Machine learning support device and machine learning support method

Abstract

To easily collect learning data used for machine learning, and improve accuracy of an AI model.SOLUTION: A machine learning support device comprises: a game control unit which sets a game state in a game map in which virtual sewerage facilities are arranged; an acquisition unit which acquires equipment information including an operation setting which a user determines with respect to the game state and is set for equipment comprised in the sewerage facilities; an inflow data generation unit which generates inflow data indicating an amount and quality of sewage flowing into the sewerage facilities, on the basis of information on at least one of weather, population, and a degree of industrial development in the game map; a simulation processing unit which simulates water treatment in the sewerage facilities, on the basis of the equipment information and the inflow data; and a generation unit which associates the equipment information, the inflow data, and the simulation result with each other and generates learning data used for machine learning.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present disclosure relates to a machine learning assistance device and a machine learning assistance method. [Background technology]

[0002] In recent years, the aging and declining number of operators at sewage treatment plants and other sewerage facilities has posed a problem of how to pass on the operating skills of machinery and equipment. As one solution to this problem, operational assistance technologies using AI (Artificial Intelligence) are being considered.

[0003] In this type of operation assistance technology, for example, AI is trained on accumulated past data, including environmental conditions such as weather and inflow water volume, operational settings for devices and equipment under those environmental conditions, and treated water information such as the quality and volume of treated water treated by the devices and equipment.The trained AI model then presents appropriate operational settings and predicted treated water information to the operator in response to the input environmental conditions. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2022-39126 [Patent Document 2] Japanese Patent Application Publication No. 2023-131390 Summary of the Invention [Problem to be solved by the invention]

[0005] However, there is a problem with AI-based driving assistance technologies: it is not easy to train the AI ​​with sufficient data. Specifically, the data that AI learns to generate an AI model is environmental conditions and operational settings that actually occurred in the past, so data from abnormal situations, such as when extreme weather occurs, is not learned, or even if it is learned, the amount of data is small. As a result, there is a certain limit to the accuracy of the trained AI model, making it difficult to present appropriate operational settings and accurate treated water information when abnormal weather occurs.

[0006] In addition, the equipment and devices that make up sewerage facilities and their appropriate operating settings are subject to change in response to the demands of the times and technological advances, and at the same time, the required treated water information may also change. In this regard, since the data that AI learns from is past operating settings and treated water information, there is a risk that the trained AI model may not reflect operating settings and treated water information based on the latest technology and knowledge.

[0007] Furthermore, the facilities and equipment that make up a sewerage facility and their appropriate operating settings vary depending on the conditions of the area in which the facility is located (population, weather, industry, etc.), so simply learning the environmental conditions, operating settings, and treated water information in one area is unlikely to produce an AI model that will function in other areas.

[0008] As such, the data used to train AI tends to be data obtained under limited conditions, making it difficult to train AI models with sufficient data to obtain highly accurate models. For this reason, there is a need to easily collect diverse training data suited to various situations.

[0009] The present disclosure has been made in consideration of the above, and aims to provide a machine learning assistance device and a machine learning assistance method that can easily collect learning data used in machine learning and improve the accuracy of an AI model. [Means for solving the problem]

[0010] According to one aspect of the present disclosure, a machine learning assistance device includes: a game control unit that sets a game situation in a game map in which virtual sewerage facilities are arranged; an acquisition unit that acquires equipment information including operation settings determined by a user for the game situation set by the game control unit and operation settings of equipment provided in the sewerage facility; an inflow data generation unit that generates inflow data indicating the amount and quality of sewage flowing into the sewerage facility based on at least one of information on weather, population, and industrial development in the game map; a simulation processing unit that simulates water treatment in the sewerage facility based on the equipment information acquired by the acquisition unit and the inflow data generated by the inflow data generation unit; and a generation unit that generates learning data to be used in machine learning by associating the equipment information, the inflow data, and simulation results by the simulation processing unit.

