollution indicator substance testing machine 23 may use water collected from the aquaculture water area 10 as a sample and perform flora analysis or genetic analysis of the pollution indicator substances contained in the sample. Although the pollution indicator substance testing machine 23 is designed to test for a specific type of pollution indicator substance, multiple types of pollution indicator substance testing machines 23 with different testing targets and testing methods may be used. The water sample is collected, for example, from a location in the aquaculture water area 10 that is topographically easy to collect. The water sample may be collected not only from the aquaculture water area 10, but also from a related water area related to the aquaculture water area 10. For example, the water sample may be collected from a location that affects the aquaculture water area 10, such as an ocean current or river flowing into the aquaculture water area 10.

[0083] 11 is a data configuration diagram showing an example of the aquaculture management database 22B according to the second embodiment. As in the first embodiment, the aquaculture management database 22B has a marine product table 220, a water area environment table 221, and a marine product damage table 222, and further has a water area pollution table 223.

[0084] The water area pollution table 223 registers the presence of one or more types of pollution indicator substances (three types in the example of FIG. 11) as the inspection results of the pollution indicator substance inspector 23.

[0085] (Machine Learning Device 3B) Fig. 12 is a block diagram showing an example of a machine learning device 3B according to the second embodiment. Fig. 13 is a diagram showing an example of a learning model 12B and learning data 13B according to the second embodiment.

[0086] The input data constituting the training data 13B, like the training data 13A according to the first embodiment, includes water area environmental information and is aquaculture condition information including water area pollution information indicating the presence of one or more types of pollution indicator substances contained in the aquaculture water area 10. The water area pollution information includes at least one water area pollution factor among the presence of general viable bacteria, the presence of E. coli, the presence of coliform bacteria, the presence of Vibrio parahaemolyticus, and the presence of norovirus contained in the aquaculture water area 10 or related water areas.

[0087] The presence of a pollution indicator substance is expressed, for example, by the presence or absence of the pollution indicator substance, its presence level, the number of individuals, etc. The presence level is a graded value or a continuous value, and in the case of a continuous value, it may be a value normalized to a predetermined range (for example, 0 to 1).

[0088] The water pollution information may also include the results of flora analysis or genetic analysis of pollution indicator substances contained in samples of water collected from the aquaculture water area 10 or related water areas.

[0089] The output data constituting the learning data 13B is, like the learning data 13A according to the first embodiment, information on the danger to marine products, and therefore a description thereof will be omitted.

[0090] The learning data acquisition unit 300 acquires the learning data 13B by referring to various information registered in the aquaculture management database 22B when the marine product 11 was cultivated in the past, or by receiving input operations from the operator via the operator terminal device 5. When the learning data acquisition unit 300 refers to the aquaculture management database 22B, it acquires the learning data 13B by, for example, acquiring the aquaculture condition information and marine product damage information associated with a specific marine product ID in the water area environment table 221, the marine product damage table 222, and the water area pollution table 223.

[0091] The machine learning unit 301 inputs multiple sets of learning data 13B into the learning model 12B, and generates a trained learning model 12B by having the learning model 12B learn the correlation between the input data (aquaculture condition information) and output data (fishery product damage information) that make up the learning data 13B.

[0092] As described above, the machine learning device 3B and machine learning method of this embodiment can provide a learning model 12B that can predict (infer) seafood hazard information for the seafood 11 from the cultivation condition information for the seafood 11.

[0093] (Information Processing Device 4B) Fig. 14 is a block diagram showing an example of an information processing device 4B according to the second embodiment. Fig. 15 is a functional explanatory diagram showing an example of the information processing device 4B according to the second embodiment.

[0094] The information acquisition unit 400 acquires the culture condition information by receiving, as a detection result by the environmental sensor group 20, water area environmental information at the time when the prediction target marine product 11 was cultured from the environmental sensor group 20, and by receiving, as an inspection result by the pollution indicator substance inspection machine 23, water area pollution information at the time when the prediction target marine product 11 was cultured from the pollution indicator substance inspection machine 23. Furthermore, when the culture condition information and water area pollution information for the prediction target marine product 11 are registered in the aquaculture management database 22B, the information acquisition unit 400 acquires the culture condition information by referring to the aquaculture management database 22B. Note that the water area pollution information may include the analysis results of flora analysis or genetic analysis of pollution indicator substances contained in a sample of water collected from the aquaculture water area 10 or a related water area.

