How to learn

Abstract

A learning device 100 of the present invention comprises: a classification unit 121 which classifies measurement value data obtained by measuring the performance of an object on the basis of condition data that indicates a condition of the object when the measurement value data was measured; a selection unit 122 which selects the measurement value data from each of the classes according to the number of pieces of the measurement value data for each of the classes; and a learning unit 123 which performs machine learning on the basis of the selected measurement value data.

Description

[Technical field]

[0001] The present invention relates to a learning method, a learning device, and a program adapted to the situation in which the device is placed. [Background technology]

[0002] In systems such as vehicles, computer products, and industrial plants, measurement data detected from various sensors installed in the system is analyzed, and an abnormal state of the system is detected from the analysis results. At this time, the measurement data is analyzed using a model that has been trained using measurement data detected and accumulated in advance from a system in a normal state as training data. In other words, when detecting an abnormal state of a system, it is necessary to generate an anomaly detection model using training data measured in advance from a system operating in a normal state. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-113216 A Summary of the Invention [Problem to be solved by the invention]

[0004] Here, the training data used for learning is measured when the system is operating, but the system operates in various environments. However, the operating conditions of the system differ depending on the environment, and the measurement data in the normal state also differs depending on the environment. For this reason, it is desirable to randomly sample training data from a system that is operating normally in all environments, but in some environments, the occurrence of certain conditions is seldom so that little data can be obtained, making it difficult to extract training data uniformly from all environments. As a result, a problem occurs in that it is not possible to generate an anomaly detection model with high accuracy.

[0005] In Patent Document 1, alerts generated in the system are classified, and the number of training data extracted from each classification is adjusted. However, alerts are generated by training data, and the number of extracted data is simply adjusted according to the content of the training data itself. As a result, there is still a problem that training data cannot be uniformly extracted from each situation in which the system is operating, and an accurate anomaly detection model cannot be generated. In addition, there is a problem that it is difficult to generate a condition detection model for detecting a specific condition, not limited to detecting an abnormal condition.

[0006] Therefore, an object of the present invention is to provide a learning method that can solve the above-mentioned problem of being unable to generate a state detection model with high accuracy. [Means for solving the problem]

[0007] A learning method according to one aspect of the present invention includes: classifying measurement data obtained by measuring the performance of an object based on situation data that represents a situation of the object when the measurement data was measured; Selecting the measurement data from each classification according to the number of measurement data for each classification; Performing machine learning based on the selected measurement data. The structure is as follows.

[0008] Furthermore, the learning device according to one aspect of the present invention comprises: a classification unit that classifies measurement value data obtained by measuring a performance of an object based on situation data that represents a situation of the object when the measurement value data was measured; a selection unit that selects the measurement data from each classification according to the number of the measurement data for each classification; A learning unit that performs machine learning based on the selected measurement data; Equipped with The structure is as follows.

[0009] A program according to one aspect of the present invention includes: In the information processing device, classifying measurement data obtained by measuring the performance of an object based on situation data that represents a situation of the object when the measurement data was measured; Selecting the measurement data from each classification according to the number of measurement data for each classification; Performing machine learning based on the selected measurement data. Execute the process, The structure is as follows. Effect of the Invention

[0010] By being configured as described above, the present invention can generate a highly accurate condition detection model. [Brief description of the drawings]

[0011] [Figure 1] 1 is a block diagram showing a configuration of an information processing device according to a first embodiment of the present invention. [Diagram 2] 2 is a diagram illustrating an example of data processed by the information processing device disclosed in FIG. 1. [Diagram 3] 2 is a diagram showing how data is processed by the information processing device disclosed in FIG. 1; [Figure 4] 2 is a diagram showing how data is processed by the information processing device disclosed in FIG. 1; [Diagram 5] 2 is a diagram showing how data is processed by the information processing device disclosed in FIG. 1; [Figure 6] 2 is a flowchart showing the operation of the information processing device disclosed in FIG. 1. [Figure 7] 2 is a flowchart showing the operation of the information processing device disclosed in FIG. 1. [Figure 8] FIG. 11 is a block diagram showing a hardware configuration of an information processing device according to a second embodiment of the present invention. [Figure 9] FIG. 11 is a block diagram showing a configuration of an information processing device according to a second embodiment of the present invention. [Figure 10] 10 is a flowchart showing the operation of an information processing device according to a second embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0012] <Embodiment 1> A first embodiment of the present invention will be described with reference to Figures 1 to 7. Figures 1 and 2 are diagrams for explaining the configuration of an information processing device, and Figures 3 to 7 are diagrams for explaining the processing operation of the information processing device.