[0011] According to another aspect of the present disclosure, a machine learning assistance method includes: setting a game situation in a game map in which virtual sewerage facilities are arranged; acquiring equipment information including operation settings determined by a user for the set game situation and operation settings of equipment provided in the sewerage facility; generating inflow data indicating the amount and quality of sewage flowing into the sewerage facility based on at least one of information on weather, population, and industrial development in the game map; simulating water treatment in the sewerage facility based on the acquired equipment information and the generated inflow data; and generating learning data to be used in machine learning by associating the equipment information, the inflow data, and the simulation results of the simulation. [Brief explanation of the drawings]

[0012] [Figure 1] FIG. 1 is a diagram illustrating an example of the configuration of a machine learning system according to an embodiment. [Figure 2] FIG. 2 is a block diagram illustrating a configuration of a machine learning device according to an embodiment. [Figure 3]FIG. 3 is a block diagram illustrating a configuration of a machine learning support device according to an embodiment. [Figure 4] FIG. 4 is a flow diagram illustrating a machine learning assistance method according to an embodiment. [Figure 5] FIG. 5 is a diagram showing a specific example of learning data. [Figure 6] FIG. 6 is a flow diagram showing the operation of the machine learning device. [Figure 7] FIG. 7 is a block diagram showing a specific example of the hardware configuration of the machine learning assistance device. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, an embodiment according to the present disclosure will be described with reference to the accompanying drawings. The embodiment described below is an example and should not be construed as being limited by this description.

[0014] Fig. 1 is a diagram showing an example of the configuration of a machine learning system according to an embodiment. In the machine learning system shown in Fig. 1, a machine learning device 100, a machine learning assistance device 200, and multiple user terminal devices 300-1 to 300-N (N is an integer equal to or greater than 2) are communicably connected via a network.

[0015] The machine learning device 100 acquires learning data including environmental conditions and operational settings related to the sewerage facility, and generates an AI model through machine learning using the learning data. The machine learning device 100 then determines and outputs optimal operational settings for the devices and equipment of the sewerage facility through inference using the trained AI model. The machine learning device 100 acquires learning data used for machine learning from the machine learning support device 200. The machine learning device 100 may also acquire actual operational settings and treated water information as learning data from devices and equipment installed in the sewerage facility (not shown).

[0016] The machine learning support device 200 is a game server that provides users of user terminal devices 300-1 to 300-N with a game that includes a simulation of water treatment in a sewerage facility. That is, the machine learning support device 200 manages a virtual sewerage facility for each user, simulates water treatment in each sewerage facility according to the user's settings, and provides the user with a game that uses the simulation results. The machine learning support device 200 then generates learning data from the user's settings during the simulation and the results of the water treatment simulation, and provides the learning data to the machine learning device 100. The specific configuration and operation of the machine learning support device 200 will be described in detail later.

[0017] The user terminal devices 300-1 to 300-N are terminal devices such as smartphones, tablet devices, or personal computers owned by users who play the game. The user terminal devices 300-1 to 300-N output game data provided by the machine learning assistance device 200 from output units such as a display and a speaker, and accept input from users playing the game from input units such as a touch panel and a keyboard. The user terminal devices 300-1 to 300-N then transmit the user input to the machine learning assistance device 200. The user input may include, for example, information specifying the layout of devices and equipment in a virtual sewerage facility in the game, or information specifying the operation settings of the devices and equipment. Hereinafter, when there is no need to particularly distinguish between the user terminal devices 300-1 to 300-N, they will also be simply referred to as "user terminal devices 300."

[0018] Fig. 2 is a block diagram showing the configuration of a machine learning device 100 according to one embodiment. The machine learning device 100 shown in Fig. 2 includes a communication interface unit (hereinafter abbreviated as "communication I / F unit") 110, a learning data acquisition unit 120, a machine learning control unit 130, an AI model storage unit 140, and an operation setting optimization unit 150.

[0019] The communication I / F unit 110 is a wired or wireless interface capable of communicating with the machine learning assistance device 200. The communication I / F unit 110 may also be capable of communicating with devices and equipment of the sewerage facility (not shown), or with a management server that manages these devices and equipment.

