Data analysis device, data analysis system, learning data generation method, and program

Abstract

To promote data collection from an information provider.SOLUTION: In a data analysis device 10, an inquiry unit 106 transmits an inquiry regarding the operating status of a device 3 to an information terminal 3 of an information provider U. A status data receiving unit 107 receives status data from the information terminal 5 as a response to the inquiry transmitted by the inquiry unit 106. A learning data generation unit 108 generates learning data in which driving data indicating detail of the driving is associated with the status data received by the status data receiving unit 107. A reward granting unit 111 gradually grants rewards to the information provider U for generation of the learning data generated by the learning data generation unit 108.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present disclosure relates to a data analysis device, a data analysis system, a method for generating training data, and a program. [Background technology]

[0002] There are known techniques for analyzing operation data of equipment and inferring conditions related to the operation of the equipment. For example, Patent Document 1 discloses a control device that calculates a predicted mean vote (PMV) value from sensor values ​​such as a temperature sensor and a humidity sensor, and controls an air conditioner based on the PMV value. Patent Document 1 also discloses correcting the PMV value based on thermal sensation information such as "hot" or "cold" input by a user, and awarding points to encourage the user to input their thermal sensation. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2016-95066 Summary of the Invention [Problem to be solved by the invention]

[0004] In the technology for inferring situations related to the operation of equipment as described above, in order to improve the accuracy of the inference process, it is necessary to collect situation data that will be the correct answer for the inference process from information providers. Therefore, there is a demand for facilitating the collection of data from information providers.

[0005] The present disclosure has been made to solve the above-mentioned problems, and aims to provide a data analysis device and the like that can promote the collection of data from information providers. [Means for solving the problem]

[0006] In order to achieve the above object, a data analysis device according to the present disclosure includes: Inquiry means for transmitting an inquiry about a status related to the operation of the equipment to an information terminal of the information provider; a situation data receiving means for receiving situation data indicating the situation from the information terminal as a response to the inquiry transmitted by the inquiry means; a learning data generating means for generating learning data in which the driving data indicating the details of the driving is associated with the situation data received by the situation data receiving means; and reward provision means for providing a reward in stages to the information provider for generating the learning data generated by the learning data generation means. [Effects of the Invention]

[0007] In the present disclosure, an inquiry about a situation related to the operation of a device is sent, situation data is received as a response to the inquiry, learning data is generated in which the operation data and the situation data are associated, and rewards for generating the learning data are gradually given to information providers. Thus, according to the present disclosure, it is possible to promote the collection of data from information providers. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram showing the overall configuration of a data analysis system according to a first embodiment. [Figure 2] FIG. 1 is a block diagram showing a hardware configuration of a data analysis device according to a first embodiment. [Figure 3] FIG. 1 is a block diagram showing a functional configuration of a data analysis device according to a first embodiment. [Figure 4] FIG. 1 shows a configuration of a device data storage unit according to the first embodiment. [Figure 5] FIG. 1 shows a configuration of a device attribute table according to the first embodiment. [Figure 6] FIG. 1 shows a configuration of a user information table according to the first embodiment. [Figure 7] FIG. 10 is a diagram showing an example of a registration screen displayed on an information terminal according to the first embodiment; [Figure 8] FIG. 1 is a diagram showing the configuration of an operation history table according to the first embodiment; [Figure 9] 1 shows examples of input data and output data of an inference model according to embodiment 1. [Figure 10] FIG. 1 shows a configuration of a learning definition storage unit according to the first embodiment. [Figure 11] FIG. 1 shows a configuration of a learning input definition table according to the first embodiment. [Figure 12] FIG. 10 is a diagram showing the configuration of a learning output definition table according to the first embodiment; [Figure 13] FIG. 1 shows a configuration of a learning definition table according to the first embodiment. [Figure 14] FIG. 10 is a diagram showing an example of an inquiry screen displayed on an information terminal according to the first embodiment; [Figure 15] FIG. 10 is a diagram showing an example of a free-writing screen displayed on an information terminal according to the first embodiment; [Figure 16] FIG. 1 shows a configuration of a situation data table according to the first embodiment. [Figure 17] FIG. 1 shows a configuration of a learning data storage unit according to the first embodiment. [Figure 18] FIG. 1 shows a configuration of a learning data table according to the first embodiment. [Figure 19] FIG. 1 shows a configuration of a learning state table according to the first embodiment. [Figure 20] FIG. 1 shows a configuration of an evaluation information table according to the first embodiment. [Figure 21] FIG. 1 shows the configuration of a remuneration management table according to the first embodiment. [Figure 22] FIG. 1 is a first diagram showing an example of a notification screen according to the first embodiment; [Figure 23] FIG. 2 is a second diagram showing an example of a notification screen according to the first embodiment; [Figure 24] 1 is a flowchart showing a flow of a device installation process according to the first embodiment. [Figure 25] 1 is a flowchart showing a flow of a learning data definition process according to the first embodiment. [Figure 26] 1 is a first flowchart showing a flow of a data analysis process according to the first embodiment; [Figure 27] 2 is a second flowchart showing the flow of the data analysis process according to the first embodiment; [Figure 28] 1 is a flowchart showing the flow of evaluation processing according to the first embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments will be described in detail with reference to the drawings, in which the same or corresponding parts are designated by the same reference numerals.

[0010] (Embodiment 1) 1, a data analysis system 1 according to the first embodiment includes a plurality of devices 3, a plurality of information terminals 5, and a data analysis device 10. The data analysis system 1 is a system that collects and analyzes operation data of the devices 3 and infers a situation related to the operation data. More specifically, the data analysis system 1 is a system that generates learning data from data obtained when the devices 3 are operating, for inferring the situation of the devices 3 or their surroundings when the devices 3 are operating.

[0011] The multiple devices 3, multiple information terminals 5, and data analysis device 10 are communicably connected to each other via a wide area network N. The wide area network N is, for example, the Internet. Each device 3 and each information terminal 5 is distinguishable from one another and is managed by the data analysis device 10.

[0012] Each of the multiple devices 3 is a home appliance such as an air conditioner, cooker, light, refrigerator, etc. Each device 3 is installed in one of multiple residences H, which are the living spaces of people from which data is collected. Only one device 3 may be installed in one residence H, or two or more devices 3 may be installed. Each device 3 has a built-in or external communication device, and transmits operating data of each device 3 to the data analysis device 10 via the wide area network N.

[0013] Each of the multiple information terminals 5 is a communication terminal such as a PC (Personal Computer), a smartphone, etc. Each information terminal 5 is operated by an information provider U. Here, the information provider U is a person who provides information about the device 3 to the data analysis device 10. In many cases, the information provider U is a user of the device 3, and more specifically, a resident of the residence H in which the device 3 is installed, the owner of the device 3, etc.

[0014] Although not shown, each information terminal 5 includes a CPU (Central Processing Unit), ROM (Read Only Memory), RAM (Random Access Memory), a communication interface, and a readable / writable non-volatile semiconductor memory. Each information terminal 5 communicates with the data analysis device 10 via the wide area network N. For example, each information terminal 5 receives messages from the data analysis device 10 and transmits situation data input by an information provider U to the data analysis device 10.

[0015] The data analysis device 10 is an information processing device such as a PC or a cloud server, and is installed under the management of the operator of the data analysis system 1. The data analysis device 10 is a device that analyzes data related to multiple devices 3. As shown in FIG. 2 , the data analysis device 10 includes a control unit 11, a storage unit 12, an operation unit 13, a display unit 14, and a communication unit 15.

[0016] The control unit 11 includes a CPU, a ROM, and a RAM. The CPU is also called a central processing unit, a processor, a microprocessor, a microcomputer, etc., and functions as a central processing unit that executes processing and calculations related to the control of the data analysis device 10. In the control unit 11, the CPU reads out programs and data stored in the ROM and uses the RAM as a work area to perform overall control of the data analysis device 10.

[0017] The storage unit 12 includes a nonvolatile semiconductor memory such as a flash memory, an EPROM (Erasable Programmable ROM), or an EEPROM (Electrically Erasable Programmable ROM), and serves as a so-called secondary storage device or auxiliary storage device. The storage unit 12 stores programs and data used by the control unit 11 to perform various processes. The storage unit 12 also stores data generated or acquired by the control unit 11 as a result of performing various processes.

[0018] The operation unit 13 includes input devices such as a touch panel, a touch pad, a keyboard, and a mouse, and receives operation inputs from a user. The information provider U can input various instructions to the data analysis device 10 by operating the operation unit 13.