[0095] As described above, the generation processing unit 401 generates seafood hazard information for the seafood 11 based on the seafood hazard information output by inputting the cultivation condition information for the seafood 11 to be predicted, acquired by the information acquisition unit 400, into the learning model 12B as input data.

[0096] As described above, according to the information processing device 4B and information processing method of this embodiment, by inputting the cultivation condition information for the fishery product 11 to be predicted into the learning model 12B, the presence of hazard-causing substances contained in the fishery product 11 to be predicted is predicted as fishery product hazard information for the fishery product 11 to be predicted.

[0097] The growth conditions of the marine products 11 are affected not only by the meteorological environment, water quality environment, and wave environment of the aquaculture waters 10, but also by the presence of pollution indicator substances contained in the aquaculture waters 10 or related water areas. For example, if the presence of pollution indicator substances in the aquaculture waters 10 or related water areas is high, the presence of hazardous substances in the marine products 11 also increases. Therefore, the presence of hazardous substances in the marine products 11 varies in accordance with fluctuations in the meteorological environment information, water quality environment information, wave environment information, and water area pollution information contained in the aquaculture condition information. Therefore, by using these parameters that affect the growth conditions of the marine products 11 as input data and inferring the presence of hazardous substances as output data, the presence of hazardous substances can be predicted with high accuracy.

[0098] Third Embodiment FIG. 16 is an overall configuration diagram showing an example of a fishery product aquaculture management system 1C according to a third embodiment. The third embodiment differs from the first embodiment in that an aquaculture management device 2C is connected to an information providing device 7 managed by an external business operator. The third embodiment also differs from the first embodiment in that the aquaculture condition information serving as input data constituting the learning data 13C includes at least one of community infection status information indicating the occurrence of food-related accidents in an area (city) that discharges sewage into the aquaculture water area 10, and animal appearance status information indicating the appearance of animals in the aquaculture water area 10 or related water areas associated with the aquaculture water area 10. The following describes a machine learning device 3C and an information processing device 4C according to the third embodiment, focusing on differences from the first embodiment.

[0099] The information providing device 7 is a device managed by a business operator such as a public institution or a private institution, and provides community infection status information and animal inhabitation information to the aquaculture management device 2C. The community infection status information and animal inhabitation information provided from the information providing device 7 via the network 6 are registered in the aquaculture management database 22C.

[0100] The community infection status information includes the infection level for each type of hazardous substance as a food accident occurrence status in the community. The infection level is determined, for example, based on the number of infected people per unit population in a specified period. The infection level is a graded value or a continuous value, and if it is a continuous value, it may be a value normalized to a specified range (e.g., 0 to 1).

[0101] The animal appearance status information includes an appearance level for each type of animal as the appearance status of the animal. Examples of animal types include mammals such as cats, and birds such as ducks, swans, and crows. The appearance level is determined, for example, based on the number of sightings in a specified area during a specified period of time. The appearance level is a graded value or a continuous value, and if it is a continuous value, it may be a value normalized to a specified range (for example, 0 to 1).

[0102] 17A and 17B are data configuration diagrams showing an example of an aquaculture management database 22C according to the third embodiment. As in the first embodiment, the aquaculture management database 22C includes a seafood table 220, an aquatic environment table 221, and a seafood harm table 222, and further includes a community infection status table 224 and an animal appearance status table 225.

[0103] The community infection status table 224 registers the infection levels of one or more types of hazardous agents (three types in the example of FIG. 17B ) as community infection status information provided by the information providing device 7. The animal appearance status table 225 registers the appearance levels of one or more types of animals (three types in the example of FIG. 17B ) as animal appearance status information provided by the information providing device 7.

[0104] (Machine Learning Device 3C) Fig. 18 is a block diagram showing an example of a machine learning device 3C according to the third embodiment. Fig. 19 is a diagram showing an example of a learning model 12C and learning data 13C according to the third embodiment.

[0105] The input data constituting the training data 13C, like the training data 13A according to the first embodiment, includes aquatic environment information and is aquaculture condition information including at least one of community infection status information and animal occurrence status information. The community infection status information includes the infection level for each type of hazardous agent. The animal occurrence status information includes the occurrence level for each type of animal.

[0106] The output data constituting the learning data 13C is, like the learning data 13A according to the first embodiment, information on the danger to marine products, and therefore a description thereof will be omitted.