[0020] The communication I / F unit 110 receives learning data transmitted from the machine learning assistance device 200. This learning data is obtained in a game played by a user of the user terminal device 300. The communication I / F unit 110 may also receive learning data including actual operation settings and treated water information from devices and equipment of the sewerage facility or their management server (not shown). Furthermore, the communication I / F unit 110 receives an optimization request, which requests the presentation of optimal operation settings for the devices and equipment, from, for example, a terminal device operated by an operator of the sewerage facility. The optimization request includes information such as weather information for the area where the sewerage facility is located and the target water quality after treatment.

[0021] The learning data acquisition unit 120 acquires learning data received by the communication I / F unit 110. That is, the learning data acquisition unit 120 acquires learning data obtained in a game played by a user of the user terminal device 300. The learning data acquisition unit 120 may also acquire learning data including actual operation settings and treated water information for a sewerage facility (not shown).

[0022] The machine learning control unit 130 performs machine learning using the training data acquired by the training data acquisition unit 120 to generate an AI model. Specifically, the machine learning control unit 130 performs supervised learning, unsupervised learning, or reinforcement learning based on training data including the amount and quality of sewage flowing into the sewerage facility, the arrangement and operation settings of the devices and equipment installed in the sewerage facility, and the amount and quality of treated water obtained as a result of water treatment, to generate an AI model. This AI model is a model that outputs optimal operation settings for the devices and equipment installed in the sewerage facility when, for example, the amount and quality of sewage flowing into the sewerage facility and the target amount and quality of treated water are input.

[0023] After generating the AI ​​model, if new learning data is acquired by the learning data acquisition unit 120, the machine learning control unit 130 continues machine learning using the new learning data and updates the AI ​​model.

[0024] The AI ​​model storage unit 140 stores the AI ​​model generated and updated by the machine learning control unit 130.

[0025] When the communication I / F unit 110 receives an optimization request, the operation setting optimization unit 150 uses the AI ​​model stored in the AI ​​model storage unit 140 to optimize the operation settings of the sewerage facility's equipment and facilities. Specifically, the operation setting optimization unit 150 inputs information included in the optimization request, such as weather information and the target post-treatment water quality, into the AI ​​model, and derives optimal operation settings for the equipment and facilities through inference using the AI ​​model. The operation setting optimization unit 150 then transmits the derived optimal operation settings to the terminal device that sent the optimization request.

[0026] Figure 3 is a block diagram showing the configuration of a machine learning assistance device 200 according to one embodiment. The machine learning assistance device 200 shown in Figure 3 includes a communication interface unit (hereinafter abbreviated as "communication I / F unit") 210, an environmental information acquisition unit 220, a facility information acquisition unit 230, a game control unit 240, an inflow data generation unit 250, a simulation processing unit 260, a game data generation unit 270, a progress data storage unit 280, and a learning data generation unit 290.

[0027] The communication I / F unit 210 is a wired or wireless interface capable of communicating with the machine learning device 100 and the user terminal devices 300-1 to 300-N. The communication I / F unit 210 receives user input from the user terminal devices 300-1 to 300-N and transmits game data including game screens and sounds to the user terminal devices 300-1 to 300-N. The communication I / F unit 210 also transmits learning data obtained through users playing games to the machine learning device 100.

[0028] The environmental information acquisition unit 220 acquires environmental information indicating the environment, such as the weather, of the area where the virtual sewerage facility is located in the game map that is the setting for the game played by the user. Specifically, when the environmental information acquisition unit 220 receives current location information from the user terminal devices 300-1 to 300-N of the users playing the game, it acquires weather information for the current locations of the user terminal devices 300-1 to 300-N from a predetermined weather information server or the like, and uses this information as environmental information indicating the weather in the game map. Alternatively, the environmental information acquisition unit 220 may directly acquire weather information for the current locations from the user terminal devices 300-1 to 300-N of the users playing the game, and use this information as environmental information indicating the weather in the game map.