[0019] The display unit 14 includes a display device such as an LCD (Liquid Crystal Display) panel, an organic EL (Electro-Luminescence) panel, etc. The display unit 14 is driven by a display drive circuit (not shown), and displays various images under the control of the control unit 11.

[0020] In addition, the operation unit 13 and the display unit 14 are not limited to being built into the data analysis device 10, but may be located outside the data analysis device 10 and exchange information with the data analysis device 10 via the communication unit 15.

[0021] The communication unit 15 includes a communication interface for the data analysis device 10 to communicate with an external device. The communication unit 15 communicates with the external device via the wide area network N. Specifically, the communication unit 15 communicates with a plurality of devices 3 and a plurality of information terminals 5 via the wide area network N.

[0022] Next, the functional configuration of the data analysis device 10 will be described with reference to FIG. 3. The data analysis device 10 includes, in the control unit 11, a registration data receiving unit 101 as an example of a registration data receiving means, a driving data receiving unit 102 as an example of a driving data receiving means, an inference unit 103 as an example of an inference means, an analyst input unit 104 as an example of an analyst input means, a conversion unit 105 as an example of a conversion means, an inquiry unit 106 as an example of an inquiry means, a situation data receiving unit 107 as an example of a situation data receiving means, a learning data generating unit 108 as an example of a learning data generating means, a learning unit 109 as an example of a learning means, an evaluation unit 110 as an example of an evaluation means, a reward granting unit 111 as an example of a reward granting means, and a notification unit 112 as an example of a notification means. Each of these functions is realized in the control unit 11 by software, firmware, or a combination of software and firmware. The software and firmware are written as programs and stored in the ROM or the memory unit 12. In the control unit 11, the CPU executes the programs stored in the ROM or the storage unit 12, thereby realizing the functions shown in FIG.

[0023] The data analysis device 10 also includes an equipment data storage unit 121, a learning definition storage unit 122, a situation data storage unit 123, a learning data storage unit 124, an evaluation information storage unit 125, and a reward storage unit 126. These are constructed in appropriate storage areas in the storage unit 12.

[0024] As shown in FIG. 4, the device data storage unit 121 stores a device attribute table T1, a user information table T2, and an operation history table T3.

[0025] The device attribute table T1 is a table that stores information about the attributes of the devices 3. The device attribute table T1 includes multiple columns as shown in FIG. 5. In the device attribute table T1, "device ID" is an ID that identifies an individual device 3, and represents identification information for uniquely identifying each of the multiple devices 3 under the management of the data analysis system 1. "Device type" represents the type of each device 3, such as an air conditioner, cooker, light, refrigerator, etc., in other words, the model of each device 3. "User ID" represents the ID of the information provider U. "Data usage permission" represents whether or not the information provider U is permitted to use the operating data as learning data.

[0026] The user information table T2 is a table that stores information about the information provider U. The user information table T2 includes multiple columns as shown in FIG. 6. In the user information table T2, the "user ID" is the ID of the information provider U, and is the same as the "user ID" in the device attribute table T1. The "email address" indicates the email address of the information provider U.

[0027] Returning to FIG. 3 , when a new device 3 is installed under the management of the data analysis system 1, the registration data receiving unit 101 receives the registration data of the device 3 from the information terminal 5 corresponding to the device 3. As an example, when an information provider U installs a new device 3 in his / her residence H, the information provider U reads a QR (Quick Response) code (registered trademark) affixed to the device 3 with the information terminal 5. The QR code may also be affixed to a document enclosed with the device 3.

[0028] The QR code includes URL (Uniform Resource Locator) information of the website of the data analysis device 10 and attribute data such as the model and product serial number of the device 3. When the QR code is read by the information terminal 5, the information terminal 5 transmits the attribute data of the device 3 contained in the QR code to the data analysis device 10. In the data analysis device 10, the registration data receiving unit 101 receives the attribute data transmitted from the information terminal 5 and thereby detects that a new device 3 has been installed.

[0029] Furthermore, when the information terminal 5 reads the QR code, it accesses the website of the data analysis device 10 indicated by the QR code. When the information terminal 5 accesses the website of the data analysis device 10, the registration screen shown in Fig. 7 is displayed on the display unit of the information terminal 5. In the "Product Serial Number" field on the registration screen, the product serial information included in the QR code is displayed by default.

[0030] The information provider U enters his / her own email address in the "Registered Email Address" field on the registration screen. The entered email address will be the recipient of points, as described below. The information provider U also selects "Data Use Permission" on the registration screen. As for "Data Use Permission," the information provider U can select whether or not to allow the information provided to be used as learning data. The information terminal 5 transmits the registration data entered by the information provider U on the registration screen to the data analysis device 10 via the wide area network N.

[0031] When the registration data receiving unit 101 receives the attribute data and registration data from the information terminal 5, it updates the device attribute table T1 and the user information table T2. Specifically, the registration data receiving unit 101 assigns a new user ID to the email address included in the received registration information, associates the assigned user ID with the email address, and stores it in the user information table T2.

[0032] Furthermore, the registration data receiving unit 101 assigns a new device ID to the product serial number included in the received attribute data. Then, the registration data receiving unit 101 associates the assigned device ID with the information on the model of the device 3 included in the received attribute data, the user ID added to the user information table T2, and the data use permission information included in the received registration data, and stores them in the "Device Type," "User ID," and "Data Use Permission" columns of the device attribute table T1, respectively.

[0033] 3, during normal operation of the data analysis apparatus 10, the operating data receiving unit 102 receives operating data from each of the plurality of devices 3 present in the data analysis system 1. The operating data is data indicating the details of the operation performed by the device 3, such as the operating state, operating mode, etc. of the device 3.

[0034] More specifically, if the device 3 is a cooker, the operation data indicates the current operating state, operation mode, etc. of the cooker. Here, the operating state of the cooker indicates whether the cooker is ON or OFF. The same applies to the operating states of other devices 3. Furthermore, the operating mode of the cooker indicates cooking modes such as high heat, medium heat, low heat, and simmering heat. If the device 3 is an air conditioner, the operation data indicates the current operating state, operation mode, set temperature, etc. of the air conditioner. If the device 3 is a light, the operation data indicates the current operating state, intensity, etc. of the light. If the device 3 is a refrigerator, the operation data indicates whether the refrigerator door is currently open or closed.

[0035] Each device 3 periodically transmits history data indicating its own operating history to the data analysis device 10 at predetermined time intervals, such as one minute or five minutes. The operating data receiving unit 102 receives the operating data periodically transmitted from each device 3 in this manner, and stores the received operating data in an operating history table T3 in the device data storage unit 121.

[0036] The operation history table T3 is a table that stores operation data of multiple devices 3. The operation history table T3 includes multiple columns as shown in FIG. 8. In the operation history table T3, the "date and time" indicates the date and time when operation data was received from one of the devices 3. The "device ID" is an ID for uniquely identifying the device 3 that sent the operation data, and is the same as the "device ID" in the device attribute table T1. The "operation data" indicates the operating state, operation mode, refrigerator door, etc. of the device 3. Different data is stored in the "operation data" depending on the model of the device 3.

[0037] The operating data receiving unit 102 adds the received operating data to the operating history table T3 every time it receives operating data from one of the plurality of devices 3. As a result, the operating data of each device 3 is stored in chronological order in the operating history table T3.

[0038] 3, the inference unit 103 executes an inference process to infer, from the operation data of the device 3, a situation related to the operation of the device 3. Here, "a situation related to the operation of the device 3" means a situation of the device 3 or its surroundings when the device 3 is being operated. "A situation related to the operation of the device 3" is, for example, the environment in which the device 3 is used when the device 3 is being operated, the behavior of the user of the device 3 when the device 3 is being operated, etc.

[0039] In order to execute the inference process, the inference unit 103 uses an inference model M that has been generated in advance by machine learning. The inference model M is generated by the learning unit 109, which will be described later, executing machine learning, and is stored in the storage unit 12.

[0040] As shown in FIG. 9, the inference model M receives operation data as input data and outputs situation data indicating a situation related to the operation data as output data. For example, if the device 3 is a cooker, the inference model M receives input of the cooking time, heating intensity, and time period of use of the cooker, and outputs the type of food. If the device 3 is a light, the inference model M receives input of the frequency with which the light is turned on / off, the length of time the light is on, and the time period during which the light is on, and outputs the type of room in which the light is installed. If the device 3 is a refrigerator, the inference model M receives input of the time period and length of time the refrigerator door is open, and outputs the time when the user went out shopping. The inference unit 103 infers the situation indicated by the situation data output from the inference model M in response to such input operation data as the situation of the device 3 or its surroundings when the device 3 is operating.