[0013] [composition] The information processing device 10 of the present invention functions as a learning device that learns measurement data measured from an object and generates a condition detection model that detects the state of the object. The information processing device 10 also functions as a condition detection device that detects the state of the object from measurement data newly measured from the object using the generated condition detection model.

[0014] In particular, in this embodiment, the target object is an automobile C. Therefore, the information processing device 10 has a function of learning measurement data that measures the performance of accessories equipped on the automobile C and generating a condition detection model that detects an abnormal condition of the automobile C. The information processing device 10 also has a function of inputting newly measured measurement data from the automobile C to the condition detection model, and detecting whether the automobile C is in an abnormal condition from the output of the model. However, the condition detection model generated by the information processing device 10 is not necessarily limited to detecting an abnormal condition of the automobile C, and may be one that detects any condition of the automobile C. Furthermore, the target object whose condition is detected is not necessarily limited to an automobile, and may be any object such as a computer device or a plant such as a manufacturing factory or a processing facility.

[0015] The information processing device 10 is composed of one or more information processing devices each having a calculation device and a storage device. As shown in FIG. 1, the information processing device 10 includes an acquisition unit 11, a classification unit 12, a selection unit 13, a learning unit 14, and a detection unit 15. The functions of the acquisition unit 11, the classification unit 12, the selection unit 13, the learning unit 14, and the detection unit 15 can be realized by the calculation device executing a program for realizing each function stored in the storage device. The information processing device 10 also includes a measurement value data storage unit 16, a situation data storage unit 17, a training data storage unit 18, and a model storage unit 19. The measurement value data storage unit 16, the situation data storage unit 17, the training data storage unit 18, and the model storage unit 19 are each composed of a storage device. Each configuration will be described in detail below.

[0016] The acquisition unit 11 acquires measurement data measured by various sensors installed in the automobile C and stores it in the measurement data storage unit 16. For example, the measurement data is a value representing the performance of each of a plurality of accessories equipped in the automobile C, and is data measured by various sensors at a predetermined time interval. As an example, the measurement data may be a vehicle speed, acceleration, engine speed, fuel injection amount, battery remaining amount, and tire pressure, as a measurement representing the performance of various devices mounted on the automobile C. However, the measurement data may be a measurement of any accessory as long as it represents the performance of the automobile C. The acquisition unit 11 may acquire measurement data accumulated inside the automobile C together with time data from the automobile C, or may acquire measurement data transmitted from the automobile C together with time data each time measurement is performed at the automobile C. Then, as shown in FIG. 2(1), the acquisition unit 11 associates the acquired measurement data with identification information (ID) for identifying the automobile C and time data, and stores it in the measurement data storage unit 16. In this embodiment, the measurement data acquired by the acquisition unit 11 is data when the operating state of the automobile C is normal, but it may also be data measured when the automobile C is in a specific operating state.

[0017] The acquisition unit 11 also acquires situation data representing the situation of the automobile C when the above-mentioned measurement data is measured, and stores the acquired situation data in the situation data storage unit 17. For example, the situation data is data representing the situation of the automobile C caused by the external situation of the automobile C (hereinafter, the situation data related to the external situation of the automobile is also referred to as "external situation data"). As an example, the external situation data includes weather, temperature, brightness, time of day, and road surface conditions as situations caused by the external environment in which the automobile C is placed. Furthermore, the external situation data includes steering direction and dozing as situations caused by the situation of the driver (operator) who operates the automobile C. Note that these situation data are acquired from various sensors equipped in the automobile C. For example, the weather, brightness, and road surface conditions can be detected from image data acquired from an exterior camera, the steering direction can be detected from data acquired from a steering angle sensor, and the driver's dozing can be detected from image data of the driver acquired from an in-vehicle camera. However, the situation data may be any data as long as it represents the situation of the automobile C. For example, the status data may be data representing the status of the vehicle itself or the status of the vehicle due to the internal status, and may be data such as the vehicle model, model number, purchase date, repair history, and total mileage of the vehicle C (hereinafter, such data representing the status of the vehicle due to the status of the vehicle itself or the internal status is also referred to as "internal status data"). The acquisition unit 11 may acquire status data accumulated inside the vehicle C together with time data, or may acquire status data transmitted together with time data each time the vehicle C is detected. Then, as shown in Figs. 2(2) and (3), the acquisition unit 11 stores the acquired status data in the status data storage unit 17 in association with identification information (ID) for identifying the vehicle C and time data. In this embodiment, the status data acquired by the acquisition unit 11 is data when the operating status of the vehicle C is normal, but may be data acquired when the vehicle C is in a specific operating status.

