Program, information processing device, method, and system

Abstract

To further facilitate understanding of a relation between a plurality of experimental conditions and a predicted experimental result in prediction of the experimental result using a prediction model.SOLUTION: A program comprises: executing machine learning based on a data set including a plurality of combinations of experimental condition data relating to three or more types of experimental conditions and experimental result data relating to one or more types of experimental results, to generate a prediction model with the three or more types of experimental conditions defined as explanatory variables and the experimental results defined as objective variables; compressing the dimension of the explanatory variables into two dimensions composed of a first variable and a second variable; using a value of the explanatory variable restored by assigning a prescribed value to each of the first variable and the second variable, and a prediction model, to infer a value of an objective variable; and presenting, on a plane with the value of the first variable as a first axis and the value of the second variable as a second axis, a distribution map of the value of the inferred objective variable in such a manner that the magnitude of the value of the inferred objective variable can be recognized.SELECTED DRAWING: Figure 9

Description

[Technical field]

[0001] The present disclosure relates to a program, an information processing device, a method, and a system. [Background technology]

[0002] For example, in research into new material development, there is a technology that performs machine learning based on data on multiple experimental conditions and data on one or more experimental results, and generates a prediction model that outputs a predicted value of the experimental result when a predetermined experimental condition is input. In addition, in inference using this prediction model, there is a data analysis method or data display method for facilitating understanding of the relationship between an input value and an output value. For example, the method of Patent Document 1 provides a data analysis method that targets two types of data, input data representing the multiple input values ​​and output values ​​having a predetermined relationship, and output data representing the multiple output values, and includes a step of determining at least one of a first index and a second index in an objective function space using the multiple output values ​​as objective functions, the first index being the distance of at least two objective function values ​​among the multiple objective function values ​​to a preset value, and the second index being expressed as a ratio of at least two objective function values ​​among the multiple objective function values. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2016-099737 A Summary of the Invention [Problem to be solved by the invention]

[0004] In the technology for predicting experimental results using a predictive model, there is a demand for technology that makes it easier to understand the relationship between experimental conditions and the predicted experimental results.

[0005] In view of the above problems, the present invention aims to make it easier to understand the relationship between multiple experimental conditions and predicted experimental results when predicting experimental results using a prediction model. [Means for solving the problem]

[0006] A program for operating a computer having a processor and a memory, the program executing the following steps: accepting a dataset including multiple combinations of experimental condition data related to three or more types of experimental conditions and experimental result data related to one or more types of experimental results; generating a prediction model using machine learning based on the dataset to set the three or more types of experimental conditions as three or more explanatory variables and the experimental result as an objective variable; compressing the dimensions of the explanatory variables to two dimensions consisting of a first variable and a second variable; inferring the value of the objective variable using the values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable and the second variable, respectively, and the prediction model; and presenting a distribution diagram of the values ​​of the objective variable inferred in the inference step on a plane with the value of the first variable as a first axis and the value of the second variable as a second axis, in a manner that allows the magnitude of the value of the objective variable inferred in the inference step to be recognized. Effect of the Invention

[0007] In predicting experimental results using a predictive model, it is possible to more easily understand the relationship between multiple experimental conditions and the predicted experimental results. [Brief description of the drawings]

[0008] [Figure 1] 1 is a block diagram showing an example of the overall configuration of a system 1. FIG. [Diagram 2] FIG. 2 is a diagram illustrating an example of a functional configuration of a terminal device 10. [Diagram 3] FIG. 2 is a diagram illustrating an example of a functional configuration of a server 20. [Figure 4] FIG. 2 is a diagram showing the data structure of a user table 2021. [Diagram 5]FIG. 2 shows the data structure of an experiment data table 2022. [Figure 6] FIG. 23 is a diagram showing the data structure of a model table 2023. [Figure 7] FIG. 2 is a diagram for explaining an example of the operations of the terminal device 10 and the server 20 when generating a prediction model based on experimental data. [Figure 8] 10 is a diagram for explaining an example of the operation of the terminal device 10 and the server 20 when generating a distribution map of predicted values ​​of experimental results and presenting it to a user. FIG. [Figure 9] 1A and 1B are schematic diagrams showing an example of a screen displaying a distribution map of predicted values ​​of experimental results and an enlarged view thereof; [Figure 10] FIG. 2 is a diagram for explaining an example of the operations of the terminal device 10 and the server 20 when optimizing a prediction model and presenting the optimization result to a user. [Figure 11] FIG. 13 is a schematic diagram illustrating an example of a screen displaying a result of optimization. [Figure 12] 13 is a schematic diagram showing an example of a screen that presents values ​​of experimental conditions and predicted values ​​of experimental results at points received from the user to the user. FIG. [Figure 13] FIG. 2 is a block diagram showing the basic hardware configuration of a computer 90. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, an embodiment will be described with reference to the drawings. Note that the embodiment described below does not unduly limit the contents of the present disclosure described in the claims. In addition, not all of the configurations described in the embodiment are necessarily essential components of the present disclosure.

[0010] <Summary> The system 1 according to the present embodiment performs machine learning based on a data set including a combination of experimental condition data on experimental conditions and experimental result data on experimental results, and generates a prediction model with the experimental conditions as explanatory variables and the experimental results as objective variables. The system 1 compresses the dimensions of the explanatory variables into two dimensions represented by a first variable and a second variable. The system 1 restores the values ​​of the explanatory variables by substituting predetermined values ​​for the first variable and the second variable, respectively. The system 1 infers the value of the objective variable using the restored values ​​of the explanatory variables and the prediction model. The system 1 presents a distribution diagram of the inferred objective variable values ​​to the user on a plane with the value of the first variable as the first axis and the value of the second variable as the second axis in a manner that allows the user to recognize the magnitude of the inferred objective variable value.

[0011] <1 Overall composition> Fig. 1 is a diagram showing an overall configuration of a system 1 of the present embodiment. As shown in Fig. 1, the system 1 includes a terminal device 10 and a server 20. The terminal device 10 and the server 20 are connected to each other via a network 80 using a wired or wireless communication standard so as to be able to communicate with each other. The network 80 is realized by, for example, the Internet and / or a communication network provided by a communication carrier. In the illustrated example, a plurality of terminal devices 10 are included in the system 1.

[0012] Although FIG. 1 shows a case where the server 20 is one computer, the server 20 may be realized by combining a plurality of computers.

[0013] The terminal device 10 is an information processing device used by a user who uses the service provided by the server 20. For example, the terminal device 10 is an information processing device used by a test designer who creates a test design document. Also, for example, the terminal device 10 is an information processing device used by a developer who develops software. Also, the terminal device 10 is an information processing device used by a tester who performs tests based on a test design document.