[0041] The input data and output data in such an inference model M are defined in the learning definition storage unit 122. The learning definition storage unit 122 stores information necessary for the inference unit 103 to define a learning process in which the inference unit 103 infers situation data from the operating data of the device 3. As shown in Figure 10, the learning definition storage unit 122 stores a learning input definition table T4, a learning output definition table T5, and a learning definition table T6.

[0042] The learning input definition table T4 is a table that defines the input data in the learning process. The learning input definition table T4 includes multiple columns as shown in Figure 11. The "input definition ID" is an ID that identifies the input data of the inference model M and is represented by an ID beginning with "LI_".

[0043] "Attribute name" represents an attribute of the operation data that can be input data for the inference model M. "Type" represents the type of the operation data. In the example of Figure 11, the learning input definition table T4 defines the operation data of the cooker, "cooker heating time," "cooker heating intensity," and "cooker usage time period," as int type, string type, and string type, respectively.

[0044] "Conversion definition" represents a program that converts the operation data stored in the operation history table T3 into data with attribute names defined in the learning input definition table T4. As a specific example, "Program 1" defines that the "cooker heating time" is to be calculated from the difference between the time when the operation status value of the cooker in the operation history table T3 switches from ON to OFF and the time when it switches from OFF to ON. "Program 2" defines that the value of the operation mode of the cooker in the operation history table T3 is to be applied as is as the "cooker heating intensity." "Program 3" defines that the "cooker usage time period" is to be calculated as follows, based on the time when the operation status of the cooker in the operation history table T3 switches from OFF to ON:

[0045] 6:00 to 9:00 is the "breakfast time." 11:00 to 13:00 is the "lunch time." Between 5:00 PM and 9:00 PM is considered "dinner time." If it is anything other than the above, select "Other."

[0046] Similarly, for "frequency of lighting ON / OFF," "length of time lighting is ON," "time period lighting is ON," etc., a program for calculating these from lighting operation data in the operation history table T3 is defined in the "conversion definition."

[0047] The learning output definition table T5 is a table that defines the output data in the learning process. The learning output definition table T5 includes multiple columns as shown in Figure 12. The "output definition ID" is an ID that identifies the output data of the inference model M and is represented by an ID beginning with "LO_".

[0048] "Attribute name" represents an attribute of situation data that can become output data for the inference model M. "Type" represents the type of situation data for a situation. "Attribute value definition" defines the value that the situation data of the attribute name can take in the learning output definition table T5. In the example of Figure 12, the learning output definition table T5 defines the attribute values ​​"boiled food," "grilled food," "stir-fried food," and "steamed food" as string types for "food type." The learning output definition table T5 also defines the attribute values ​​"living room," "bedroom," "kitchen," "study," "bathroom," and "toilet" as string types for "room type."

[0049] The learning definition table T6 is a table that defines the correspondence between input data and output data in the learning process. The learning definition table T6 includes multiple columns as shown in Figure 13. The "learning definition ID" is an ID assigned in association with the learning process and is represented by an ID beginning with "L_".

[0050] "Input" represents the input data of the inference model M as a set of "input definition IDs" defined in the learning input definition table T4. "Output" represents the output data of the inference model M as a set of "output definition IDs" defined in the learning output definition table T5. In the example of Figure 13, the learning definition table T6 defines learning to infer "type of food" from "cooker heating time," "cooker heating intensity," and "time period when the cooker is in use." The learning definition table T6 also defines learning to infer "room type" from "frequency of lights on / off," "length of time lights are on," and "time period when lights are on."

[0051] "Answer points" represent points awarded to the information provider U by the reward awarding unit 111 (described later) when the information provider U answers an inquiry. "Answer points" are set according to the value of the situation data provided by the information provider U. Here, the value of the situation data is determined by the market size of the corresponding device 3, the difficulty of collecting the situation data, and the like. Specifically, situation data used in inference processing related to major products, products with large market sizes, etc., is highly valuable. Furthermore, situation data that is likely to lead to the identification of individuals, situation data that is difficult to collect because it is sensitive data, and the like are highly valuable. For example, while an estimate of a type of cuisine is unlikely to lead to the identification of an individual, the time a person went shopping is likely to lead to the identification of an individual. Therefore, the time a person went shopping is more valuable as situation data than an estimate of a type of cuisine. "Answer points" may be automatically set according to the type of device 3, or may be arbitrarily set by an analyst.

[0052] 3 , the analyst input unit 104 accepts information about data analysis input by an analyst. The analyst is a person who analyzes data in the data analysis system 1. The analyst can operate the operation unit 13 to input the definition of the learning data in the learning definition storage unit 122.

[0053] For example, when an analyst wishes to register new input data in the learning input definition table T4, the analyst inputs the "attribute name," "type," and "conversion definition" of the input data to be registered. Alternatively, when an analyst wishes to register new output data in the learning output definition table T5, the analyst inputs the "attribute name," "type," and "attribute value definition" of the output data to be registered. Furthermore, when an analyst wishes to define new learning data in the learning definition table T6, the analyst inputs the correspondence between the input data and output data in the learning data to be defined, as well as the "answer points." The analyst input unit 104 accepts the information input by the analyst in this way and updates the learning definition storage unit 122.

[0054] 3, the conversion unit 105 converts the driving data stored in the driving history table T3 into data in the format of input data for the learning process. Specifically, when the driving data receiving unit 102 receives driving data from any of the multiple devices 3, the conversion unit 105 converts the received driving data into data in the format of input data for the inference model M in accordance with the program specified in the "conversion definition" column of the learning input definition table T4.

[0055] For example, when the operation data receiving unit 102 receives operation data from a cooker, the conversion unit 105 calculates the "heating time of the cooker," the "heating intensity of the cooker," and the "time period during which the cooker is in use" according to Programs 1 to 3 defined in the learning input definition table T4. Alternatively, when the operation data receiving unit 102 receives operation data from a light, the conversion unit 105 calculates the "frequency of turning the light on / off," the "length of time the light is on," and the "time period during which the light is on" according to Programs 4 to 6 defined in the learning input definition table T4.

[0056] After converting the format of the operation data in this manner, the conversion unit 105 determines whether all of the input data defined in the "Input" column of any learning definition ID in the learning definition table T6 has been obtained. For example, when the conversion unit 105 calculates the "heating time of the cooker," the "heating intensity of the cooker," and the "time period during which the cooker is used" from the operation data of a cooker, it determines that all of the input data for the learning definition ID "L_0001" has been obtained. In this case, the conversion unit 105 determines that the operation data of the cooker is a target for inference processing.

[0057] When the conversion unit 105 determines that the driving data is a target for inference processing, the inference unit 103 executes inference processing to infer, from the driving data, a situation related to the device 3 during the driving. Specifically, the inference unit 103 inputs the driving data converted by the conversion unit 105 into the inference model M. Then, the inference unit 103 infers that the output data output from the inference model M is a situation related to the device 3 during the driving.

[0058] The inquiry unit 106 transmits an inquiry about the situation related to the driving of the device 3 to the information terminal 5. Specifically, when the driving data receiving unit 102 receives driving data from any of the devices 3 and the conversion unit 105 determines that the driving data corresponds to the target of inference processing, the inquiry unit 106 generates a message inquiring about the situation of the device 3 or its surroundings when the driving data was being driven. Then, the inquiry unit 106 transmits the generated message to the information terminal 5 of the user of the device 3 that is the source of the driving data. In this way, the inquiry unit 106 requests correct answer data for the inference processing executed by the inference unit 103.

[0059] More specifically, the inquiry unit 106 refers to the device attribute table T1 to identify the user ID linked to the device ID of the device 3 that transmitted the driving data, and further refers to the user information table T2 to identify the email address linked to the identified user ID. Furthermore, when the conversion unit 105 determines that the driving data is a target for inference processing, the inquiry unit 106 refers to the "output" and "answer point" columns in the learning definition table T6 that are linked to the input data converted from the driving data by the conversion unit 105. In this way, the inquiry unit 106 identifies the output definition ID and the answer point.

[0060] Next, the inquiry unit 106 refers to the learning output definition table T5 and identifies the attribute of the situation data defined in the "attribute name" column associated with the identified output definition ID. For example, if the operation data receiving unit 102 receives operation data from a cooker, the inquiry unit 106 identifies the attribute of the situation data as the food type. Alternatively, if the operation data receiving unit 102 receives operation data from a light, the inquiry unit 106 identifies the attribute of the situation data as the room type.