[0018] In addition, the acquisition unit 11 generates a state detection model as described below, and when detecting the state of the vehicle C using this model, the acquisition unit 11 acquires measurement value data newly measured by the vehicle C from the vehicle C and passes it to the detection unit 15. At this time, the acquisition unit 11 does not acquire situation data, but acquires only the measurement value data and passes it to the detection unit 15.

[0019] The classification unit 12 classifies the measurement value data stored in the measurement value data storage unit 16 based on the situation data stored in the situation data storage unit 17. That is, the classification unit 12 classifies the measurement value data for each situation of the automobile C specified by the situation data. At this time, the classification unit 12 associates measurement value data and situation data whose associated ID and time data match as being acquired at the same time by the same automobile C. Then, the classification unit 12 creates a classification for each content of the situation data, and assigns and classifies the measurement value data associated with the situation data corresponding to each classification so that it belongs to each classification. As an example, it is assumed that the classification unit 12 creates a classification A in which the situation data is "weather: sunny", a classification B in which the situation data is "road surface: bad road", and a classification C in which the situation data is "driver: dozing". In this case, the classification unit 12 classifies the measurement value data associated with each classification A, B, and C, as shown by the black circles in FIG. 3. Although only three categories are created in the above, the classification unit 12 may create categories for every content of the situation data and classify the measurement data by allocating them to the created categories.

[0020] The selection unit 13 selects measurement data to be used as training data from the measurement data classified into categories as described above, and stores the data in the training data storage unit 18. At this time, the selection unit 13 selects measurement data from each category as training data according to the number of measurement data for each category. For example, the selection unit 13 selects measurement data such that the number of data is equal from each category. As an example, a case where the measurement data is classified as shown by the black circles in FIG. 3 will be described. First, in FIG. 3, the measurement data in category A corresponds to the situation data of "weather: fine", and the number of the measurement data is large. On the other hand, the measurement data in category B corresponds to the situation data of "road surface: bad road", and the measurement data in category C corresponds to the situation data of "driver: dozing", and therefore the number of the measurement data is small because the situation indicated by the situation data is unlikely to occur. In such a case, the selection unit 13 selects the same number of measurement data (for example, four each) from each of the categories A, B, and C as training data, as shown by the white circles in FIG. 4. That is, as shown in FIG. 5, the number of measurement data belonging to each of the categories A, B, and C is different, but the number of measurement data selected as training data is the same in each of the categories A, B, and C. For example, the selection unit 13 sets the minimum value of the measurement data in each category as the selection number, and selects measurement data from each category by the number of the set minimum values. The selection unit 13 is not limited to selecting the exact same number of measurement data from each classification, but may select a number of measurement data that is determined to be approximately the same number (approximately equal) within a preset range.

[0021] However, the selection unit 13 is not necessarily limited to selecting measurement value data from each classification approximately equally. For example, the selection unit 13 may receive an input of a ratio of the number of measurement value data to be selected set for each classification, and select measurement value data from each classification according to the ratio. As an example, when a ratio of "Classification A:Classification B:Classification C = 2:1:1" is set, the selection unit 13 performs selection so that the number of measurement value data to be selected from each classification is the set ratio. Note that the ratio may be input and set by an operator or the like, or may be calculated and set by the selection unit 13 according to the distribution and number of measurement value data for each classification. For example, the selection unit 13 may set the ratio so that the wider the distribution of the measurement value data belonging to the classification, the higher the ratio.

[0022] The learning unit 14 performs machine learning based on the training data selected from each classification as described above and stored in the training data storage unit 18, and stores the state detection model generated as a result of the learning in the model storage unit 19. At this time, the learning unit 14 learns all the training data selected from each classification together. That is, the learning unit 14 learns only the measurement value data without including the situation data. In addition, the learning unit 14 may additionally learn the training data from which the measurement value data is selected as described above each time measurement value data is added, and update the state detection model. Here, since the training data to be learned is measurement value data measured when the automobile C is in a normal state, the state detection model generated by learning is configured to detect whether the state is normal or abnormal. However, if the measurement value data measured from the automobile C is measured when the automobile C is in a specific state set in advance, a model for detecting whether the automobile C is in the specific state or not is generated.