[0014] The terminal device 10 is realized, for example, by a desktop PC (Personal Computer), a laptop PC, a head mounted display, etc. The terminal device 10 may also be a portable computer such as a smartphone or a tablet terminal.

[0015] 1, the terminal device 10 includes a communication IF (Interface) 12, an input device 13, an output device 14, a memory 15, a storage 16, and a processor 19. The communication IF 12, the input device 13, the output device 14, the memory 15, the storage 16, and the processor 19 are connected to each other so as to be able to communicate with each other via a bus, for example.

[0016] The communication IF 12 is an interface for transmitting and receiving signals so that the terminal device 10 can communicate with external devices. The input device 13 is an input device for receiving input operations from a user (employee). The input device 13 includes, for example, a touch panel, a touch pad, a pointing device such as a mouse, a keyboard, etc. The output device 14 is an output device for presenting information to a user. The output device 14 includes, for example, a display, a speaker, etc.

[0017] The memory 15 is for temporarily storing programs and data processed by the programs, and is realized by a volatile memory such as a dynamic random access memory (DRAM). The storage 16 is a storage device for saving data and is realized by a non-volatile memory such as a flash memory or a hard disk drive (HDD). The processor 19 is hardware for executing an instruction set described in a program, and is configured by an arithmetic unit, a register, peripheral circuits, and the like.

[0018] The server 20 is an information processing device that provides a user with predicted values ​​of experimental results for predetermined experimental conditions. The server 20 is, for example, a computer connected to a network 80.

[0019] 1, the server 20 includes a communication IF 22, an input / output IF 23, a memory 25, a storage 26, and a processor 29. The communication IF 22, the input / output IF 23, the memory 25, the storage 26, and the processor 29 are communicatively connected to each other via, for example, a bus.

[0020] The communication IF 22 is an interface for transmitting and receiving signals so that the server 20 can communicate with an external device. The input / output IF 23 functions as an interface with an input device for receiving input operations from a user and an output device for presenting information to the user. The memory 25 is for temporarily storing programs and data processed by the programs, etc., and is realized by a volatile memory such as a DRAM.

[0021] The storage 26 is a storage device for saving data, and is realized by a non-volatile memory such as a flash memory or a HDD. The storage 26 does not necessarily have to be realized by a single circuit. The storage 26 may be realized by, for example, a plurality of storage circuits. The processor 29 is hardware for executing an instruction set described in a program, and is configured by an arithmetic unit, a register, peripheral circuits, etc.

[0022] <1.1 Terminal device configuration> Fig. 2 is a block diagram showing an example of the configuration of the terminal device 10 shown in Fig. 1. As shown in Fig. 2, the terminal device 10 includes a communication unit 120, an input device 13, an output device 14, an audio processing unit 17, a microphone 171, a speaker 172, a camera 160, a position information sensor 150, a storage unit 180, and a control unit 190. The blocks included in the terminal device 10 are electrically connected to each other, for example, by a bus or the like.

[0023] The communication unit 120 performs processing for the terminal device 10 to communicate with other devices. The communication unit 120 performs transmission processing on a signal generated by the control unit 190 and transmits the signal to the outside (for example, the server 20). The communication unit 120 performs reception processing on a signal received from the outside and outputs the signal to the control unit 190.

[0024] The input device 13 is a device for inputting instructions or information by a user who operates the terminal device 10. The input device 13 is realized, for example, by a touch-sensitive device 131 or the like in which an instruction is input by touching an operation surface. In the case where the terminal device 10 is a PC or the like, the input device 13 may be realized by a reader, a keyboard, a mouse, or the like. The input device 13 converts an instruction input by a user into an electrical signal, and outputs the electrical signal to the control unit 190. Note that the input device 13 may include, for example, a receiving port that receives an electrical signal input from an external input device.

[0025] The output device 14 is a device for presenting information to a user who operates the terminal device 10. The output device 14 is realized, for example, by a display 141 or the like. The display 141 displays data according to the control of the control unit 190. The display 141 is realized, for example, by an LCD (Liquid Crystal Display) or an organic EL (Electro-Luminescence) display or the like.

[0026] The audio processing unit 17 performs, for example, digital-analog conversion processing of an audio signal. The audio processing unit 17 converts a signal provided from the microphone 171 into a digital signal and provides the converted signal to the control unit 190. The audio processing unit 17 also provides the audio signal to the speaker 172. The audio processing unit 17 is realized, for example, by a processor for audio processing. The microphone 171 accepts audio input and provides an audio signal corresponding to the audio input to the audio processing unit 17. The speaker 172 converts the audio signal provided from the audio processing unit 17 into audio and outputs the audio to the outside of the terminal device 10.

[0027] The camera 160 is a device for receiving light with a light receiving element and outputting the received light as an image capturing signal.

[0028] The position information sensor 150 is a sensor that detects the position of the terminal device 10, and is, for example, a GPS (Global Positioning System) module. The GPS module is a receiving device used in a satellite positioning system. In the satellite positioning system, signals are received from at least three or four satellites, and the current position of the terminal device 10 equipped with the GPS module is detected based on the received signals. The position information sensor 150 may detect the current position of the terminal device 10 from the position of the wireless base station to which the terminal device 10 is connected.

[0029] The storage unit 180 is realized by, for example, the memory 15, the storage 16, etc., and stores data and programs used by the terminal device 10. The storage unit 180 stores, for example, user information 181. The user information 181 includes information about a user who uses the terminal device 10.

[0030] The control unit 190 is realized by the processor 19 reading a program stored in the storage unit 180 and executing instructions included in the program. The control unit 190 controls the operation of the terminal device 10. The control unit 190 performs the functions of an operation reception unit 191, a transmission / reception unit 192, and a presentation control unit 193 by operating according to the program.

[0031] The operation reception unit 191 performs processing for receiving instructions or information input from the input device 13. Specifically, for example, the operation reception unit 191 receives instructions or information input from the touch-sensitive device 131 or the like.

[0032] Furthermore, the operation reception unit 191 receives voice instructions input from the microphone 171. Specifically, for example, the operation reception unit 191 receives a voice signal input from the microphone 171 and converted into a digital signal by the voice processing unit 17. For example, the operation reception unit 191 analyzes the received voice signal and extracts a predetermined noun, thereby acquiring an instruction from the user.