[0061] When the attributes of the situation data are identified from the learning output definition table T5, the inquiry unit 106 generates a message inquiring about the situation data of the identified attributes. Then, the inquiry unit 106 sends the generated message to the email address identified from the user information table T2. As a result, the inquiry unit 106 displays the message inquiring about the situation on the information terminal 5 corresponding to the device 3 that sent the driving data.

[0062] As an example, when the operation data receiving unit 102 receives operation data from a cooker, the inquiry unit 106 displays the inquiry screen shown in Fig. 14 on the information terminal 5. Specifically, the inquiry unit 106 displays, on the inquiry screen, a message inquiring about the type of food cooked by the cooker when it was operating, such as "What food were you cooking?", along with multiple options. The information provider U can respond to the inquiry by selecting the correct type of food from the multiple options.

[0063] Furthermore, the query unit 106 further displays a message indicating the inference result inferred by the inference unit 103, such as "It is estimated to be baked goods." The query unit 106 also displays a message indicating that, if an answer is obtained, answer points identified from the learning definition table T6 will be awarded, such as "If you can help us by answering, we will provide you with 10 points." In this way, the query unit 106 generates an inquiry including information on the situation inferred by the inference unit 103 and information on points that will be obtained if the information provider U answers the inquiry, and transmits the inquiry to the information terminal 5.

[0064] Furthermore, when the information provider U answers the inquiry on the inquiry screen shown in FIG. 14 and selects the "Next" icon, the information terminal 5 switches the inquiry screen to a free-text screen shown in FIG. 15. On the free-text screen, the information provider U can freely write an answer to the inquiry, such as "I was grilling salmon for breakfast," or "Curry." In such a description, the information provider U can optionally enter supplementary information to the answer. Such information can prompt the analyst to consider the matter, so points are added to the information provider U who fills in the remarks column.

[0065] The inquiry unit 106 generates and transmits such a message when the "state" in the learning state table T9 described later is "continue." Specifically, in the example of the learning state table T9 shown in FIG. 19, when the learning definition ID corresponding to the driving data determined to be the target of the inference process is "L_0001," the inquiry unit 106 transmits an inquiry message because the "state" is "continue." In contrast, when the learning definition ID corresponding to the driving data determined to be the target of the inference process is "L_0002," the inquiry unit 106 does not transmit an inquiry message because the "state" is "end."

[0066] Such message templates and message generation rules are prepared in advance and stored in the storage unit 12. The query unit 106 generates a query by embedding information such as the date and time, the attributes of the situation data to be queried, the model of the device 3 that has transmitted the driving data, the inference result by the inference unit 103, the number of points, etc. In this way, the query unit 106 dynamically generates a query based on the attributes of the situation data, the type of device 3, the points for the response, the inference result, etc.

[0067] When the information terminal 5 receives input of a response to the inquiry from the information provider U, it transmits situation data, which is data indicating the input response, to the data analysis device 10 as a response to the inquiry transmitted by the inquiry unit 106.

[0068] 3, the situation data receiving unit 107 receives situation data transmitted from the information terminal 5. When situation data is transmitted from one of the multiple information terminals 5, the situation data receiving unit 107 receives the transmitted situation data and stores the received situation data in the situation data storage unit 123.

[0069] The situation data storage unit 123 stores a situation data table T7. The situation data table T7 is a table that stores a reception history of situation data that the information provider U has provided in response to questions from the data analysis device 10. The situation data table T7 includes multiple columns shown in FIG.

[0070] In the situation data table T7, "inquiry sending time" indicates the time when the inquiry unit 106 sent an inquiry to the information provider U prompting the information provider U to input situation data. "Response reception time" indicates the time when the situation data receiving unit 107 received situation data indicating an answer to the inquiry from the information provider U. "Device ID" indicates the ID of the device 3 corresponding to the information terminal 5 that sent the received situation data. "Learning definition ID" is an ID that identifies the learning process corresponding to the received situation data, and is represented by the learning definition ID in the learning definition table T6.

[0071] "Operating data" refers to operating data that is input for the learning process. When the situation data receiving unit 107 receives situation data, it stores the operating data corresponding to the received situation data in the corresponding input definition ID field of the "operating data" column of the situation data table T7. For example, when the operating data receiving unit 102 receives operating data from a cooker, the situation data receiving unit 107 stores the operating data converted by the conversion unit 105 in the "cooker heating time," "cooker heating intensity," and "cooker usage time period" fields of the "operating data" in the situation data table T7. Alternatively, when the operating data receiving unit 102 receives operating data from a light, the situation data receiving unit 107 stores the operating data converted by the conversion unit 105 in the "light ON / OFF frequency," "light ON time length," and "light ON time period" fields of the "operating data" in the situation data table T7.

[0072] "Context data" represents the context data that is the output of the learning process. When the context data receiving unit 107 receives context data, it stores the answer indicated in the received context data in the corresponding output definition ID field in the "Context Data" column of the context data table T7. For example, when context data is received indicating that the "Cooking Type" of a cooker is "Grilled Food," the context data receiving unit 107 stores "Grilled Food" in the "Cooking Type" field of the "Context Data" in the context data table T7. Alternatively, when context data is received indicating that the "Room Type" of a light is "Toilet," the context data receiving unit 107 stores "Toilet" in the "Room Type" field of the "Cooking Data" in the context data table T7.

[0073] If the situation data received from the information provider U includes a description entered on the free description screen shown in Figure 15, the situation data receiving unit 107 stores the description in the ``Notes (free description)'' column of the situation data table T7.

[0074] The information stored in the "driving data" column of the situation data table T7 is included in the inquiry sent by the inquiry unit 106 using hidden tags in HTML (HyperText Markup Language). The driving data specified in the hidden tags is not displayed on the information terminal 5, but is included in the response sent from the information terminal 5 and is received together with the situation data by the situation data receiving unit 107. This allows the situation data receiving unit 107 to store data in the situation data table T7 only from the response received from the information terminal 5.

[0075] 3, the learning data generation unit 108 generates learning data in which driving data indicating the details of driving of the device 3 is associated with situation data indicating situations related to the driving, which is received by the situation data receiving unit 107. The learning data is teacher data for machine learning to generate the inference model M. The learning data includes multiple data sets, each of which is a data set combining driving data with corresponding situation data.

[0076] The learning data generation unit 108 extracts data to be input and output of the learning process defined in the learning definition storage unit 122 from the situation data storage unit 123. Specifically, the learning data generation unit 108 extracts the "answer reception time," "learning definition ID," "driving data," and "situation data" stored in the situation data table T7. Then, the learning data generation unit 108 stores the extracted data in the learning data storage unit 124 as learning data.

[0077] The learning data storage unit 124 stores the learning data defined in the learning definition storage unit 122. As shown in FIG. 17, the learning data storage unit 124 stores a learning data table T8 and a learning state table T9.

[0078] The learning data table T8 is a table in which data related to learning is extracted from the situation data table T7. The learning data table T8 includes multiple columns as shown in Figure 18. In the learning data table T8, the "date and time," "learning definition ID," "driving data," and "situation data" are the same as the "answer reception time," "learning definition ID," "driving data," and "situation data" in the situation data table T7, respectively.

[0079] The learning status table T9 is a table for determining whether or not to collect further learning data for each learning definition ID. The learning status table T9 includes multiple columns as shown in Figure 19. The "learning definition ID" is an ID assigned in association with the learning process and is the same as the "learning definition ID" in the learning definition table T6. The "number of data" indicates the number of data sets stored in the learning data table T8 for the learning data of each "learning definition ID." The "status" indicates whether collection of learning data for each "learning definition ID" is continuing, or whether sufficient learning data has been collected and collection has ended.

[0080] Each time new situation data is stored in the situation data table T7, the learning data generation unit 108 extracts the data of these columns from the new situation data from the situation data table T7 and stores the data in the corresponding columns of the learning data table T8. At that time, the learning data generation unit 108 updates the number of data items in the learning status table T9. As a result, learning data of various learning definition IDs is accumulated in the learning data table T8.

[0081] Returning to FIG. 3, the learning unit 109 performs machine learning using the learning data generated by the learning data generation unit 108. Specifically, the learning unit 109 uses a known learning algorithm such as a neural network or a support vector machine to perform supervised learning using multiple data sets of driving data and situation data stored in the learning data table T8 as training data. In this way, the learning unit 109 learns the correspondence between the driving data and the situation data. Through this learning process, the learning unit 109 generates an inference model M that outputs situation data corresponding to input driving data, as shown in FIG. 9.