[0107] The learning data acquisition unit 300 acquires the learning data 13C by referring to various information registered in the aquaculture management database 22C when the marine product 11 was cultivated in the past, or by receiving input operations from the operator via the operator terminal device 5. When the learning data acquisition unit 300 refers to the aquaculture management database 22C, it acquires the learning data 13C by acquiring, for example, the aquaculture condition information and marine product harm information associated with a specific marine product ID in the aquatic environment table 221, the marine product harm table 222, the community infection status table 224, and the animal appearance status table 225.

[0108] The machine learning unit 301 inputs multiple sets of learning data 13C into the learning model 12C, and generates a trained learning model 12C by having the learning model 12C learn the correlation between the input data (aquaculture condition information) and output data (fishery product damage information) that make up the learning data 13C.

[0109] As described above, the machine learning device 3C and machine learning method of this embodiment can provide a learning model 12C that can predict (infer) seafood hazard information for the seafood 11 from the cultivation condition information for the seafood 11.

[0110] (Information processing device 4C) Fig. 20 is a block diagram showing an example of an information processing device 4C according to the third embodiment. Fig. 21 is a functional explanatory diagram showing an example of the information processing device 4C according to the third embodiment.

[0111] The information acquisition unit 400 acquires the culture condition information by receiving, as a detection result by the environmental sensor group 20, water area environment information at the time when the prediction target marine product 11 was cultured from the environmental sensor group 20, and by receiving, from the information providing device 7, community infection status information and animal appearance status information at the time when the prediction target marine product 11 was cultured. Furthermore, when the culture condition information and water area contamination information for the prediction target marine product 11 are registered in the culture management database 22C, the information acquisition unit 400 acquires the culture condition information by referring to the culture management database 22C.

[0112] As described above, the generation processing unit 401 generates seafood hazard information for the seafood 11 based on the seafood hazard information output by inputting the cultivation condition information for the seafood 11 to be predicted, acquired by the information acquisition unit 400, into the learning model 12C as input data.

[0113] As described above, according to the information processing device 4C and information processing method of this embodiment, by inputting the cultivation condition information for the fishery product 11 to be predicted into the learning model 12C, the presence of hazard-causing substances contained in the fishery product 11 to be predicted is predicted as fishery product hazard information for the fishery product 11 to be predicted.

[0114] For example, if sewage treatment is not performed properly due to heavy rain or the like, hazardous substances contained in the sewage will flow into the aquaculture area 10, and the water quality of the aquaculture area 10 where the marine products 11 are cultivated will be affected by the infection level of the hazardous substances in the city. Furthermore, because animals living in the aquaculture area 10 or related water areas can be carriers of hazardous substances, the water quality of the aquaculture area 10 where the marine products 11 are cultivated will be affected by the occurrence level of animals living in the aquaculture area 10 or related water areas. Therefore, the growth conditions of the marine products 11 and the water quality of the aquaculture area 10 are affected not only by the meteorological environment, water quality environment, and wave environment of the aquaculture area 10, but also by the infection level of hazardous substances in the city and the occurrence level of animals in the aquaculture area 10 or related water areas. Therefore, by using these as input data and inferring the presence level of the hazardous substances as output data, the presence level of the hazardous substances can be predicted with high accuracy.

[0115] (Other Embodiments) The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit and scope of the present invention, and all such modifications are included in the technical concept of the present invention.

[0116] The above embodiments may be combined as appropriate. For example, by combining the second embodiment with the third embodiment, the environmental condition information may include water environment information, water pollution information, and at least one of community infection status information and animal appearance status information. In this case, the data configuration of the learning model and the learning data may be changed as appropriate.

[0117] In the above embodiment, the information processing devices 4A to 4C have been described as predicting seafood hazard information for the time when the aquaculture condition information was acquired (the present time), but they may also be configured to predict seafood hazard information for a time later than the present time (future time). In this case, the information processing devices 4A to 4C may predict seafood hazard information for the present time and future time, or may predict changes in seafood hazard information over time by predicting seafood hazard information for multiple future time points (for example, every week or every month). In this case, the data configurations of the learning models 12A to 12C and the learning data 13A to 13C may be changed as appropriate.

[0118] In the above embodiment, the aquaculture management devices 2A-2C, the machine learning devices 3A-3C, and the information processing devices 4A-4C are described as being configured as separate devices, but these three devices may be configured as a single device, or any two of these three devices may be configured as a single device. Furthermore, at least one of the machine learning devices 3A-3C and the information processing devices 4A-4C may be incorporated into the operator terminal device 5.