[0033] The transmitting / receiving unit 192 performs processing for the terminal device 10 to transmit and receive data to and from an external device such as the server 20 in accordance with a communication protocol. Specifically, for example, the transmitting / receiving unit 192 transmits instructions or information input by a user to the server 20. In addition, the transmitting / receiving unit 192 receives information provided by the server 20.

[0034] The presentation control unit 193 controls the output device 14 to present information output from the server 20 to the user. Specifically, for example, the presentation control unit 193 causes the display 141 to display a screen for inputting experimental condition data and experimental result data. Also, for example, the presentation control unit 193 causes the display 141 to display a distribution map of the experimental result data.

[0035] <1.2 Functional configuration of the server> 3 is a block diagram showing a functional configuration of the server 20 included in the system 1 of the present embodiment. As shown in FIG. 3, the server 20 fulfills the functions of a communication unit 201, a storage unit 202, and a control unit 203.

[0036] The communication unit 201 performs processing for the server 20 to communicate with external devices.

[0037] The storage unit 202 stores data and programs used by the server 20. For example, the storage unit 202 has a user table 2021, an experiment data table 2022, a model table 2023, etc. The tables stored in the storage unit 202 are not limited to these.

[0038] The user table 2021 is a table for storing user information, the details of which will be described later.

[0039] The experimental data table 2022 is a table for storing a data set of experimental data, which will be described in detail later.

[0040] The model table 2023 is a table for storing information related to a prediction model, the details of which will be described later.

[0041] The control unit 203 is realized by the processor 29 reading a program stored in the storage unit 202 and executing instructions included in the program. The control unit 203 controls the operation of the server 20. Specifically, for example, the control unit 203 operates according to the program stored in the storage unit 202 to fulfill the functions of a reception control module 2031, a transmission control module 2032, a learning module 2033, a dimensional compression module 2034, an inverse conversion module 2035, an inference module 2036, a distribution map generation module 2037, an optimization module 2038, and a presentation control module 2039.

[0042] The reception control module 2031 controls the process in which the server 20 receives a signal from an external device in accordance with a communication protocol.

[0043] The transmission control module 2032 controls the process in which the server 20 transmits signals to external devices in accordance with a communication protocol.

[0044] The learning module 2033 generates a prediction model (trained model) by performing machine learning using a predetermined machine learning algorithm based on the experimental data. Specifically, for example, the learning module 2033 performs preprocessing such as removing outliers, normalizing data, and complementing missing values ​​on the experimental data stored in the experimental data table 2022. The learning module 2033 divides the experimental data and creates data sets such as training data, test data, and validation data. The learning module 2033 performs learning using the experimental condition data as input data and the experimental result data as output data from the experimental data, and generates a prediction model.

[0045] The machine learning algorithm used in this embodiment may be any algorithm that constructs a known learning model such as a general regression model or classification model, and examples of such algorithms include regression models such as Gaussian process regression, linear regression, neural nets, and decision trees (LightGBM, etc.). In this embodiment, the learning module 2033 performs learning using Gaussian process regression as a machine learning algorithm, and generates a prediction model with experimental condition data as explanatory variables and experimental result data as objective variables. Since Gaussian process regression is a non-parametric model, it is preferable in that it does not require a model to be assumed in advance as in linear regression, and fitting can be performed flexibly. In addition, since Gaussian process regression is a Bayesian estimation, it is possible to evaluate the uncertainty of the prediction. This makes it preferable in that it has high scalability in the proposed part of the experimental candidate points. In addition, while the number of data required for learning is likely to be enormous in the case of neural nets and decision tree analysis, in the case of Gaussian process regression, the part with less data can be interpreted as having less uncertainty in the prediction, and therefore it is suitable for an experimental design method such as this embodiment.

[0046] The dimension reduction module 2034 reduces the dimension of the explanatory variables of the prediction model to two dimensions consisting of a first variable and a second variable. Specifically, for example, the dimension reduction module 2034 reduces the dimension of the explanatory variables of the prediction model to two dimensions using a known dimension reduction method. Examples of the dimension reduction method include Principal Component Analysis (PCA) and a method using a neural network (AutoEncoder, etc.). In this embodiment, the dimension reduction module 2034 performs principal component analysis on the explanatory variables of the prediction model and converts them into principal components consisting of the first variable and the second variable.

[0047] The inverse transformation module 2035 inversely transforms a function including a first variable and a second variable, and restores the dimensions before compression. Specifically, the inverse transformation module 2035 inversely transforms the compressed variables using a predetermined method corresponding to the dimension compression method used by the dimension compression module 2034. For example, when the inverse transformation module 2035 receives the values ​​of the first variable and the second variable for the compressed prediction model, it inversely transforms the compressed explanatory variables to obtain the values ​​of the explanatory variables before compression that correspond to the received first variable and second variable.

[0048] The inference module 2036 inputs the values ​​of the explanatory variables into the prediction model and infers the predicted values ​​of the experimental results.

[0049] The distribution map generating module 2037 generates a distribution map of the predicted values ​​of the experimental results on a two-dimensional plane with the value of the first variable as the first axis and the value of the second variable as the second axis in such a manner that the magnitude of the predicted values ​​of the experimental results can be recognized. Specifically, for example, the distribution map generating module 2037 generates a contour map showing the magnitude of the predicted values ​​of the experimental results inferred by the inference module 2036 with contour lines on a plane with the value of the first variable as the first axis and the value of the second variable as the second axis. The distribution map generating module 2037 colors each area separated by the contour lines in a different color. This allows the user to easily recognize the magnitude of the predicted values ​​of the experimental results.

[0050] The optimization module 2038 executes optimization of the prediction model. Specifically, for example, the optimization module 2038 analyzes the prediction model under given predetermined constraint conditions, and obtains values ​​of explanatory variables and values ​​of objective variables that satisfy a predetermined purpose. As the optimization method, a known optimization method such as a method using a genetic algorithm (NSGA-II, etc.) is appropriately used.

[0051] At least one of the conditions regarding the search range of the experimental conditions (explanatory variables) and the condition regarding the range of values ​​that the experimental results (objective variables) should satisfy is given to the optimization module 2038 as a constraint condition.

[0052] The condition regarding the search range of explanatory variables is, for example, that for one or more predetermined explanatory variables, at least one of the values ​​of the explanatory variables or the sum, difference, product, and ratio of the values ​​of multiple explanatory variables is within a predetermined range. Examples of the predetermined range include being equal to or greater than a specific value, being greater than a specific value, being equal to or less than a specific value, and the like.

[0053] The condition regarding the range of the objective variable value is, for example, that for one or more predetermined objective variables, at least one of the objective variable value or the sum, difference, product, or ratio of the values ​​of multiple objective variables is within a predetermined range. Examples of the predetermined range include being equal to or greater than a specific value, being greater than a specific value, being equal to or less than a specific value, and the like.