[0082] The learning unit 109 executes this learning process for each learning definition ID, thereby generating an inference model M that outputs data of the corresponding output definition ID in response to input of data of the input definition ID defined for each learning definition ID in the learning definition table T6.

[0083] Note that when a certain number of data sets of training data have been accumulated, the learning unit 109 performs machine learning using the training data. On the other hand, when the number of data sets of training data is small, effective machine learning cannot be performed, and therefore the learning unit 109 does not perform machine learning when the number of data sets is small. Specifically, the learning unit 109 refers to the "number of data sets" in the learning status table T9 for each learning definition ID. Then, when the number of data sets for any learning definition ID is equal to or greater than a first threshold, the learning unit 109 performs machine learning using the training data of that learning definition ID and generates an inference model M that can correspond to the data of that learning definition ID. The first threshold is set in advance to the minimum number of data sets required to perform effective machine learning.

[0084] When the driving data receiving unit 102 receives new driving data, the inference unit 103 infers a situation corresponding to the received new driving data using the latest inference model M generated by the learning unit 109 in this manner. Then, the query unit 106 generates a query including information on the situation corresponding to the new driving data inferred by the inference unit 103, and transmits the query to the information terminal 5. Note that if the number of data corresponding to the received new driving data is insufficient, the inference model M has not yet been generated, and therefore the inference unit 103 does not execute the inference process. In this case, the query transmitted by the query unit 106 to the information terminal 5 does not include information on the situation inferred by the inference unit 103.

[0085] Returning to FIG. 3, the evaluation unit 110 evaluates the value of the training data generated by the training data generation unit 108. The value of the training data refers to the extent to which the training data was useful for generating a highly accurate inference model M. Specifically, the evaluation unit 110 evaluates (1) the accuracy rate and (2) the results as the value of the training data.

[0086] (1) Correct answer rate The accuracy rate of the learning data is the accuracy rate of the inference model M generated from the learning data, and means the probability that the inference model M generated from the learning data will output situation data that is correct for input driving data. The evaluation unit 110 refers to the "number of data" in the learning status table T9 and determines whether the number of data for any learning definition ID has reached a second threshold or more. The second threshold is a predetermined threshold for determining whether a sufficient number of data items has been accumulated in the learning data. When the number of data items for any learning definition ID has reached the second threshold or more, the evaluation unit 110 updates the "status" of the learning definition ID for which the number of data items has reached the second threshold or more in the learning status table T9 from "continue" to "end."

[0087] When the state of any learning definition ID is completed, the evaluation unit 110 evaluates the accuracy rate of the learning data of that learning definition ID. The evaluation unit 110 calculates the accuracy rate by applying a known method. As an example, the evaluation unit 110 performs K-fold tolerance verification. Specifically, the evaluation unit 110 sets one of multiple datasets included in the learning data to be evaluated as test data and sets the remaining datasets as training data. Then, the evaluation unit 110 changes the dataset set as test data in multiple ways and calculates the accuracy rate of the inference process by the inference model M generated from the training data.

[0088] (2) Results The results of the training data refer to the results of the service realized using the training data. For example, if the processing for inferring food types is useful for a recipe suggestion service, the greater the contribution of the training data to business profits, the higher the evaluation unit 110 will evaluate the results of the training data.

[0089] When a result is obtained from any of the learning data stored in the learning data storage unit 124, the analyst operates the operation unit 13 to input evaluation information for the learning data. The evaluation unit 110 evaluates the result of the learning data based on the evaluation information input by the analyst.

[0090] Specifically, the evaluation unit 110 calculates the result value and result contribution rate of the learning data as values ​​indicating the result of the learning data. Here, the result value means a value expressed in monetary terms of the value of the result. The result contribution rate means the proportion of the result contributed by the learning data. Information on the result value and result contribution rate is included in the evaluation information input by the analyst. The evaluation unit 110 calculates the result value and result contribution rate of the learning data based on the evaluation information input by the analyst.

[0091] When the evaluation unit 110 calculates the accuracy rate or the achievement value as the value of the learning data, it stores the calculated accuracy rate, or the achievement value and achievement contribution rate in the evaluation information storage unit 125. The evaluation information storage unit 125 stores an evaluation information table T10. The evaluation information table T10 is a table that manages the value of the learning data corresponding to the learning definition ID. The evaluation information table T10 includes multiple columns shown in FIG. 20.

[0092] In the evaluation information table T10, "date and time" represents the update date of the evaluation information. "learning definition ID" represents an ID that identifies the learning process. "outcome ID" represents an ID that identifies the outcome. "outcome value" and "outcome contribution rate" represent the outcome value and outcome contribution rate corresponding to the outcome ID. "correct answer rate" represents the correct answer rate of the learning data corresponding to the learning definition ID.

[0093] After calculating the accuracy rate or the achievement value and achievement contribution rate of the learning data, the evaluation unit 110 assigns a new achievement ID. Then, the evaluation unit 110 associates the achievement ID and the calculated accuracy rate or the achievement value and achievement contribution rate with the learning definition ID of the learning data and stores them in the corresponding columns in the evaluation information table T10.

[0094] Returning to FIG. 3 , the reward granting unit 111 grants rewards in stages to the information provider U for generating the learning data generated by the learning data generation unit 108. Here, the reward has economic, honorary, or other value to the information provider U and is used to increase the information provider U's motivation when responding to inquiries from the inquiry unit 106. As an example, the reward granting unit 111 grants points to the information provider U as reward. Points are units of value that can be exchanged for goods or services, such as money. The reward granting unit 111 grants such points in stages to the information provider U who provided information to generate the learning data, each time a predetermined reason for granting is met.

[0095] More specifically, the reward granting unit 111 grants a reward to the information provider U according to the information provider U's degree of contribution to the generation of the training data generated by the training data generation unit 108. Here, the information provider U's degree of contribution to the generation of the training data means the degree of contribution of the information provider U to the generation of valuable training data. For example, if the information provider U provides valuable situation data as a response to an inquiry sent from the inquiry unit 106, the information provider U's degree of contribution will be high. Furthermore, if a highly accurate inference model M is generated from the training data, the information provider U's degree of contribution, who provided the situation data used to generate the training data, will be high.

[0096] The reward granting unit 111 calculates points to be granted to the information provider U according to the information provider U's degree of contribution. More specifically, the reward granting unit 111 uses five indices, namely, (A) permission to use, (B) answer, (C) additional description, (D) accuracy rate, and (E) result, as indices indicating the information provider U's degree of contribution. The reward granting unit 111 then calculates points to be granted to the information provider U in stages according to these five indices. Each indices will be explained below.

[0097] (A) Permission to use The permission for use indicates whether or not the information provider U is permitted to use the driving data for generating learning data. When the information provider U permits the use of the driving data for generating learning data, the reward granting unit 111 grants points to the information provider U. The permission for use is determined based on whether or not the information provider U has selected permission to use the driving data as learning data in the "Permission for Use of Data" section of the registration screen shown in FIG. 7. Specifically, the reward granting unit 111 grants a predetermined number of points to the information provider U who has selected "Permission for Use as Learning Data."

[0098] (B)Answer The response is a response to the inquiry sent by the inquiry unit 106. When the situation data receiving unit 107 receives situation data from the information terminal 5 as a response to the inquiry, the reward granting unit 111 grants points to the information provider U of the information terminal 5.

[0099] Specifically, the reward granting unit 111 refers to the "answer points" in the learning definition table T6, and grants the answer points stored in association with the situation data received by the situation data receiving unit 107 to the information provider U. As described above, the "answer points" in the learning definition table T6 stores points whose value corresponds to the value of the situation data, i.e., the market size of the corresponding device 3, the difficulty of collecting the situation data, etc. Therefore, the reward granting unit 111 grants points to the information provider U according to the value of the situation data received by the situation data receiving unit 107.

[0100] (C) Additional Description The additional description is a description that can be entered on the free description screen shown in Fig. 15. When the information provider U enters an additional description on the free description screen, the reward granting unit 111 grants additional points in addition to the points for the answer entered on the inquiry screen shown in Fig. 14. In other words, when the situation data received by the situation data receiving unit 107 includes an additional description, the reward granting unit 111 grants bonus points.