[0119] In the above embodiment, a case has been described in which neural networks are used as the learning models 12A to 12C that realize machine learning by the machine learning unit 301. However, other machine learning models may also be used. Examples of other machine learning models include tree-type models such as decision trees and regression trees, ensemble learning such as bagging and boosting, neural network-type models (including deep learning) such as recurrent neural networks, convolutional neural networks, and LSTM, clustering-type models such as hierarchical clustering, non-hierarchical clustering, k-nearest neighbors, and k-means, multivariate analyses such as principal component analysis, factor analysis, and logistic regression, and support vector machines.

[0120] In the above embodiment, the output processing unit 402 of the information processing devices 4A to 4C is described as performing output processing to output the seafood damage information generated by the generation processing unit 401, but the seafood damage information generated by the generation processing unit 401 may be used for any purpose other than the above embodiment. For example, the seafood damage information generated by the generation processing unit 401 may be used to determine processing conditions before shipping the seafood 11.

[0121] Specifically, the control unit 40 of the information processing devices 4A to 4C, like each of the units 400 to 402, functions as a processing condition determination unit (processing condition determination process) that determines processing conditions before shipping based on the seafood damage information generated by the generation processing unit 401.

[0122] The processing condition determination unit determines, as the processing conditions before shipping, whether to perform a purification process in a water tank, a high-pressure process using high-pressure washing water, or a combination of the purification process and the high-pressure process, based on, for example, the presence level of each hazardous substance contained in the marine product 11. In this case, the processing condition determination unit determines, as detailed conditions for the purification process, for example, the purification time depending on the presence level of each hazardous substance, and determines, as detailed conditions for the high-pressure process, for example, the water pressure, water temperature, and washing time of the high-pressure washing water depending on the presence level of each hazardous substance.

[0123] (Machine Learning Program and Information Processing Program) The present invention can also be provided in the form of a program (machine learning program) that causes a computer 900 to function as each unit included in the machine learning devices 3A to 3C, or a program (machine learning program) that causes the computer 900 to execute each step included in a machine learning method. The present invention can also be provided in the form of a program (information processing program) that causes a computer 900 to function as each unit included in the information processing devices 4A to 4C, or a program (information processing program) that causes the computer 900 to execute each step included in the information processing method according to the above embodiment.

[0124] (Inference device, inference method, and inference program) The present invention can be provided not only in the form of the information processing devices 4A to 4C (information processing method or information processing program) according to the above embodiments, but also in the form of an inference device (inference method or inference program) used to infer seafood hazard information. In this case, the inference device (inference method or inference program) can include a memory and a processor, and the processor executes a series of processes. The series of processes includes an information acquisition process (information acquisition step) for acquiring culture condition information for the seafood 11 that is the prediction target, and an inference process (inference step) for inferring seafood hazard information for the seafood 11 once the culture condition information has been acquired in the information acquisition process.

[0125] By providing it in the form of an inference device (inference method or inference program), it can be more easily applied to various devices than when an information processing device is implemented. It will be naturally understood by those skilled in the art that when the inference device (inference method or inference program) infers fishery product hazard information, it may apply the inference method implemented by the generation processing unit using the trained learning model generated by the machine learning device and machine learning method according to the above embodiments.

[0126] Various aspects of the present disclosure are summarized below as appendices.