[0054] Also, for example, when performing optimization by associating a first prediction model that predicts a first objective variable with the first to fifth explanatory variables with a second prediction model that predicts a second objective variable with the first to fifth explanatory variables, a constraint condition may be set on at least one of the first prediction model and the second prediction model, and the first prediction model and the second prediction model may be optimized by searching the values ​​of the first to fifth explanatory variables that satisfy the constraint condition in the first prediction model and the second prediction model. In other words, by sharing the constraint condition at the time of optimization between the prediction models, a plurality of prediction models may be associated and each prediction model may be optimized. This makes it possible to derive the trade-off relationship between the objective variables when optimizing the first objective variable and the second objective variable.

[0055] The presentation control module 2039 controls the process for presenting to the user the information generated by each module of the control unit 203. For example, the presentation control module 2039 presents to the user the distribution map generated by the distribution map generation module 2037.

[0056] <2 Data Structure> 4 to 6 are diagrams showing the data structures of tables stored in the server 20. Note that Fig. 4 to 6 are merely examples and do not exclude data that is not listed. In addition, data that is listed in the same table may be stored in separate storage areas in the storage unit 202.

[0057] Fig. 4 is a diagram showing the data structure of the user table 2021. The user table 2021 shown in Fig. 4 is a table having columns such as a user name, an email address, and the like, with a user ID as a primary key.

[0058] The item "user ID" is an item for storing an ID for identifying a user.

[0059] The item "user name" is an item for storing the name of the user. The user name may be set to any character string such as the user's name or nickname.

[0060] The item "email address" is an item for storing the user's email address.

[0061] Fig. 5 is a diagram showing the data structure of the experiment data table 2022. The experiment data table 2022 shown in Fig. 5 is a table having columns such as experiment conditions, experiment results, etc., with a data ID as a primary key.

[0062] The storage unit 202 stores the experimental data sets in different tables for each data set. That is, the storage unit 202 has an experimental data table 2022a, an experimental data table 2022b, ... corresponding to each data set, and stores the experimental data included in each data set in a different experimental data table for each data set. Each experimental data table is identified by a table ID (not shown).

[0063] The item "data ID" is an item for storing an ID for identifying the experiment data.

[0064] The item "experimental condition" is an item that stores experimental condition data. Specifically, the item "experimental condition" stores the parameters of each experimental condition that the user changed in the experiment. The item "experimental condition" has a different column for each type of parameter that the user changed in the experiment. For example, a predetermined type of experimental condition data, such as temperature, is stored in the item "X1," humidity, etc., in the item "X2," etc., as appropriate.

[0065] The item "Experimental Results" is an item that stores experimental result data. Specifically, the item "Experimental Results" stores the values ​​of each experimental result obtained by the user in an experiment. The item "Experimental Results" has a different column for each type of experimental result obtained by the user in an experiment. For example, predetermined types of experimental result data, such as the viscosity of the product in item "Y1" and the hardness of the product in item "Y2", are appropriately stored.

[0066] Fig. 6 is a diagram showing the data structure of the model table 2023. The model table 2023 shown in Fig. 6 is a table having columns such as a user ID, a table ID, a model name, and a model file, with a model ID as a primary key.

[0067] The item "model ID" is an item that stores an ID that identifies a prediction model.

[0068] The item "user ID" is an item for storing an ID for identifying a user.

[0069] The item "table ID" is an item that stores an ID that identifies the experimental data table used to generate the prediction model identified by the model ID.

[0070] The item "model name" is an item that stores the name of the prediction model.

[0071] The item "model file" is an item that stores a data file of a prediction model. The item "model file" stores a data file of a prediction model saved in a predetermined format. The item "model file" may also store reference information (path) for a data file of a prediction model located in another location.

[0072] <3 operations> The operations of the terminal device 10 and the server 20 will now be described.

[0073] (Modeling process) FIG. 7 is a diagram for explaining an example of the operations of the terminal device 10 and the server 20 when a prediction model is generated based on experimental data obtained by a user performing an experiment.

[0074] First, the user prepares a data set including a combination of experimental condition data on a plurality of types of experimental conditions (also referred to as experimental parameters, or factors in experimental design) and experimental result data on one or more types of experimental results (also referred to as characteristic values, or responses in experimental design). For example, the experimental condition data may include the numerical values ​​of five types of experimental conditions, namely, temperature (X1), humidity (X2), the ratio of the first material (X3), the ratio of the second material (X4), and the ratio of the third material (X5), as experimental conditions, when the five types of experimental conditions are changed. For example, the experimental result data may include the numerical values ​​of two types of experimental results, namely, the viscosity (Y1) of the product and the hardness (Y2) of the product, when an experiment is performed by changing X1 to X5, as each experimental result. Note that the types of experimental conditions and the types of experimental results are not limited to the above examples, and can be set appropriately according to the purpose of the experiment.

[0075] In step S101, the terminal device 10 accepts a request to execute modeling based on experimental data. Specifically, for example, a user operates the terminal device 10 to input experimental data. The user performs an operation to transmit information including the user's user ID and the input experimental data to the server 20 as a request to execute modeling. The user may also perform an operation to transmit a data file of the experimental data saved in a predetermined format such as a CSV file to the server 20.

[0076] In step S102, the control unit 203 receives a data set of experimental data and a user ID from the terminal device 10. The control unit 203 creates a new experimental data table 2022, and stores the received experimental condition data in a different column for each type of experimental condition. The control unit 203 stores the received experimental result data in a different column for each type of experimental result. The control unit 203 assigns a predetermined table ID to the created experimental data table 2022, and stores it in the storage unit 202.

[0077] In step S103, the control unit 203 generates a prediction model with the experimental conditions as explanatory variables and the experimental results as objective variables by performing machine learning based on the data set using the learning module 2033. Specifically, for example, the learning module 2033 performs preprocessing such as removing outliers, normalizing data, and complementing missing values ​​on the experimental data received in step S102. The learning module 2033 divides the experimental data and creates datasets such as training data, test data, and validation data. The learning module 2033 performs learning using the experimental condition data as input data and the experimental result data as output data from the experimental data, and generates a prediction model. As a result, the learning module 2033 generates a prediction model with the experimental condition data as explanatory variables and the experimental result data as objective variables.