[0023] The detection unit 15 detects the state of the automobile C using the state detection model stored in the model storage unit 19. Specifically, the detection unit 15 acquires measurement data newly measured by the automobile C via the acquisition unit 11, and inputs the measurement data to the state detection model. The detection unit 15 then detects the state of the automobile C according to the output result of the state detection model. For example, the detection unit 15 detects whether the automobile C is in a normal state or an abnormal state. When the detection unit 15 detects that the automobile C is in an abnormal state, it performs a notification process, such as notifying the automobile C that an abnormal state has been detected, or notifying a mobile terminal of a driver registered in advance or a business that maintains the automobile C.

[0024] Here, the above-mentioned model storage unit 19 and detection unit 15 may be mounted on the automobile C. In other words, an information processing device equipped with a calculation device and a storage device may be mounted inside the automobile C, the model storage unit 19 storing the state detection model may be formed in the storage device, and the detection unit 15 may be constructed by having the calculation device execute a program. In this way, the automobile C can detect the state of the automobile C by inputting newly measured measurement value data into the state detection model in the detection unit 15.

[0025] [Operation] Next, the operation of the above-mentioned information processing device 10 will be described mainly with reference to the flowcharts of Figures 6 to 7. First, the operation of generating a state detection model will be described with reference to the flowchart of Figure 6.

[0026] The information processing device 10 acquires measurement data and situation data from a plurality of automobiles C (step S1). The measurement data are values ​​that represent the performance of each of a plurality of accessories equipped on the automobile C, such as vehicle speed, acceleration, engine speed, fuel injection amount, remaining battery charge, tire air pressure, etc. The situation data are data that represent the situation of the automobile C when the measurement data is measured, such as weather, temperature, brightness, time of day, road surface condition, steering direction by the driver, whether the driver is dozing, etc.

[0027] Next, the information processing device 10 classifies the measurement data based on the situation data (step S2). At this time, the classification unit 12 of the information processing device 10 distributes and classifies the associated measurement data for each situation of the automobile C specified by the situation data. For example, as shown by the black circles in Fig. 3, the information processing device 10 classifies the measurement data measured in each situation into category A where the situation data is "weather: sunny", category B where the situation data is "road surface: bad road", and category C where the situation data is "driver: dozing".

[0028] Next, the information processing device 10 selects measurement data to be used as training data from each classification (step S3). At this time, the information processing device 10 selects measurement data from each classification approximately evenly. That is, the information processing device 10 selects approximately the same number of measurement data from each classification even if the number of measurement data belonging to each classification is different, as shown in Fig. 3 and Fig. 4. Note that the information processing device 10 is not necessarily limited to selecting approximately the same number of measurement data from each classification, and may select a number of measurement data according to the ratio between the classifications set for each classification.

[0029] Next, the information processing device 10 performs machine learning using the measurement data selected from each classification as training data to generate a state detection model (step S4). This makes it possible to generate a model that evenly considers all conditions of the automobile C, thereby improving the accuracy of the model.

[0030] Next, with reference to the flowchart in FIG. 7, an operation for detecting the state of the automobile C using the state detection model generated as described above will be described.

[0031] The information processing device 10 acquires new measurement data from the vehicle C every time the vehicle C measures new measurement data (step S11). Then, the information processing device 10 inputs the new measurement data into a state detection model (step S12) and detects the state of the vehicle C from the output result of the state detection model (step S13). For example, the information processing device 10 detects whether the vehicle C is in a normal state or an abnormal state. This allows the information processing device 10 to detect the state of the vehicle C with high accuracy.

[0032] <Embodiment 2> Next, a second embodiment of the present invention will be described with reference to Fig. 8 to Fig. 10. Fig. 8 to Fig. 9 are block diagrams showing the configuration of a learning device in the second embodiment, and Fig. 10 is a flowchart showing the operation of the learning device. Note that this embodiment shows an outline of the configuration of the information processing device and information processing method described in the above-mentioned embodiments.