[0127] (Supplementary Note 1) An information processing device comprising: an information acquisition unit that acquires aquaculture condition information including water area environmental information that indicates the environment of aquaculture waters where a fishery product to be predicted is cultivated; and a generation processing unit that generates the fishery product harm information for the fishery product to be predicted by inputting the aquaculture condition information acquired by the information acquisition unit into a learning model that has learned by machine learning a correlation between the aquaculture condition information for the fishery product to be learned and fishery product harm information that indicates the presence of one or more types of hazard-causing substances contained in the fishery product. (Supplementary Note 2) The information processing device according to Supplementary Note 1, wherein the water area environmental information includes at least one of meteorological environment information that indicates the meteorological environment of the aquaculture waters, water quality environment information that indicates the water quality environment of the aquaculture waters, and wave environment information that indicates the wave environment of the aquaculture waters. (Supplementary Note 3) The information processing device according to Supplementary Note 2, wherein the meteorological environment information includes at least one meteorological factor among weather, temperature, humidity, wind direction, wind speed, precipitation, and amount of sunlight, the water quality environment information includes at least one water quality factor among water surface temperature, underwater water temperature, dissolved oxygen content, chlorophyll-a concentration, pH value, salinity water level, and river flow rate, and the wave environment information includes at least one wave factor among wave height, current direction, and current speed. (Supplementary Note 4) The information processing device according to any one of Supplementary Notes 1 to 3, wherein the culture condition information includes water area pollution information indicating the presence of one or more types of pollution indicator substances contained in the culture water area or related water areas associated with the culture water area. (Supplementary Note 5) The information processing device according to Supplementary Note 4, wherein the water area pollution information includes at least one water area pollution factor among the presence of general viable bacteria in the aquaculture water area, the presence of E. coli in the aquaculture water area, the presence of coliform bacteria in the aquaculture water area, the presence of Vibrio parahaemolyticus in the aquaculture water area, and the presence of norovirus in the aquaculture water area. (Supplementary Note 6) The information processing device according to Supplementary Note 4 or Supplementary Note 5, wherein the water area pollution information includes the analysis results of flora analysis or genetic analysis of the pollution indicator substance contained in a sample of water collected from the aquaculture water area or the related water area.(Supplementary Note 7) The information processing device according to claim 1, wherein the culture condition information includes community infection status information indicating the occurrence status of food accidents in an area that discharges sewage into the culture waters. (Supplementary Note 8) The information processing device according to any one of Supplements 1 to 7, wherein the culture condition information includes animal appearance status information indicating the appearance status of animals in the culture waters or related waters associated with the culture waters. (Supplementary Note 9) The information processing device according to any one of Supplements 1 to 8, wherein the seafood damage information is at least one of the presence of general viable bacteria in the seafood, the presence of Escherichia coli in the seafood, the presence of coliform bacteria in the seafood, the presence of Vibrio parahaemolyticus in the seafood, and the presence of norovirus in the seafood. (Supplementary Note 10) The information processing device according to any one of Supplements 1 to 9, further comprising a processing condition determination unit that determines processing conditions before shipping the seafood based on the seafood damage information generated by the generation processing unit. (Appendix 11) The information processing device described in Appendix 10, wherein the processing condition determination unit determines whether to perform purification processing in a water tank, high-pressure processing using high-pressure cleaning water, or a combination of the purification processing and the high-pressure processing, based on the presence of the hazardous substance contained in the marine product.

[0128] DESCRIPTION OF THE SYMBOLS 1A to 1C...Aquaculture management system, 2A to 2C...Aquaculture management device, 3A to 3C...Machine learning device, 4A to 4C...Information processing device, 5...Worker terminal device, 6...Network, 10...Aquaculture water area, 11...Aquatic product, 12A to 12C...Learning model, 13A to 13C...Learning data, 20...Environmental sensor group, 21...Hazardous substance inspection device, 22A to 22C...Aquaculture management database, 23...Pollution indicator substance inspection device, 30...Control unit, 31...Communication unit, 32...Learning data storage unit, 33...Learned model storage unit, 40...Control unit, 41...Communication unit, 42...Learned model storage unit, 300...Learning data acquisition unit, 301...Machine learning unit, 400...Information acquisition unit, 401...Generation processing unit, 402...Output processing unit, 900...Computer

Claims

1. An information processing device comprising: an information acquisition unit that acquires aquaculture condition information including water area environmental information that indicates the environment of the aquaculture waters where the aquatic product to be predicted is cultivated; and a generation processing unit that generates the aquaculture condition information for the aquatic product to be predicted by inputting the aquaculture condition information acquired by the information acquisition unit into a learning model that has been trained by machine learning to determine the correlation between the aquaculture condition information for the aquatic product to be learned and aquatic product hazard information that indicates the presence of one or more types of hazard-causing substances contained in the aquatic product.

2. The information processing device according to claim 1, wherein the water area environmental information includes at least one of weather environment information indicating the weather environment of the aquaculture water area, water quality environment information indicating the water quality environment of the aquaculture water area, and wave environment information indicating the wave environment of the aquaculture water area.

3. The information processing device of claim 2, wherein the meteorological environment information includes at least one meteorological factor among weather, temperature, humidity, wind direction, wind speed, precipitation, and amount of sunlight; the water quality environment information includes at least one water quality factor among water surface temperature, underwater water temperature, dissolved oxygen content, chlorophyll a concentration, pH value, salinity level, and river flow rate; and the wave environment information includes at least one wave factor among wave height, current direction, and current speed.