[0078] The learning module 2033 stores the user's user ID in the user ID field of the new record in the model table 2023. The learning module 2033 stores the table ID assigned to the experiment data table 2022 in step S102 in the table ID field of the new record. The learning module 2033 stores the model name in the model name field of the new record. The model name may be received from the user in step S101, or may be determined according to a predetermined naming rule. The learning module 2033 stores the data file of the generated prediction model in a predetermined format in the model file field of the new record.

[0079] The control unit 203 transmits information for notifying the user that the generation of the prediction model is completed to the terminal device 10 via the transmission control module 2032. The terminal device 10 notifies the user that the generation of the prediction model is completed via the display 141 or the like.

[0080] (Distribution map generation process) FIG. 8 is a diagram for explaining an example of the operation of the terminal device 10 and the server 20 when a distribution map of predicted values ​​of experimental results is generated based on a user request and presented to the user.

[0081] First, the user operates the terminal device 10 to access the server 20. The control unit 203 transmits information for presenting a screen for accepting a request to generate a distribution diagram of predicted values ​​of experimental results to the terminal device 10. Furthermore, for example, the control unit 203 searches the model table 2023 based on the user ID of the user, and acquires information on the prediction model associated with the user. The control unit 203 transmits the information on the prediction model associated with the user to the terminal device 10.

[0082] In step S201, the terminal device 10 accepts a request to generate a distribution map from a user. Specifically, for example, the terminal device 10 presents information on a prediction model associated with the user to the user via the display 141. The user selects a prediction model for which a distribution map is to be generated. The terminal device 10 transmits the model ID of the prediction model selected by the user to the server 20 as a request to generate a distribution map.

[0083] The control unit 203 may execute step S202 (to be described later) immediately after step S103 of the modeling process described above, without executing step S201. In this case, the prediction model generated in step S103 is used to generate a distribution diagram.

[0084] In step S202, the control unit 203 compresses the dimension of the explanatory variables of the prediction model into two dimensions consisting of a first variable and a second variable by the dimension compression module 2034. Specifically, for example, the control unit 203 searches the model ID item of the model table 2023 based on the accepted model ID, and acquires the prediction model for which a distribution diagram is to be generated. The dimension compression module 2034 executes a principal component analysis with the number of components set to 2 on the explanatory variables of the acquired prediction model, and compresses the explanatory variables of the prediction model into the first variable and the second variable.

[0085] In step S203, the control unit 203 generates a two-dimensional grid by the distribution map generation module 2037. Specifically, for example, the distribution map generation module 2037 creates a two-dimensional grid in which the coordinates are arranged in a lattice pattern in a two-dimensional space having values ​​in the first axis direction (first coordinate) and values ​​in the second axis direction (second coordinate). The conditions for the intervals between the coordinates are determined in advance by the user or the business providing the service of this embodiment.

[0086] The control unit 203 performs inverse transformation of the explanatory variables compressed in step S202 for each coordinate of the two-dimensional grid, with the first coordinate as the value of the first variable and the second coordinate as the value of the second variable, using the inverse transformation module 2035. This restores the values ​​of the explanatory variables based on the values ​​of the first variable and the second variable corresponding to each coordinate of the two-dimensional grid.

[0087] In step S204, the control unit 203 infers a predicted value of the experimental result by the inference module 2036. Specifically, for example, the inference module 2036 inputs the inversely transformed explanatory variable values ​​for each coordinate of the two-dimensional grid to a prediction model, and infers the objective variable (predicted value of the experimental result). As a result, the predicted value of the experimental result corresponding to each coordinate of the two-dimensional grid is inferred.

[0088] Here, since information loss occurs in the process of compressing the dimensions of the prediction model, the accuracy of inference may decrease in the compressed prediction model. Therefore, in steps S203 to S204, the control unit 203 does not input the value of the first variable and the value of the second variable directly to the compressed model, but instead inversely converts the value of the first variable and the value of the second variable into the value of the explanatory variable by the inverse conversion module 2035, and then inputs the value to the prediction model before the dimension compression, thereby inferring the objective variable. As a result, even if a two-dimensional value such as a predetermined coordinate on a plane is given to a prediction model having explanatory variables of three or more dimensions, inference can be performed while maintaining accuracy.

[0089] In step S205, the control unit 203 uses the distribution diagram generation module 2037 to draw a distribution diagram of the values ​​of the objective variables on a plane with the value of the first variable as the first axis and the value of the second variable as the second axis in a manner that allows the magnitude of the inferred objective variable value to be recognized. Specifically, for example, the distribution diagram generation module 2037 draws a distribution diagram of the predicted values ​​of the experimental results on a plane with the value of the first variable as the first axis and the value of the second variable as the second axis by coloring each coordinate of the two-dimensional grid based on a predetermined condition regarding the magnitude of the predicted values ​​of the experimental results. Note that the distribution diagram generation module 2037 may perform coloring based on any condition as long as the magnitude relationship of the predicted values ​​of the experimental results can be understood. For example, the distribution diagram generation module 2037 may divide the range of the predicted values ​​of the normalized experimental results at a predetermined interval and assign a different color to each divided range. For example, the two-dimensional grid may be colored with a warmer color for a larger predicted value of the experimental results and a cooler color for a smaller predicted value of the experimental results. Furthermore, the distribution map generating module 2037 may add contour lines to the distribution map by highlighting the boundaries of each divided area.

[0090] In addition, in step S205, the distribution map generation module 2037 may plot the following additional information on the distribution map.

[0091] (Data points for experimental conditions) The distribution diagram generation module 2037 may plot data points indicating the values ​​of the first variable and the second variable corresponding to the values ​​of the experimental condition data on the distribution diagram. Specifically, for example, the distribution diagram generation module 2037 searches the model ID item of the model table 2023 based on the model ID acquired in step S201, and acquires a table ID. The distribution diagram generation module 2037 searches the storage unit 202 based on the table ID, and acquires a data set of experimental data used in learning the prediction model identified by the model ID. The dimensionality reduction module 2034 converts the values ​​of the experimental condition data of each record of the experimental data table 2022 into the corresponding values ​​of the first variable and the second variable, respectively, in the same manner as in the process of step S202. The distribution diagram generation module 2037 plots data points on the generated distribution diagram at coordinates corresponding to the converted values ​​of the first variable and the second variable. This allows the conditions under which the user performed the experiment to be visualized on the distribution diagram, making it easier for the user to understand the predicted values ​​of the experimental results.