[0029] The facility information acquisition unit 230 acquires facility information such as operation settings determined by the user for devices and equipment placed in the virtual sewerage facility during the game. Specifically, the facility information acquisition unit 230 acquires facility information from each of the user terminal devices 300-1 to 300-N, which is various settings determined by the user during the game for the virtual sewerage facility, including the water treatment method in the sewerage facility, the arrangement of the devices and equipment, and operation settings of the devices and equipment.

[0030] When a user starts a game, the game control unit 240 reads the progress status on the game map for each user from the progress data storage unit 280 and executes game processing. That is, the game control unit 240 reads game map information such as the population and industrial development level of cities on the user's game map, and sets a virtual game situation with game properties on this game map. Furthermore, when the game control unit 240 obtains simulation results of water treatment in a sewerage facility from the simulation processing unit 260, it updates the game status on the game map to reflect the simulation results.

[0031] Specifically, the game control unit 240 sets game conditions such as weather in accordance with environmental information on a game map including a virtual sewerage facility, and requests a user to make a decision regarding operation settings for the equipment and facilities of the sewerage facility.The game control unit 240 then obtains from the simulation processing unit 260 the results of a water treatment simulation in which the equipment and facilities are operated according to the user's operation settings, and updates game conditions such as city population and industrial development level on the game map based on the simulation results.In addition, instead of setting the weather on the game map according to the environmental information, the game control unit 240 may generate an event accompanied by abnormal weather such as heavy rain or drought, and request a user to make a decision regarding operation settings for the equipment and facilities of the sewerage facility.

[0032] The game control unit 240 stores the game status on the game map for each user as progress data in the progress data storage unit 280. The progress data includes environmental information indicating the weather on the game map, facility information indicating the operation settings of devices and facilities determined by the user, simulation results by the simulation processing unit 260, and the like.

[0033] The inflow data generating unit 250 generates inflow data related to sewage flowing into the virtual sewerage facility for each user based on the environmental information in the game map. Specifically, the inflow data generating unit 250 calculates the amount and quality of sewage flowing into the sewerage facility based on the weather, city population, industrial development level, etc. of the game map, and generates inflow data indicating the calculated inflow amount and quality.

[0034] The simulation processing unit 260 executes a simulation of water treatment in a virtual sewerage facility based on the inflow data and facility information, and outputs the volume and quality of treated water as simulation results. Specifically, the simulation processing unit 260 acquires the volume and quality of sewage flowing into the virtual sewerage facility from the inflow data, acquires the layout and operation settings of the devices and facilities in the virtual sewerage facility from the facility information, and executes a water treatment simulation using a model such as an ASM (Activated Sludge Model). The simulation processing unit 260 then outputs simulation data indicating the volume and quality of treated water obtained as a result of the simulation to the game control unit 240. In addition to the volume and quality of treated water, the simulation data may also include, for example, the treatment cost and sludge generation amount due to the operation of the devices and facilities in the sewerage facility.

[0035] The game data generation unit 270 generates game data including screens and sounds corresponding to the game map whose game situation has been updated by the game control unit 240. That is, the game data generation unit 270 generates, as game data for each user, for example, a map screen that displays the city population and industrial development level on the game map for each user, and a setting screen that sets the layout of devices and equipment in a virtual sewerage facility. Then, the game data generation unit 270 transmits the generated game data from the communication I / F unit 210 to the user terminal device 300.

[0036] The progress data storage unit 280 stores progress data indicating the game status for each user in association with the identification information of each user. Specifically, the progress data storage unit 280 stores progress data indicating the game status on the game map, such as game map information for each user, inflow data related to sewage flowing into the virtual sewerage facility, facility information related to devices and equipment in the virtual sewerage facility, and simulation data output from the simulation processing unit 260.

[0037] The learning data generation unit 290 generates learning data to be used for machine learning based on the game progress data stored in the progress data storage unit 280. Specifically, the learning data generation unit 290 acquires inflow data related to sewage flowing into the virtual sewerage facility, facility information related to the devices and equipment in the virtual sewerage facility, simulation data output from the simulation processing unit 260, and so on from the progress data storage unit 280, and generates learning data by associating these pieces of data. Then, the learning data generation unit 290 transmits the generated learning data from the communication I / F unit 210 to the machine learning device 100.