4. The information processing device according to claim 1, wherein the aquaculture condition information includes water area pollution information indicating the presence of one or more types of pollution indicator substances contained in the aquaculture water area or related water areas associated with the aquaculture water area.

5. The information processing device according to claim 4, wherein the water pollution information includes at least one water pollution factor from among the presence of general viable bacteria in the aquaculture waters, the presence of Escherichia coli in the aquaculture waters, the presence of coliform bacteria in the aquaculture waters, the presence of Vibrio parahaemolyticus in the aquaculture waters, and the presence of norovirus in the aquaculture waters.

6. The information processing device according to claim 4, wherein the water area pollution information includes the results of flora analysis or genetic analysis of the pollution indicator substances contained in samples of water collected from the aquaculture water area or the related water area.

7. The information processing device according to claim 1, wherein the aquaculture condition information includes community infection status information indicating the occurrence of food accidents in areas that discharge sewage into the aquaculture waters.

8. The information processing device according to claim 1, wherein the aquaculture condition information includes animal appearance status information indicating the appearance status of animals in the aquaculture water area or related water areas associated with the aquaculture water area.

9. The information processing device according to claim 1, wherein the seafood hazard information is at least one of the presence of general viable bacteria in the seafood, the presence of E. coli in the seafood, the presence of coliform bacteria in the seafood, the presence of Vibrio parahaemolyticus in the seafood, and the presence of norovirus in the seafood.

10. The information processing device according to claim 1, further comprising a processing condition determination unit that determines processing conditions for the marine products before shipping based on the marine product damage information generated by the generation processing unit.

11. The information processing device according to claim 10, wherein the processing condition determination unit determines whether to perform purification processing in a water tank, high-pressure processing using high-pressure cleaning water, or a combination of the purification processing and the high-pressure processing, based on the presence of the harmful substances contained in the marine product.

12. An inference device comprising a memory and a processor, wherein the processor executes an information acquisition process to acquire aquaculture condition information including water area environmental information indicating the environment of the aquaculture water area in which the aquatic product to be predicted is cultivated, and an inference process to infer aquatic product hazard information indicating the presence of one or more types of hazard-causing substances contained in the aquatic product to be predicted, once the aquaculture condition information is acquired through the information acquisition process.

13. A machine learning device comprising: a learning data storage unit that stores multiple sets of learning data consisting of input data and output data; a machine learning unit that inputs the multiple sets of learning data into a learning model, thereby causing the learning model to learn a correlation between the input data and the output data; and a trained model storage unit that stores the learning model that has learned the correlation by the machine learning unit, wherein the input data is aquaculture condition information including aquaculture area environmental information that indicates the environment of aquaculture waters in which the aquatic product to be learned is cultivated, and the output data is aquatic product hazard information that indicates the presence of one or more types of hazard-causing substances in the aquatic product to be learned.

14. An information processing method executed by a computer, comprising: an information acquisition step of acquiring aquaculture condition information including aquaculture area environmental information indicating the environment of the aquaculture area in which the aquatic product to be predicted is cultivated; and a generation processing step of generating the aquaculture condition information for the aquatic product to be predicted by inputting the aquaculture condition information acquired by the information acquisition step into a learning model that has been trained by machine learning to determine the correlation between the aquaculture condition information for the aquatic product to be learned and aquatic product hazard information indicating the presence of one or more types of hazard-causing substances contained in the aquatic product.

15. An inference method executed by an inference device having a memory and a processor, wherein the processor executes an information acquisition process to acquire aquaculture condition information including water area environmental information indicating the environment of the aquaculture water area in which the aquatic product to be predicted is cultivated, and an inference process to infer aquatic product hazard information indicating the presence of one or more types of hazard-causing substances contained in the aquatic product to be predicted, once the aquaculture condition information has been acquired through the information acquisition process.

16. A machine learning method executed by a computer, comprising: a learning data storage step of storing multiple sets of learning data consisting of input data and output data; a machine learning step of inputting the multiple sets of learning data into a learning model, thereby causing the learning model to learn the correlation between the input data and the output data; and a trained model storage step of storing the learning model that has learned the correlation by the machine learning step, wherein the input data is aquaculture condition information including aquaculture area environmental information that indicates the environment of the aquaculture area in which the aquatic product to be learned is cultivated, and the output data is aquatic product hazard information that indicates the presence of one or more types of hazard-causing substances contained in the aquatic product to be learned.

Images