[0033] First, the hardware configuration of the learning device 100 in this embodiment will be described with reference to Fig. 8. The learning device 100 is configured as a general information processing device, and is equipped with the following hardware configuration, as an example. ·CPU(Central Processing Unit)101(Arithmetic unit) ROM (Read Only Memory) 102 (storage device) RAM (Random Access Memory) 103 (storage device) · Programs 104 loaded into RAM 103 A storage device 105 for storing a program group 104 A drive device 106 that reads and writes data from a storage medium 110 external to the information processing device A communication interface 107 for connecting to a communication network 111 outside the information processing device Input / output interface 108 for inputting and outputting data Bus 109 connecting each component

[0034] The learning device 100 can be equipped with the classification unit 121, the selection unit 122, and the learning unit 123 shown in Fig. 9 by the CPU 101 acquiring the program group 104 and executing it. The program group 104 is stored in the storage device 105 or the ROM 102 in advance, for example, and the CPU 101 loads the program group 104 into the RAM 103 and executes it as necessary. The program group 104 may be supplied to the CPU 101 via the communication network 111, or may be stored in the storage medium 110 in advance, and the drive device 106 may read out the program and supply it to the CPU 101. However, the classification unit 121, the selection unit 122, and the learning unit 123 described above may be constructed of dedicated electronic circuits for realizing such means.

[0035] 8 shows an example of the hardware configuration of the information processing device that is the learning device 100, and the hardware configuration of the information processing device is not limited to the above-mentioned case. For example, the information processing device may be configured with a part of the above-mentioned configuration, such as not having the drive device 106.

[0036] The learning device 100 executes the learning method shown in the flowchart of FIG. 10 using the functions of the classification unit 121, the selection unit 122, and the learning unit 123, which are constructed by the program as described above.

[0037] As shown in FIG. 10, the learning device 100 Classifying measurement data obtained by measuring the performance of an object based on situation data that represents the situation of the object when the measurement data was measured (step S101); Selecting the measurement data from each classification according to the number of measurement data for each classification (step S102); Machine learning is performed based on the selected measurement data (step S103). The following process is executed.

[0038] By being configured as described above, the present invention can generate a state detection model that takes into account all situations of an object equally, and can improve the accuracy of the model.

[0039] The above-mentioned program can be stored and supplied to a computer using various types of non-transitory computer readable media. The non-transitory computer readable medium includes various types of tangible storage media. Examples of the non-transitory computer readable medium include magnetic recording media (e.g., flexible disks, magnetic tapes, hard disk drives), magneto-optical recording media (e.g., magneto-optical disks), CD-ROM (Read Only Memory), CD-R, CD-R / W, and semiconductor memory (e.g., mask ROM, PROM (Programmable ROM), EPROM (Erasable PROM), flash ROM, RAM (Random Access Memory)). The program may also be supplied to a computer by various types of transitory computer readable media. Examples of the transitory computer readable medium include electric signals, optical signals, and electromagnetic waves. The transitory computer readable medium can supply the program to a computer via a wired communication path such as an electric wire and an optical fiber, or via a wireless communication path.

[0040] Although the present invention has been described above with reference to the above-mentioned embodiments, the present invention is not limited to the above-mentioned embodiments. Various modifications that can be understood by a person skilled in the art can be made to the configuration and details of the present invention within the scope of the present invention. In addition, at least one or more functions of the above-mentioned classification unit 121, selection unit 122, and learning unit 123 may be executed by an information processing device installed and connected anywhere on a network, that is, they may be executed by so-called cloud computing.