[0101] (D) Correct answer rate As described above, the accuracy rate is one of the values ​​of the learning data evaluated by the evaluation unit 110. When the evaluation unit 110 calculates the accuracy rate of the learning data, the reward granting unit 111 grants points to the information provider U according to the calculated accuracy rate. As an example, the relationship between the accuracy rate and the points to be granted is set in advance in a conversion table. The reward granting unit 111 converts the accuracy rate calculated by the evaluation unit 110 into points based on the conversion table. Then, the reward granting unit 111 grants the converted points to all of the information providers U who provided information for the corresponding learning data.

[0102] Here, the information provider U who provided information for the training data is the information provider U who provided the situation data for generating the training data. One training data is generated from situation data transmitted from the information terminals 5 of various information providers U. Therefore, the reward granting unit 111 grants points according to the calculated accuracy rate to all of the information providers U who provided situation data for generating the training data whose value has been evaluated by the evaluation unit 110.

[0103] (E) Results As described above, the result is one of the values ​​of the learning data evaluated by the evaluation unit 110, and is represented by the result value and the result contribution rate. When the evaluation unit 110 calculates the result value and the result contribution rate of the learning data, the reward granting unit 111 grants points to the information provider U according to the calculated result value and the result contribution rate. As an example, the relationship between the result value, the result contribution rate, and the points to be granted is set in advance in a conversion table. The reward granting unit 111 converts the result value and the result contribution rate calculated by the evaluation unit 110 into points based on the conversion table. Then, the reward granting unit 111 grants the converted points to all of the information providers U who provided information for the corresponding learning data.

[0104] After calculating the points to be awarded to the information provider U, the reward awarding unit 111 updates the reward storage unit 126 based on the calculated points. The reward storage unit 126 stores a reward management table T11. The reward management table T11 is a table that manages the points for each information provider U. The reward management table T11 includes multiple columns shown in FIG. 21.

[0105] In the reward management table T11, "date and time" represents the date and time when points were awarded or consumed. "User ID" represents the ID that identifies the information provider U. "Awarded points" represents the points awarded on the corresponding date and time. "Consumed points" represents the points consumed on the corresponding date and time. "Remaining points" represents the points held by the information provider U, updated by the generation of awarded points or consumed points.

[0106] The "reason for awarding" indicates the reason why points were awarded. When the reward awarding unit 111 awards points to the information provider U, it stores one of the five indicators described above, (A) permission to use, (B) answer, (C) additional description, (D) accuracy rate, and (E) result, in the "reason for awarding" column in the reward management table T11 according to the reason.

[0107] Returning to Fig. 3, the notification unit 112 notifies the information provider U of the reward granted by the reward granting unit 111. The notification unit 112 generates a message notifying the information provider U of the points granted by the reward granting unit 111, and transmits the generated message to the information terminal 5 of the information provider U to whom the points have been granted.

[0108] First, when the reward granting unit 111 grants points according to the accuracy rate of the learning data, the notification unit 112 transmits a message to the information terminals 5 of all the information providers U to whom points have been granted, notifying them that points have been granted based on the accuracy rate of the learning data. As a result, the notification unit 112 causes the information terminals 5 of the information providers U to display a notification screen, for example, as shown in FIG. 22 .

[0109] Second, when the reward granting unit 111 grants points according to the results of the learning data, the notification unit 112 transmits a message to the information terminals 5 of all the information providers U to whom the points have been granted, notifying them that the points have been granted based on the results of the learning data. As a result, the notification unit 112 causes the information terminals 5 of the information providers U to display a notification screen, for example, as shown in FIG. 23 .

[0110] This notification screen allows the information provider U to confirm that the situation data he or she provided has been effectively utilized, thereby further increasing the motivation of the information provider U to provide situation data.

[0111] Next, the flow of processing executed by the data analysis device 10 will be described with reference to Figures 24 to 28. The processing shown in Figures 24 to 28 is an example of a data analysis method. First, the device installation processing shown in Figure 24 is executed by the control unit 11 when a new device 3 is installed under the management of the data analysis system 1.

[0112] When the device installation process starts, the control unit 11 detects the installation of the device 3 (step S11). Specifically, the control unit 11 detects that a new device 3 has been installed when it receives attribute data of the device 3 contained in a QR code attached to the newly installed device 3.

[0113] When detecting the installation of the device 3, the control unit 11 functions as the registration data receiving unit 101 and receives the registration data of the information provider U for the installed device 3 (step S12). Specifically, the control unit 11 receives the registration data input by the information provider U on the registration screen shown in FIG. 7 from the information terminal 5.

[0114] Upon receiving the registration data, the control unit 11 updates the device attribute table T1 and the user information table T2 based on the attribute data of the device 3 received in step S11 and the registration data received in step S12 (step S13). As a result, the control unit 11 stores information about the newly installed device 3 in the device attribute table T1, and stores information about the user of the device 3 in the user information table T2.

[0115] Next, the control unit 11 functions as a reward granting unit 111 and grants points for permission to use the driving data (step S14). Specifically, if the information provider U has given permission to use the driving data for generating learning data in the registration data received in step S12, the control unit 11 grants points to the information provider U and registers the granted points in the reward management table T11.

[0116] After the points are awarded, the control unit 11 starts the data analysis process shown in Figures 26 and 27 (step S15). With the above, the device installation process shown in Figure 24 ends.

[0117] 25 is executed by the control unit 11 when an analyst registers a new definition of learning data. In the learning data definition process, the control unit 11 functions as the analyst input unit 104.

[0118] When the learning data definition process starts, first, the control unit 11 receives input from the analyst into the learning input definition table T4 (step S21). Then, the control unit 11 updates the learning input definition table T4 in accordance with the received input (step S22). Specifically, the analyst operates the operation unit 13 to input the "attribute name," "type," and "conversion definition" of the input data that the analyst wishes to newly register. The control unit 11 assigns a new input definition ID and registers the information input by the analyst into the learning input definition table T4.

[0119] Second, the control unit 11 receives input for the learning output definition table T5 from the analyst (step S23). Then, the control unit 11 updates the learning output definition table T5 in accordance with the received input (step S24). Specifically, the analyst operates the operation unit 13 to input the "attribute name," "type," and "attribute value definition" of the output data that the analyst wishes to newly register. The control unit 11 assigns a new output definition ID and registers the information input by the analyst in the learning output definition table T5.

[0120] Third, the control unit 11 receives input of the learning definition table T6 from the analyst (step S25). Then, the control unit 11 updates the learning definition table T6 according to the received input (step S26). Specifically, the analyst operates the operation unit 13 to input the correspondence between input data and output data for each piece of learning data that the analyst wishes to newly register, as well as the "answer points." The control unit 11 assigns a new learning definition ID and registers the information input by the analyst in the learning definition table T6. This completes the learning data definition process shown in FIG. 25.

[0121] It is also possible to use data previously registered in the learning input definition table T4 or the learning output definition table T5 as input data or output data for the learning data to be newly registered. When using data already registered in the learning input definition table T4, steps S21 and S22 are omitted. When using data already registered in the learning output definition table T5, steps S23 and S24 are omitted.

[0122] The data analysis processing shown in FIGS. 26 and 27 is executed by the control unit 11 when operation data is transmitted from any of the plurality of devices 3 present in the data analysis system 1 during normal operation of the data analysis apparatus 10.

[0123] When the data analysis process starts, the control unit 11 functions as the operating data receiving unit 102 and receives operating data transmitted from any of the plurality of devices 3 (step S301). Upon receiving the operating data, the control unit 11 stores the received operating data in the operating history table T3 of the device data storage unit 121 and updates the operating history table T3 (step S302).

[0124] After updating the driving history table T3, the control unit 11 functions as the conversion unit 105 and converts the received driving data into input data for the inference model M according to the program specified in the "conversion definition" column of the learning input definition table T4 (step S303).

[0125] Then, the control unit 11 determines whether all input data defined in the "Input" column of any learning definition ID in the learning definition table T6 has been obtained from the received driving data. As a result, the control unit 11 determines whether the received driving data is a target for inference processing (step S304).

[0126] If the received driving data does not correspond to the target of the inference process (step S304; NO), the control unit 11 skips the processes from step S305 onwards and ends the data analysis process.

[0127] On the other hand, if the received driving data corresponds to the target of the inference process (step S304; YES), the control unit 11 functions as the inference unit 103 and executes the inference process (step S305). Specifically, the control unit 11 inputs input data converted from the received driving data into the inference model M. Then, the control unit 11 infers that the situation indicated in the output data output from the inference model M is a situation related to the device 3 during driving.