[0038] Next, a machine learning assistance method using the machine learning assistance device 200 configured as described above will be described with reference to the flowchart shown in Fig. 4. In Fig. 4, the operation of the machine learning assistance device 200 is shown together with the operation of the user terminal device 300.

[0039] When a user playing a game starts the game on the user terminal device 300 (step S101), identification information identifying this user is sent to the machine learning assistance device 200, and the game control unit 240 reads out progress data corresponding to the user's identification information from the progress data storage unit 280 (step S102).

[0040] The game control unit 240 then executes game processing on the game map of the read progress data. That is, for example, the population and industrial development level of cities may be set on the user's game map, and weather may be set based on environmental information. At this time, the environmental information acquisition unit 220 may acquire environmental information including the weather at the current location of the user terminal device 300, and the weather on the game map may be set based on this environmental information. Alternatively, instead of setting weather based on environmental information, an event accompanied by abnormal weather, such as heavy rain or drought, may occur on the game map. The game map with the situation set in this way is transmitted to the user terminal device 300 and output from the display, speaker, etc. of the user terminal device 300.

[0041] The user, checking the game map output from the user terminal device 300, inputs facility information for the virtual sewerage facility within the game map according to the game situation (step S104). Specifically, the user specifies the layout of the devices and equipment in the sewerage facility, and specifies the operation settings of the devices and equipment. The facility information input by the user is transmitted from the user terminal device 300 to the machine learning assistance device 200.

[0042] The facility information is then acquired by the facility information acquisition unit 230 (step S105) and input to the game control unit 240. As a result, facility information related to the devices and equipment of the virtual sewerage facility on the game map, along with environmental information such as weather and game map information such as city population and industrial development level, is acquired by the game control unit 240. This information is stored in the progress data storage unit 280 in association with the user's identification information.

[0043] Furthermore, the game control unit 240 outputs the environmental information and game map information to the inflow data generation unit 250, and outputs the facility information to the simulation processing unit 260. When the environmental information and game map information are input to the inflow data generation unit 250, inflow data is generated from this information (step S106). That is, inflow data indicating the amount and quality of sewage flowing into the virtual sewerage facilities in the game map is generated based on the weather, city population, and industrial development level in the game map. This inflow data may, for example, be such that the more precipitation there is, the greater the inflow amount, or the larger the city population and the higher the industrial development level, the lower the inflow water quality. The generated inflow data is output to the simulation processing unit 260 and stored as progress data by the progress data storage unit 280.

[0044] Then, when the inflow data and facility information are input to the simulation processing unit 260, a water treatment simulation is executed (step S107). Specifically, a water treatment simulation is executed using the quality and volume of sewage flowing into the virtual sewerage facility and the arrangement and operation settings of the devices and facilities in the virtual sewerage facility, and the volume and quality of the treated water are derived. Simulation data indicating the quality and volume of the treated water obtained as a result of the simulation is output to the game control unit 240.

[0045] When the simulation data is input to the game control unit 240, the game situation on the game map is updated based on the simulation data (step S108). Specifically, for example, the city population and industrial development level on the game map are updated based on the water treatment results indicated by the simulation data. Here, for example, if favorable results are obtained for the amount and quality of treated water, the game situation is updated so that the city population further increases and industry further develops.

[0046] When the game situation on the game map is updated, the game data generation unit 270 generates game data including screens and sounds corresponding to the game map (step S109). The generated game data is transmitted to the user terminal device 300 and output from the display, speaker, etc. of the user terminal device 300 (step S110).

[0047] Meanwhile, the simulation data input to the game control unit 240 is also stored in the progress data storage unit 280 (step S111). That is, simulation data indicating the volume and quality of the treated water and game map information such as updated city population and industrial development level are stored as progress data for each user in the progress data storage unit 280. As a result, environmental information, facility information, inflow data, simulation data, and game map information are stored in the progress data storage unit 280 as progress data.