[0041] <Additional Notes> A part or all of the above-described embodiments can be described as follows: The following is an outline of the configurations of the learning method, learning device, and program of the present invention. However, the present invention is not limited to the following configurations. (Appendix 1) classifying measurement data obtained by measuring the performance of an object based on situation data that represents a situation of the object when the measurement data was measured; Selecting the measurement data from each classification according to the number of measurement data for each classification; Performing machine learning based on the selected measurement data. How to learn. (Appendix 2) 2. The learning method according to claim 1, further comprising: classifying the measurement data, which measures the performance of each of the plurality of pieces of equipment equipped on the object, based on the situation data; How to learn. (Appendix 3) 3. The learning method according to claim 1 or 2, classifying the measurement data based on external condition data representing a condition of the object caused by an external condition of the object; How to learn. (Appendix 4) A learning method according to any one of claims 1 to 3, classifying the measurement data based on internal condition data representing a condition of the object caused by an internal condition of the object; How to learn. (Appendix 5) A learning method according to any one of claims 1 to 4, The measurement data is selected approximately equally from each classification. How to learn. (Appendix 6) 6. The learning method according to claim 1, further comprising: selecting the measurement data from each classification according to a ratio set for each classification; How to learn. (Appendix 7) 7. The learning method according to claim 1, further comprising: the object is a car, When classifying the measurement value data based on the situation data, the situation data represents at least one of a condition of a road surface on which the vehicle is traveling and a weather condition at the time of traveling. How to learn. (Appendix 8) A learning method according to any one of claims 1 to 7, the object is a car, When classifying the measurement data based on external condition data representing a condition of the vehicle caused by an external condition of the vehicle, the external condition data is at least one of weather, temperature, brightness, time of day, road surface condition, driver's operation condition, and driver's state when the vehicle is traveling. How to learn. (Appendix 9) A learning method according to any one of Supplementary Notes 1 to 8, the object is a car, When classifying the measurement data based on internal condition data representing the condition of the vehicle caused by the internal condition of the vehicle, the internal condition data is at least one of the vehicle model, model number, purchase date, repair history, and total mileage of the vehicle. How to learn. (Appendix 10) A state detection method using the learning method according to any one of Supplementary Notes 1 to 9, The measurement data newly measured from the object is input to a model generated by performing the machine learning, and a state of the object is detected according to an output from the model. Condition detection method. (Appendix 11) a classification unit that classifies measurement value data obtained by measuring a performance of an object based on situation data that represents a situation of the object when the measurement value data was measured; a selection unit that selects the measurement data from each classification according to the number of the measurement data for each classification; A learning unit that performs machine learning based on the selected measurement data; A learning device equipped with (Appendix 12) 12. The learning device according to claim 11, The classification unit classifies the measurement value data, which measures performance of each of a plurality of pieces of equipment equipped to the object, based on the situation data. Learning device. (Appendix 13) 13. The learning device according to claim 11 or 12, The classification unit classifies the measurement value data based on external condition data representing a condition of the object caused by an external condition of the object. Learning device. (Appendix 14) A learning device according to any one of claims 11 to 13, The classification unit classifies the measurement value data based on internal condition data representing a condition of the object caused by the internal condition of the object. Learning device. (Appendix 15) A learning device according to any one of claims 11 to 14, The selection unit selects the measurement value data from each classification approximately equally. Learning device. (Appendix 16) 16. The learning device according to any one of claims 11 to 15, the selection unit selects the measurement value data from each classification in accordance with a ratio set for each classification; Learning device. (Appendix 17) 17. The learning device according to any one of claims 11 to 16, the object is a car, When classifying the measurement value data based on the situation data, the situation data represents at least one of a condition of a road surface on which the vehicle is traveling and a weather condition at the time of traveling. Learning device. (Appendix 18) 18. The learning device according to any one of claims 11 to 17, the object is a car, When classifying the measurement data based on external condition data representing a condition of the vehicle caused by an external condition of the vehicle, the external condition data is at least one of weather, temperature, brightness, time of day, road surface condition, driver's operation condition, and driver's state when the vehicle is traveling. Learning device. (Appendix 19) 19. A learning device according to any one of claims 11 to 18, the object is a car, When classifying the measurement data based on internal condition data representing the condition of the vehicle caused by the internal condition of the vehicle, the internal condition data is at least one of the vehicle model, model number, purchase date, repair history, and total mileage of the vehicle. Learning device. (Appendix 20) A state detection device including the learning device according to any one of appendixes 11 to 19, A detection unit is provided that inputs the measurement data newly measured from the object to a model generated by performing the machine learning, and detects a state of the object according to an output from the model. Condition detection device. (Appendix 21) In the information processing device, classifying measurement data obtained by measuring the performance of an object based on situation data that represents a situation of the object when the measurement data was measured; Selecting the measurement data from each classification according to the number of measurement data for each classification; Performing machine learning based on the selected measurement data. A computer-readable storage medium that stores a program for executing processing. [Explanation of symbols]

[0042] 10. Information processing device 11 Acquisition Department 12 Classification section 13 Selection Department 14 Learning Department 15 Detection unit 16 Measurement data storage section 17 Situation data storage unit 18 Training data storage unit 19 Model memory section C car 100 Learning Device 101 CPU 102 ROM 103 RAM 104 Programs 105 Storage device 106 Drive device 107 Communication Interface 108 Input / Output Interface 109 Bus 110 Storage medium 111 Communication Network 121 Classification Department 122 Selection Department 123 Learning Department

Claims

1. classifying measurement data that measures the performance of a vehicle based on situation data that represents a situation of the vehicle that is attributable to a situation of a driver of the vehicle at the time the measurement data was measured; When selecting the measurement data from each classification according to the number of the measurement data for each classification, the measurement data is selected from each classification according to a ratio set for the classification; The ratio is set so as to be higher as the distribution of the measurement data becomes wider, Performing machine learning based on the selected measurement data. How to learn.