[0128] When the inference process is executed, the control unit 11 functions as the inquiry unit 106 and sends an inquiry about the situation to the information terminal 5 of the user of the device 3 that is the source of the received driving data (step S306). As a result, the control unit 11 causes the information terminal 5 to display the inquiry screens shown in FIGS. 14 and 15. More specifically, the control unit 11 refers to the learning status table T9, and sends the inquiry if the "status" linked to the learning definition ID for which the input data was obtained in step S303 in the learning status table T9 is "continue." On the other hand, if the "status" is "end," the control unit 11 does not send the inquiry and ends the data analysis process.

[0129] When the inquiry is sent, the control unit 11 functions as the situation data receiving unit 107 and determines whether or not a response to the sent inquiry has been received (step S307). If no response has been received within a predetermined time since the inquiry was sent (step S307; NO), the control unit 11 skips the processing from step S308 onwards and ends the data analysis processing.

[0130] On the other hand, if a reply has been received (step S307; YES), the control unit 11 stores the situation data indicated in the received reply in the situation data table T7, and updates the situation data table T7 (step S308).

[0131] After updating the situation data table T7, the control unit 11 functions as the learning data generation unit 108 and generates learning data (step S309). Specifically, the control unit 11 extracts learning data from the situation data table T7 and stores it in a learning data table T8. The control unit 11 also updates the number of data items in the learning status table T9.

[0132] After generating the learning data, the control unit 11 functions as the learning unit 109 and executes a learning process using the learning data stored in the learning data table T8 (step S310). As a result, the control unit 11 generates or updates an inference model M for inferring situation data from driving data.

[0133] 27, when the learning process is executed, the control unit 11 functions as the reward granting unit 111 and grants points for the received answer (step S311). Specifically, the control unit 11 grants points for the received answer to the information provider U. Furthermore, if the received answer includes an additional description, the control unit 11 grants additional points to the information provider U. The control unit 11 registers the granted points in the reward management table T11.

[0134] After awarding the points, the control unit 11 refers to the "number of data" in the learning state table T9 and determines whether the number of data in any of the learning data has accumulated to a threshold or more (step S312). If the number of data in all of the learning data is less than the threshold (step S312; NO), the control unit 11 skips the processes from step S313 onward and ends the data analysis process.

[0135] On the other hand, if the number of data items accumulated in any of the learning data items exceeds the threshold value (step S312; YES), the control unit 11 ends collection of that learning data item (step S313). Specifically, the control unit 11 updates the "status" of the learning data item whose number of data items exceeds the threshold value in the learning status table T9 from "continue" to "end."

[0136] When the collection of the learning data is completed, the control unit 11 functions as the evaluation unit 110, calculates the accuracy rate of the learning data, and updates the evaluation information table T10 (step S314).

[0137] After calculating the accuracy rate of the learning data, the control unit 11 functions as a reward granting unit 111 and grants points corresponding to the calculated accuracy rate to all information providers U who provided information for the learning data (step S315). The control unit 11 registers the granted points in the reward management table T11.

[0138] When the points are awarded, the control unit 11 functions as the notification unit 112 and notifies the information providers 5 of the awarded points (step S316). Specifically, the control unit 11 sends a message to the information terminals 5 of all the information providers U to whom the points have been awarded, and causes the notification screen shown in Fig. 22 to be displayed, for example. This completes the data analysis process shown in Figs. 26 and 27.

[0139] The evaluation process shown in FIG. 28 is executed by the control unit 11 when the points related to the results of any of the learning data stored in the learning data storage unit 124 are updated.

[0140] When the evaluation process starts, the control unit 11 functions as the evaluation unit 110 and receives input of evaluation information regarding the business contribution of the learning data from the analyst (step S41). Upon receiving the input of the evaluation information, the control unit 11 calculates the performance value and performance contribution rate of the learning data based on the input evaluation information, and updates the evaluation information table T10 (step S42).

[0141] After calculating the result value and result contribution rate of the learning data, the control unit 11 functions as a reward granting unit 111 and grants points for the result to all information providers U who provided information for the learning data based on the calculated result value and result contribution rate (step S43). The control unit 11 registers the awarded points in the reward management table T11.

[0142] When the points are awarded, the control unit 11 functions as the notification unit 112 and notifies the information providers U of the awarded points (step S44). Specifically, the control unit 11 sends a message to the information terminals 5 of all the information providers U to whom the points have been awarded, and causes the notification screen shown in Fig. 23 to be displayed, for example. This completes the evaluation process shown in Fig. 28.

[0143] As described above, the data analysis device 10 according to the first embodiment transmits an inquiry about a situation related to the operation of the device 3 to the information terminal 5, receives situation data from the information terminal 5 as a response to the inquiry, and generates learning data in which the operation data and the situation data are associated with each other. The data analysis device 10 then gradually awards points to the information provider U for generating the learning data. In this way, the data analysis device 10 according to the first embodiment gradually awards points for generating the learning data, thereby increasing the motivation of the information provider U to provide situation data corresponding to the correct answer to the inference process and promoting the collection of situation data. This makes it easier to generate valuable learning data and improves the accuracy of the inference model M. As a result, it becomes possible to infer the corresponding situation from the operation data of the device 3 with high accuracy, which can lead to improved comfort for the user of the device 3 and behavioral support.

[0144] (Variation) Although the embodiments have been described above, it is possible to combine the embodiments, or to modify or omit the embodiments as appropriate.

[0145] For example, in the above embodiment, the reward granting unit 111 grants points as a reward to the information provider U. However, the reward granted by the reward granting unit 111 is not limited to points, and may be, for example, a gift certificate, a service voucher, a coupon, or the like, as long as it can increase the information provider U's motivation to provide information. Furthermore, the reward may be something that confers honorary value on the information provider U, such as promoting the information provider U to a higher rank member when a certain number of points are accumulated.

[0146] In the above embodiment, when the information provider U installs a new device 3 in his / her residence H, he / she reads the QR code attached to the device 3 with the information terminal 5 and accesses the website of the data analysis device 10. However, the information provider U may access the website of the data analysis device 10 by a method other than the QR code and transmit the attribute data of the newly installed device 3 to the data analysis device 10.

[0147] In the above embodiment, the data analysis device 10 includes the units shown in FIG. 3. However, these units are not limited to being included in a single device, and may be separated into different devices independent of each other in the data analysis system 1. For example, the data analysis device 10 may not include the function of the learning unit 109, and a device external to the data analysis device 10 may include the function of the learning unit 109. In this case, the learning data generated by the learning data generation unit 108 is transmitted to the external device, and machine learning is performed in the external device to generate an inference model M. Then, the inference unit 103 receives the inference model M generated by the external device and uses the received inference model M to infer a situation related to the driving data.

[0148] Similarly, the data analysis device 10 may not have any of the functions of the inference unit 103, the learning data generation unit 108, the evaluation unit 110, the reward granting unit 111, etc., and an external device may have these functions. Note that if the data analysis device 10 does not have the function of the inference unit 103, the query sent by the query unit 106 may not include the inference result by the inference unit 103.

[0149] In the above embodiment, the CPU in the control unit 11 of the data analysis device 10 executes a program stored in the ROM or the storage unit 12, thereby functioning as each unit shown in FIG. 3. However, the control unit 11 may be dedicated hardware. Dedicated hardware is, for example, a single circuit, a composite circuit, a programmed processor, an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or a combination of these. When the control unit 11 is dedicated hardware, the functions of each unit may be realized by individual hardware, or the functions of each unit may be realized together by a single piece of hardware.

[0150] In addition, some of the functions of each unit may be realized by dedicated hardware, and other parts may be realized by software or firmware. In this way, the control unit 11 can realize each of the above-mentioned functions by hardware, software, firmware, or a combination of these.

[0151] It is also possible to make an existing computer such as a personal computer or an information terminal device function as data analysis device 10 by applying a program that defines the operation of data analysis device 10 to the computer.

[0152] Furthermore, the method of distribution of such a program is arbitrary, and for example, it may be stored on a computer-readable recording medium such as a CD-ROM (Compact Disk ROM), a DVD (Digital Versatile Disk), an MO (Magneto Optical Disk), or a memory card and distributed, or it may be distributed via a communication network such as the Internet.

[0153] The present disclosure allows various embodiments and modifications without departing from the broad spirit and scope of the present disclosure. Furthermore, the above-described embodiments are intended to illustrate the present disclosure and do not limit the scope of the present disclosure. That is, the scope of the present disclosure is defined by the claims, not the embodiments. Various modifications made within the scope of the claims and the meaning of equivalent disclosures are considered to be within the scope of the present disclosure.

[0154] Various aspects of the present disclosure are summarized below as appendices.