[0092] (Assigning identifiers to data points) The distribution diagram generating module 2037 may display a predetermined identifier associated with the value of the experimental condition data in association with the data point drawn on the distribution diagram. Specifically, for example, the distribution diagram generating module 2037 may draw an object indicating a value based on the data ID of the experimental data table 2022, on the data point of the experimental condition data associated with the data ID, in a manner that allows the user to recognize that the object is associated with the data point. In this way, for example, when the experimental data is stored in the experimental data table 2022 in the order in which the experiments were performed, the user can check the history of the experiments by checking the numbers assigned to the data points in comparison with the distribution diagram. Note that, for example, when information such as the order in which the experiments were performed, the date and time of the experiments, and an identification number appropriately assigned by the user to the experimental data is stored in the storage unit 202 or the like in association with the experimental condition data, the distribution diagram generating module 2037 may draw an object indicating the information associated with the experimental condition data indicated by the data point, in a manner that allows the user to recognize that the object is associated with the data point.

[0093] In addition, the distribution diagram generation module 2037 may draw a line object connecting data points having adjacent identifier values ​​on the distribution diagram, which allows the user to check the experiment history more easily.

[0094] The control unit 203 presents the distribution diagram to the user by displaying information on the distribution diagram of predicted values ​​of the experimental results on the terminal device 10 via the presentation control module 2039. Specifically, the presentation control module 2039 causes the transmission control module 2032 to transmit information on the distribution diagram of predicted values ​​of the experimental results to the terminal device 10.

[0095] In step S206, the terminal device 10 presents to the user a distribution map of predicted values ​​of the experimental results by the presentation control unit 193. Specifically, for example, the presentation control unit 193 presents information on the distribution map of predicted values ​​of the experimental results to the user via the display 141 based on the information received from the server 20.

[0096] Fig. 9 is a schematic diagram showing an example of a screen displaying a distribution diagram of predicted values ​​of experimental results and an enlarged view thereof. The presentation control unit 193 displays a distribution diagram 1411 of predicted values ​​of experimental results for each type of experimental result. In the example of Fig. 9, the presentation control unit 193 displays a distribution diagram 1411a of a first type of experimental result Y1 and a distribution diagram 1411b of a second type of experimental result Y2.

[0097] In the distribution diagram 1411, the horizontal axis indicates the value of the first variable, and the vertical axis indicates the value of the second variable. In the distribution diagram 1411, the two-dimensional grid objects generated in step S203 are colored based on a predetermined criterion regarding the magnitude of the predicted value of the experimental result. Note that while FIG. 9 shows a grayscale image, in reality, the two-dimensional grid objects in the distribution diagrams 1411a and 1411b are colored in a darker red for larger predicted values ​​of the experimental result, and in a darker blue for smaller predicted values ​​of the experimental result. Note that the same is true for FIGS. 11 and 12, which will be described later. This allows the user to recognize the magnitude of the predicted values ​​of the experimental result.

[0098] The display control unit 193 displays the distribution diagram 1411a of the experimental result Y1 and the distribution diagram 1411b of the experimental result Y2 in adjacent positions. This makes it easier to check the relationship between the predicted values ​​of the respective experimental results. For example, the user can more easily understand the trade-off relationship between the experimental result Y1 and the experimental result Y2, and can more easily consider the experimental conditions after understanding the trade-off relationship.

[0099] As shown in the enlarged view of Figure 9, the distribution plot 1411 includes data point objects 14111 (14111a, 14111b, 14111c, 14111d, ...) indicating data points for the experimental conditions under which the user conducted the experiment, number objects 14112 (14112a, 14112b, 14112c, 14112d, ...) indicating identification numbers for the experimental conditions under which the user conducted the experiment, and line objects 14113 indicating lines connecting the data points.

[0100] The data point object 14111 visualizes the experimental conditions performed by the user on the distribution diagram 1411. This allows the user to understand the experiment that he or she actually performed by relating it to the distribution diagram 1411. Therefore, the content of the distribution diagram 1411 can be more easily understood, and the predicted experimental results shown by the distribution diagram 1411 can be more convincing. Furthermore, the terminal device 10 may accept the selection of a data point by the user, and display on the display 141 the values ​​of the experimental conditions and the experimental results associated with the data point.

[0101] The number object 14112 visualizes, for example, the identification number of an experiment conducted by the user on the distribution diagram 1411. The identification number indicates, for example, the order in which the experiment was conducted by the user. This makes it easy to understand the user's experiment history (how the user changed the experimental conditions and conducted the experiment) on the distribution diagram 1411, and allows the user to understand the distribution diagram 1411 with a sense of satisfaction.

[0102] Furthermore, when the distance between data point objects with adjacent numbers on the distribution diagram 1411 is large, for example when the distance between the 11th data point object and the 12th data point object is large, it is suggested that there is room for further consideration of experimental conditions between the 11th and 12th experimental conditions, which can promote the user's consideration of experimental conditions.

[0103] The line object 14113 indicates a line connecting data points adjacent to the identification numbers indicated by the number object 14112. This allows the user to more easily check the history of an experiment.

[0104] (Optimization process) FIG. 10 is a diagram for explaining an example of the operations of the terminal device 10 and the server 20 when optimizing a prediction model based on a user request and presenting the optimization result to the user.

[0105] First, the user operates the terminal device 10 to access the server 20. The control unit 203 transmits information for presenting a screen for accepting an optimization request to the terminal device 10. Furthermore, for example, the control unit 203 searches the model table 2023 based on the user ID of the user, and acquires information on the prediction model associated with the user. The control unit 203 transmits the information on the prediction model associated with the user to the terminal device 10.

[0106] In step S301, the terminal device 10 accepts input of information on settings for performing optimization from the user. Specifically, for example, the terminal device 10 accepts selection of a prediction model to be optimized. The terminal device 10 also accepts information on constraint conditions for performing optimization using the prediction model. For example, the user inputs a search range of the experimental conditions (explanatory variables) and a range of values ​​that the experimental results (objective variables) should satisfy. In addition, for example, the terminal device 10 accepts information on goals for performing optimization using the prediction model (setting of maximum / minimum values ​​in the search range of a predetermined explanatory variable, maximization / minimization of a predetermined objective variable, etc.) from the user. The terminal device 10 transmits information on the optimization settings to the server 20.

[0107] Note that step S301 may be executed following step S103 of the modeling process described above. In this case, the prediction model generated in step S103 is the target for optimization.

[0108] The server 20 receives, via the reception control module 2031, information on the prediction model to be optimized, and information on constraint conditions related to at least one of the values ​​of the explanatory variables and the values ​​of the objective variables.