2. classifying measurement data that measures the performance of a vehicle based on situation data that represents a situation of the vehicle that is attributable to a situation of a driver of the vehicle at the time the measurement data was measured; When selecting the measurement data from each classification according to the number of the measurement data for each classification, the measurement data is selected from each classification according to a different ratio set for each classification; Performing machine learning based on the selected measurement data. How to learn.

3. 3. The learning method according to claim 1 or 2, The measurement data is classified based on the situation data representing at least one of an operation situation of the driver and a state of the driver. How to learn.

4. 3. The learning method according to claim 1 or 2, classifying the measurement data based on the situation data indicative of at least one of a steering direction of the driver and a drowsy state of the driver; How to learn.

5. A learning method according to any one of claims 1 to 4, classifying the measurement data, which measures the performance of each of a plurality of equipment items equipped on the vehicle, based on the situation data; How to learn.

6. A state detection method using the learning method according to any one of claims 1 to 5, The measurement data newly measured from the vehicle is input to a model generated by performing the machine learning, and the state of the vehicle is detected according to an output from the model. Condition detection method.

7. a classification unit that classifies measurement data that measures the performance of a vehicle based on situation data that represents a situation of the vehicle that is caused by a situation of a driver of the vehicle when the measurement data is measured; a selection unit that selects the measurement data from each classification according to a ratio set for each classification when selecting the measurement data from each classification according to the number of the measurement data for each classification; A learning unit that performs machine learning based on the selected measurement data; Equipped with The ratio is set so as to be higher as the distribution of the measurement data becomes wider. Learning device.

8. a classification unit that classifies measurement data that measures the performance of a vehicle based on situation data that represents a situation of the vehicle that is caused by a situation of a driver of the vehicle when the measurement data is measured; a selection unit that selects the measurement data from each classification according to a different ratio set for each classification when selecting the measurement data from each classification according to the number of the measurement data for each classification; A learning unit that performs machine learning based on the selected measurement data; A learning device equipped with

9. The learning device according to claim 7 or 8, The classification unit classifies the measurement data based on the situation data representing at least one of an operation situation of the driver and a state of the driver. Learning device.

10. The learning device according to claim 7 or 8, The classification unit classifies the measurement data based on the situation data representing at least one of a steering direction of the driver and a dozing state of the driver. Learning device.

11. A learning device according to any one of claims 7 to 10, The classification unit classifies the measurement value data, which measures the performance of each of a plurality of accessories equipped on the automobile, based on the situation data. Learning device.

12. A state detection device including the learning device according to any one of claims 7 to 11, A detection unit is provided that inputs the measurement data newly measured from the vehicle to a model generated by performing the machine learning, and detects a state of the vehicle according to an output from the model. Condition detection device.

13. In the information processing device, classifying measurement data that measures the performance of a vehicle based on situation data that represents a situation of the vehicle that is attributable to a situation of a driver of the vehicle at the time the measurement data was measured; When selecting the measurement data from each classification according to the number of the measurement data for each classification, the measurement data is selected from each classification according to a ratio set for the classification; The ratio is set so as to be higher as the distribution of the measurement data becomes wider, Performing machine learning based on the selected measurement data. A program for executing a process.

14. In the information processing device, classifying measurement data that measures the performance of a vehicle based on situation data that represents a situation of the vehicle that is attributable to a situation of a driver of the vehicle at the time the measurement data was measured; When selecting the measurement data from each classification according to the number of the measurement data for each classification, the measurement data is selected from each classification according to a different ratio set for each classification; Performing machine learning based on the selected measurement data. A program for executing a process.

15. 15. The program according to claim 13 or 14, The information processing device includes: The measurement data is classified based on the situation data representing at least one of an operation situation of the driver and a state of the driver. A program for executing a process.

16. 15. The program according to claim 13 or 14, The information processing device includes: classifying the measurement data based on the situation data indicative of at least one of a steering direction of the driver and a drowsy state of the driver; A program for executing a process.

Images