[0048] Then, the learning data generation unit 290 generates learning data from the progress data stored in the progress data storage unit 280 (step S112). Specifically, the learning data is generated by associating the inflow data, facility information, and simulation data included in the progress data with each other.

[0049] The relationships between the various data contained in the learning data will now be described with reference to Fig. 5. As described above, the inflow data generating unit 250 generates inflow data indicating the amount and quality of sewage flowing into the virtual sewerage facility from game map information such as city population and industrial development level on the game map and environmental information such as weather. In addition, facility information including the water treatment method, arrangement and operation settings of devices and equipment in the virtual sewerage facility is obtained based on user input to the user terminal device 300.

[0050] The simulation processing unit 260 then performs a water treatment simulation based on the inflow data and facility information, and obtains simulation data indicating the volume, quality, and treatment cost of treated water, as shown in FIG. 5. Game map information, such as city population and industrial development level, is updated based on the simulation data. Of these various data, learning data is generated from the inflow data, facility information, and simulation data, as surrounded by a dashed line in FIG. 5. In other words, the learning data is generated by associating the simulation data based on the inflow data and facility information with each other. The generated learning data is transmitted to the machine learning device 100 (step S113) and used by the machine learning device 100 to generate or update an AI model.

[0051] In this way, through gameplay, users can acquire inflow data and facility information for virtual sewerage facilities under various conditions, as well as simulation data showing the results of water treatment, thereby collecting a large amount of diverse learning data. In other words, this makes it easy to collect learning data for machine learning, thereby improving the accuracy of AI models.

[0052] Next, the operation of the machine learning device 100 that performs machine learning using training data will be described with reference to the flowchart shown in FIG.

[0053] The learning data transmitted from the machine learning support device 200 is received by the communication I / F unit 110. The communication I / F unit 110 may also receive actual operational settings and treated water information as learning data from devices and equipment installed in the sewerage facility (not shown). The received learning data is acquired by the learning data acquisition unit 120 (step S201). Then, the machine learning control unit 130 performs machine learning using the learning data to generate an AI model (step S202). That is, supervised learning, unsupervised learning, or reinforcement learning is performed based on the learning data, which includes the amount and quality of sewage flowing into the sewerage facility, the arrangement and operational settings of the devices and equipment installed in the sewerage facility, and the amount and quality of treated water obtained as a result of water treatment, to generate an AI model.

[0054] The generated AI model is stored in the AI ​​model storage unit 140. Thereafter, when new learning data is acquired by the learning data acquisition unit 120, the machine learning control unit 130 executes machine learning using the new learning data, and the AI ​​model stored in the AI ​​model storage unit 140 is updated. This AI model is a model that outputs optimal operation settings for the devices and equipment installed in the sewerage facility when, for example, the amount and quality of sewage flowing into the sewerage facility and the target amount and quality of treated water are input.

[0055] Then, when an optimization request for presenting optimal operation settings for devices and equipment is received by the communication I / F unit 110 from a terminal device operated by an operator of the sewerage facility, for example, conditions for optimizing the operation settings are input to the operation setting optimization unit 150 (step S203). That is, the optimization request includes information such as weather information for the area where the sewerage facility is located and the target water quality after treatment, and this weather information and target water quality information are input as conditions.

[0056] When the conditions are input, the operation setting optimization unit 150 reads the AI ​​model from the AI ​​model storage unit 140 and performs inference on the operation settings using the AI ​​model (step S204). That is, inference is performed by the AI ​​model based on the input conditions, and optimal operation settings for the equipment and facilities of the sewerage facility are derived. The derived optimal operation settings are sent to the terminal device that sent the optimization request, and output by this terminal device (step S205). This allows the operator of the sewerage facility operating the terminal device to understand the optimal operation settings for the equipment and facilities, and even an operator with little experience or knowledge, for example, can set the operation settings for the equipment and facilities of the sewerage facility.