[0109] In step S302, the control unit 203 executes optimization of the prediction model by the optimization module 2038. Specifically, the optimization module 2038 executes optimization of the prediction model based on the constraint conditions received from the terminal device 10 by a predetermined optimization method. For example, the optimization module 2038 executes inference on the values ​​of the explanatory variables within a range that satisfies the constraint conditions, and obtains a result according to the goal of optimization of the objective variable. The optimization module 2038 obtains the optimization results of the explanatory variables and the optimization results of the objective variable obtained by the optimization.

[0110] In step S303, the control unit 203 compresses the dimensions of the explanatory variables of the prediction model to two dimensions using the dimension compression module 2034, and obtains the values ​​of the first and second variables corresponding to the optimization results obtained in step S302. The details of this process are the same as those in step S202, so duplicated explanations will be omitted.

[0111] In step S304, the control unit 203 plots data points indicating the values ​​of the first variable and the second variable obtained by the optimization on a distribution map using the distribution map generation module 2037. Specifically, for example, the control unit 203 generates a distribution map of predicted values ​​of the experimental results in the same manner as in steps S202 to S205 of the distribution map generation process. The distribution map generation module 2037 plots a plot object indicating the optimization result on the generated distribution map at the coordinates corresponding to the values ​​of the first variable and the second variable identified in step S303.

[0112] The control unit 203 presents the optimization results to the user by causing the presentation control module 2039 to display information on the optimization results on the terminal device 10. Specifically, the presentation control module 2039 causes the transmission control module 2032 to transmit, to the terminal device 10, information on a distribution diagram of predicted values ​​of the experiment results, including a plot showing the optimization results.

[0113] In step S305, the terminal device 10 presents the optimization result to the user. Specifically, for example, the terminal device 10 presents information on a distribution map of predicted values ​​of the experimental results, including a plot showing the optimization result, to the user via the display 141, based on the information received from the server 20.

[0114] 11 is a schematic diagram showing an example of a screen displaying the optimization results. The display control unit 193 displays a distribution diagram 1411 of predicted values ​​of the experimental results, and a window 1412 showing the values ​​of the optimized experimental conditions and the values ​​of the optimized experimental results.

[0115] The presentation control unit 193 displays optimization point objects 14114a and 14114b indicating the results optimized by the optimization module 2038 in the distribution diagram 1411a and the distribution diagram 1411b. This allows the user to visually understand the optimization results of the experimental conditions. In addition, when the optimization point object 14114 is displayed together with the data point object 14111 as shown in FIG. 9, the user can check the optimization results in comparison with the experiment that the user performed. This allows the user to check which experiment was closest to the optimization result, etc., which can further promote the consideration of the experimental conditions.

[0116] In addition, the optimization module 2038 of the server 20 may present a predetermined number of results from the top of the results predicted by the prediction model as optimization results in the optimization process. Then, the presentation control unit 193 of the terminal device 10 may receive the results and display an optimization point object 14114 indicating a predetermined number of optimization results from the top on the distribution diagram 1411. This makes it possible to present multiple candidates for experimental conditions to the user.

[0117] The window 1412 displays specific values ​​of the experimental conditions and experimental results indicated by the optimization point object 14114. This allows the optimized experimental conditions and specific values ​​of the experimental results to be easily confirmed.

[0118] As described above, in the above embodiment, the control unit 203 of the server 20 receives a data set including a plurality of combinations of experimental condition data related to three or more types of experimental conditions and experimental result data related to one or more types of experimental results through the reception control module 2031. The learning module 2033 performs machine learning based on the data set to generate a prediction model in which three or more types of experimental conditions are set as three or more explanatory variables and the experimental result is set as the objective variable. The dimensionality reduction module 2034 reduces the dimension of the explanatory variables to two dimensions consisting of a first variable and a second variable. The inference module 2036 infers the value of the objective variable by using the value of the explanatory variable restored by substituting a predetermined value for each of the first variable and the second variable and the prediction model. The presentation control module 2039 presents a distribution diagram of the value of the objective variable inferred in the inference step on a plane with the value of the first variable as the first axis and the value of the second variable as the second axis in a manner that allows the magnitude of the value of the objective variable inferred in the inference step to be recognized. This makes it possible to visualize the predicted results of an experimental result prediction model in which three or more experimental conditions are used as explanatory variables in a two-dimensional distribution chart.

[0119] Therefore, according to the program, server 20, and system 1 of this embodiment, when predicting experimental results using a prediction model, it is possible to further facilitate understanding of the relationship between multiple experimental conditions and the predicted experimental results.

[0120] In the above embodiment, the presentation control module 2039 presents data points on a plane that indicate the values ​​of the first and second variables corresponding to the values ​​of the experimental condition data, which allows the user to more easily understand the predicted values ​​of the experimental results.

[0121] In the above embodiment, the presentation control module 2039 displays a predetermined identifier associated with the value of the experimental condition data in association with the data point, which allows the user to more easily understand the predicted value of the experimental result.

[0122] In the above embodiment, the experimental condition data is arranged in a predetermined order. The presentation control module 2039 then displays an object showing a line connecting two data points adjacent to each other in the experimental condition data order. This allows the user to easily understand the relationship between the data points displayed in the distribution diagram.

[0123] In the above embodiment, the reception control module 2031 acquires constraint conditions including conditions related to at least one of the explanatory variable values ​​and the target variable values. The optimization module 2038 optimizes the target variable based on the constraint conditions. The dimensionality reduction module 2034 acquires the first variable value and the second variable value corresponding to the optimized explanatory variable value. The presentation control module 2039 presents data points indicating the acquired first variable value and second variable value on a plane. This allows the user to understand the optimization result in association with the distribution diagram.

[0124] In the above embodiment, the optimization module 2038 maximizes or minimizes the value of the objective variable within the range of the constraint conditions, allowing the user to grasp the predicted value of the maximized or minimized experimental result.

[0125] In the above embodiment, the constraint conditions include at least one of the conditions that the value of the objective variable is equal to or greater than a first specific value, is greater than the first specific value, is equal to or less than a second specific value, and is smaller than the second specific value, allowing the user to obtain optimization results according to desired conditions.

[0126] In the above embodiment, the experimental result data includes a plurality of types of experimental results. The learning module 2033 generates a plurality of prediction models for each type of experimental result. The reception control module 2031 acquires constraint conditions related to at least one of the plurality of prediction models. The optimization module 2038 optimizes at least one of the explanatory variable values ​​and the objective variable values ​​for each of the plurality of prediction models based on the constraint conditions related to at least one of the plurality of prediction models. This allows the user to consider the experimental conditions while understanding the trade-off relationship between the plurality of objective variables.