[0155] (Appendix 1) Inquiry means for transmitting an inquiry about a status related to the operation of the equipment to an information terminal of the information provider; a situation data receiving means for receiving situation data indicating the situation from the information terminal as a response to the inquiry transmitted by the inquiry means; a learning data generating means for generating learning data in which the driving data indicating the details of the driving is associated with the situation data received by the situation data receiving means; and a reward granting means for granting a reward in stages to the information provider for generating the learning data generated by the learning data generating means. Data analysis equipment. (Appendix 2) the reward granting means grants the reward to the information provider in accordance with the degree of contribution of the information provider to generation of the training data generated by the training data generation means. 2. The data analysis apparatus of claim 1. (Appendix 3) the reward granting means grants the reward to the information provider in accordance with the value of the situation data received by the situation data receiving means. 3. The data analysis apparatus according to claim 1 or 2. (Appendix 4) the reward granting means grants the reward to the information provider when the information provider permits the driving data to be used in generating the learning data. 4. A data analysis apparatus according to any one of appendices 1 to 3. (Appendix 5) further comprising evaluation means for evaluating the value of the training data generated by the training data generation means; the reward granting means grants the reward to the information provider in accordance with the value of the learning data evaluated by the evaluation means. 5. A data analysis apparatus according to any one of appendices 1 to 4. (Appendix 6) the evaluation means evaluates the accuracy rate of an inference model generated from the training data as the value of the training data. 6. The data analysis apparatus of claim 5. (Appendix 7) the evaluation means evaluates the results of the service realized by the learning data as the value of the learning data. 7. The data analysis apparatus according to claim 5 or 6. (Appendix 8) the inquiry means transmits the inquiry to the information terminal when the driving data received from the device corresponds to a target of an inference process for inferring a situation related to the driving. 8. The data analysis apparatus of any one of appendices 1 to 7. (Appendix 9) the inquiry means transmits the inquiry, which includes information on a reward that the information provider will receive if the information provider answers the inquiry, to the information terminal; 9. A data analysis apparatus according to any one of appendices 1 to 8. (Appendix 10) further comprising an inference means for inferring the situation from the driving data, the query means transmits the query including the information on the situation inferred by the inference means to the information terminal; 10. The data analysis apparatus of any one of appendices 1 to 9. (Appendix 11) a learning means for generating an inference model for inferring the situation from the driving data by machine learning using the learning data generated by the learning data generating means; the inference means, when new driving data is received from the device, infers a situation corresponding to the new driving data using the inference model generated by the learning means; The inquiry means transmits the inquiry to the information terminal, the inquiry including information on the situation inferred by the inference means and corresponding to the new driving data. 11. The data analysis apparatus of claim 10. (Appendix 12) Inquiry means for transmitting an inquiry about a status related to the operation of the equipment to an information terminal of the information provider; a situation data receiving means for receiving situation data indicating the situation from the information terminal as a response to the inquiry transmitted by the inquiry means; a learning data generating means for generating learning data in which the driving data indicating the details of the driving is associated with the situation data received by the situation data receiving means; and a reward granting means for granting a reward in stages to the information provider for generating the learning data generated by the learning data generating means. Data analysis system. (Appendix 13) Sending an inquiry about the status related to the operation of the equipment to the information terminal of the information provider; receiving status data indicating the status from the information terminal as a response to the transmitted inquiry; generating learning data in which the driving data indicating the details of the driving is associated with the received situation data; A reward for generating the generated training data is given to the information provider in stages. How to generate training data. (Appendix 14) Computer, an inquiry means for transmitting an inquiry about a status related to the operation of the equipment to an information terminal of the information provider; a situation data receiving means for receiving situation data indicating the situation from the information terminal as a response to the inquiry transmitted by the inquiry means; a learning data generating means for generating learning data in which the driving data indicating the details of the driving is associated with the situation data received by the situation data receiving means; a reward providing means for providing a reward for the generation of the learning data generated by the learning data generating means to the information provider in stages; Program for. [Explanation of symbols]

[0156] 1 Data analysis system, 3 Equipment, 5 Information terminal, 10 Data analysis device, 11 Control unit, 12 Memory unit, 13 Operation unit, 14 Display unit, 15 Communication unit, 101 Registration data receiving unit, 102 Driving data receiving unit, 103 Inference unit, 104 Analyst input unit, 105 Conversion unit, 106 Inquiry unit, 107 Situation data receiving unit, 108 Learning data generation unit, 109 Learning unit, 110 Evaluation unit, 111 Reward assignment unit, 112 Notification unit, 121 Equipment data storage unit, 122 Learning definition storage unit, 123 Situation data storage unit, 124 Learning data storage unit, 125 Evaluation information storage unit, 126 Reward storage unit, H Residence, M Inference model, N Wide area network, T1 Equipment attribute table, T2 User information table, T3 Driving history table, T4 Learning input definition table, T5 Learning output definition table, T6 learning definition table, T7 situation data table, T8 learning data table, T9 learning status table, T10 evaluation information table, T11 reward management table, U information provider

Claims

1. Inquiry means for transmitting an inquiry about a status related to the operation of the equipment to an information terminal of the information provider; a situation data receiving means for receiving situation data indicating the situation from the information terminal as a response to the inquiry transmitted by the inquiry means; a learning data generating means for generating learning data in which the driving data indicating the details of the driving is associated with the situation data received by the situation data receiving means; and a reward granting means for granting a reward in stages to the information provider for generating the learning data generated by the learning data generating means. Data analysis equipment.

2. the reward granting means grants the reward to the information provider in accordance with the degree of contribution of the information provider to generation of the training data generated by the training data generation means. The data analysis device according to claim 1 .

3. the reward granting means grants the reward to the information provider in accordance with the value of the situation data received by the situation data receiving means. The data analysis device according to claim 1 .

4. the reward granting means grants the reward to the information provider when the information provider permits the driving data to be used in generating the learning data. The data analysis device according to claim 1 .

5. further comprising evaluation means for evaluating the value of the training data generated by the training data generation means; the reward granting means grants the reward to the information provider in accordance with the value of the learning data evaluated by the evaluation means. The data analysis device according to claim 1 .

6. the evaluation means evaluates the accuracy rate of an inference model generated from the training data as the value of the training data. The data analysis device according to claim 5 .

7. the evaluation means evaluates the results of the service realized by the learning data as the value of the learning data. The data analysis device according to claim 5 .

8. the inquiry means transmits the inquiry to the information terminal when the driving data received from the device corresponds to a target of an inference process for inferring a situation related to the driving. The data analysis device according to claim 1 .

9. the inquiry means transmits the inquiry, which includes information on a reward that the information provider will receive if the information provider answers the inquiry, to the information terminal; The data analysis device according to claim 1 .

10. further comprising an inference means for inferring the situation from the driving data, the query means transmits the query including the information on the situation inferred by the inference means to the information terminal; The data analysis device according to claim 1 .

11. a learning means for generating an inference model for inferring the situation from the driving data by machine learning using the learning data generated by the learning data generating means; the inference means, when new driving data is received from the device, infers a situation corresponding to the new driving data using the inference model generated by the learning means; The inquiry means transmits the inquiry to the information terminal, the inquiry including information on the situation inferred by the inference means and corresponding to the new driving data. The data analysis device according to claim 10.

12. Inquiry means for transmitting an inquiry about a status related to the operation of the equipment to an information terminal of the information provider; a situation data receiving means for receiving situation data indicating the situation from the information terminal as a response to the inquiry transmitted by the inquiry means; a learning data generating means for generating learning data in which the driving data indicating the details of the driving is associated with the situation data received by the situation data receiving means; and a reward granting means for granting a reward in stages to the information provider for generating the learning data generated by the learning data generating means. Data analysis system.

13. Sending an inquiry about the status related to the operation of the equipment to the information terminal of the information provider; receiving status data indicating the status from the information terminal as a response to the transmitted inquiry; generating learning data in which the driving data indicating the details of the driving is associated with the received situation data; A reward for generating the generated training data is given to the information provider in stages. How to generate training data.

14. Computer, an inquiry means for transmitting an inquiry about a status related to the operation of the equipment to an information terminal of the information provider; a situation data receiving means for receiving situation data indicating the situation from the information terminal as a response to the inquiry transmitted by the inquiry means; a learning data generating means for generating learning data in which the driving data indicating the details of the driving is associated with the situation data received by the situation data receiving means; a reward providing means for providing a reward for the generation of the learning data generated by the learning data generating means to the information provider in stages; Program for.

Images