[0057] As described above, according to this embodiment, the user determines the operation settings of the equipment and facilities of the virtual sewerage facility during the game, a simulation of water treatment when the user's operation settings are applied is performed, and learning data is generated that associates the user's operation settings with the simulation results. Therefore, through the user's gameplay, a large amount of diverse learning data about the virtual sewerage facility can be collected under various conditions. In other words, learning data used for machine learning can be easily collected, and the accuracy of the AI ​​model can be improved.

[0058] The machine learning assistance device 200 according to the embodiment can be configured using a processor and a memory. Fig. 7 is a block diagram showing an example of the hardware configuration of the machine learning assistance device 200 according to the embodiment. As shown in Fig. 7, the machine learning assistance device 200 includes a processor 201, a memory 202, a storage 203, and a communication I / F 204.

[0059] The processor 201 includes, for example, a central processing unit (CPU), a field programmable gate array (FPGA), or a digital signal processor (DSP), and controls the entire machine learning support device 200 and executes various types of arithmetic processing.

[0060] The memory 202 includes, for example, a random access memory (RAM) or a read only memory (ROM), and stores information used in the arithmetic processing executed by the processor 201.

[0061] The storage 203 includes, for example, a hard disk drive (HDD) or a solid state drive (SSD), and stores various types of data.

[0062] The communication I / F 204 is a wired or wireless interface that connects to a network, and transmits and receives various types of data via the network.

[0063] The processing performed by the machine learning assistance device 200 described in the above embodiment can also be written as a computer-executable program. In this case, the program can be stored on a computer-readable, non-transitory recording medium and installed on a computer. Examples of such recording media include portable recording media such as CD-ROMs, DVD discs, and USB memory, as well as semiconductor memories such as flash memories. [Explanation of symbols]

[0064] 110, 210 Communication I / F section 120 Learning data acquisition unit 130 Machine Learning Control Unit 140 AI model memory unit 150 Operation setting optimization section 220 Environmental Information Acquisition Department 230 Equipment information acquisition department 240 Game control unit 250 Inflow data generation unit 260 Simulation Processing Section 270 Game Data Generation Unit 280 Progress Data Storage Unit 290 Learning Data Generation Unit

Claims

1. a game control unit that sets a game situation on a game map on which a virtual sewerage facility is arranged; an acquisition unit that acquires equipment information including operation settings determined by a user in response to the game situation set by the game control unit, the operation settings being for equipment provided in the sewerage facility; an inflow data generating unit that generates inflow data indicating the amount and quality of sewage flowing into the sewerage facility based on at least one of information on weather, population, and industrial development level in the game map; a simulation processing unit that simulates water treatment in the sewerage facility based on the facility information acquired by the acquisition unit and the inflow data generated by the inflow data generation unit; a generation unit that generates learning data to be used in machine learning by associating the facility information, the inflow data, and a simulation result by the simulation processing unit; A machine learning assistance device having the above.

2. The game control unit The game situation is updated based on the simulation result by the simulation processing unit. The machine learning assistance device according to claim 1 .

3. The simulation processing unit The amount and quality of treated water obtained in the sewerage facility are derived as simulation results. The machine learning assistance device according to claim 1 .

4. The game control unit The game situation is set to the population and industrial development of the city on the game map. The machine learning assistance device according to claim 1 .

5. The acquisition unit Acquire facility information that further includes the location of facilities provided in the sewerage facility, the location of facilities being determined by the user for the game situation. The machine learning assistance device according to claim 1 .

6. setting a game situation on a game map in which a virtual sewerage facility is arranged; Acquiring facility information including operation settings determined by a user for a set game situation, the operation settings being for facilities provided in the sewerage facility; generating inflow data indicating the amount and quality of sewage flowing into the sewerage facility based on at least one of information on weather, population, and industrial development in the game map; simulating water treatment in the sewerage facility based on the acquired facility information and the generated inflow data; generating learning data to be used in machine learning by associating the facility information, the inflow data, and the simulation results of the simulation; A machine learning assisted method having

Images