[0127] <Modification> The control unit 203 may accept a selection of a predetermined point on the distribution map from the user and present the values ​​of the experimental conditions and the predicted values ​​of the experimental results at the accepted point to the user, for example, after step S206 of the distribution map generation process and step S305 of the optimization process in the above embodiment. An example of the operation of the terminal device 10 and the server 20 when accepting a selection of a point on the distribution map from the user and presenting the experimental conditions and the predicted values ​​of the experimental results at the point to the user will be described below.

[0128] As a premise, a distribution map of predicted values ​​of the experimental results as presented to the user in steps S206 and S305 is displayed on the display 141 of the terminal device 10.

[0129] The terminal device 10 receives a selection of a predetermined point on the distribution map from the user. Specifically, for example, the user operates the terminal device 10 to select a predetermined point on the distribution map. The terminal device 10 acquires the coordinates (first coordinate, second coordinate) of the selected point. The terminal device 10 transmits the value of the received coordinates to the server 20.

[0130] The control unit 203 receives information indicating that a predetermined point on the distribution map has been selected through the reception control module 2031. The control unit 203 performs inverse transformation through the inverse transformation module 2035, using the value of the first coordinate received from the terminal device 10 as the value of the first variable and the value of the second coordinate as the value of the second variable. This allows the value of the explanatory variable (value of the experimental condition) corresponding to the coordinates of the received point to be acquired.

[0131] The control unit 203 inputs the values ​​of the identified explanatory variables to the prediction model by the inference module 2036, and infers the objective variable. As a result, the value of the objective variable (predicted value of the experimental result) corresponding to the coordinates of the point received from the user is acquired.

[0132] The control unit 203 presents the acquired values ​​of the experimental conditions and the predicted values ​​of the experimental results to the user via the presentation control module 2039. Specifically, for example, the presentation control module 2039 transmits, via the transmission control module 2032, information for presenting the values ​​of the experimental conditions and the predicted values ​​of the experimental results to the user to the terminal device 10.

[0133] The terminal device 10 presents the accepted values ​​of the experiment conditions and the predicted values ​​of the experiment results to the user via the display 141 or the like.

[0134] In addition, in steps S203 to S204 of the distribution map generation process, the control unit 203 may associate the explanatory variable values ​​corresponding to each point of the two-dimensional grid with the inferred objective variable values ​​and store them in the storage unit 202. Then, in this modified example, when the control unit 203 receives a selection of a specific point on the distribution map from the user, the control unit 203 may identify, from among the points of the two-dimensional grid, a point that is closest to the point received from the user. Then, the control unit 203 may present to the user the explanatory variable values ​​and objective variable values ​​corresponding to the identified point in the two-dimensional grid as values ​​of the experimental conditions and predicted values ​​of the experimental results.

[0135] 12 is a schematic diagram showing an example of a screen that presents the values ​​of the experimental conditions and the predicted values ​​of the experimental results at the points received from the user to the user. The presentation control unit 193 displays a distribution chart 1411 of the predicted values ​​of the experimental results and a window 1414 showing the values ​​of the experimental conditions and the predicted values ​​of the experimental results at the points received from the user.

[0136] When the terminal device 10 receives a selection of a predetermined point from the user on either the distribution map 1411a or the distribution map 1411b, the display control unit 193 displays a selected point object 14115 (14115a, 14115b) indicating the point selected by the user. The display control unit 193 also displays the values ​​of the experimental conditions and the predicted values ​​of the experimental results at the point received from the user in a window 1414. This allows the user to easily check the experimental conditions and experimental results at the point in the distribution map that interests the user. This can further promote the user's consideration of the experimental conditions.

[0137] In the above embodiment, the system 1 compresses the dimension of the explanatory variables of the prediction model into two dimensions represented by the first variable and the second variable. The system 1 restores the values ​​of the explanatory variables by substituting predetermined values ​​for the first variable and the second variable, respectively. The system 1 infers the value of the objective variable by using the restored explanatory variable values ​​and the prediction model. The system 1 presents a distribution diagram of the inferred objective variable values ​​to the user on a plane with the value of the first variable as the first axis and the value of the second variable as the second axis in a manner that allows the user to recognize the magnitude of the inferred objective variable value. However, the system 1 may compress the dimension of the explanatory variables of the prediction model into three dimensions represented by the first variable, the second variable, and the third variable. The system 1 may restore the values ​​of the explanatory variables by substituting predetermined values ​​for the first variable, the second variable, and the third variable, respectively. The system 1 may infer the value of the objective variable by using the restored explanatory variable values ​​and the prediction model. In addition, system 1 may present to the user a distribution diagram of the values ​​of the inferred objective variables in a space with the value of the first variable as the first axis, the value of the second variable as the second axis, and the value of the third variable as the third axis, in a manner that allows the user to recognize the magnitude of the values ​​of the inferred objective variables.

[0138] At this time, the dimensionality reduction module 2034 may reduce the dimension of the explanatory variables of the prediction model to three dimensions consisting of a first variable, a second variable, and a third variable. The inverse transformation module 2035 may inversely transform a function including the first variable, the second variable, and the third variable to restore the dimension before compression. The distribution diagram generation module 2037 may generate a distribution diagram of the predicted values ​​of the experimental results in a three-dimensional space in which the value of the first variable is the first axis, the value of the second variable is the second axis, and the value of the third variable is the third axis, in a manner in which the magnitude of the predicted values ​​of the experimental results can be recognized.

[0139] <4 Basic hardware configuration of computer> 13 is a block diagram showing the basic hardware configuration of a computer 90. The computer 90 includes at least a processor 91, a main storage device 92, an auxiliary storage device 93, and a communication IF (interface) 99. These are electrically connected to each other by a bus.

[0140] The processor 91 is hardware for executing an instruction set described in a program, and is composed of an arithmetic unit, a register, a peripheral circuit, and the like.

[0141] The main storage device 92 is for temporarily storing programs, data to be processed by the programs, etc. For example, it is a volatile memory such as a DRAM (Dynamic Random Access Memory).

[0142] The auxiliary storage device 93 is a storage device for saving data and programs, such as a flash memory, a hard disk drive (HDD), a magneto-optical disk, a CD-ROM, a DVD-ROM, or a semiconductor memory.

[0143] The communication IF 99 is an interface for inputting and outputting signals for communicating with other computers via a network using a wired or wireless communication standard. The network is composed of the Internet, a LAN, various mobile communication systems constructed by wireless base stations, etc. For example, the network includes 3G, 4G, 5G mobile communication systems, LTE (Long Term Evolution), wireless networks that can connect to the Internet via a specified access point (e.g., Wi-Fi (registered trademark)), etc. In the case of wireless connection, communication protocols include, for example, Z-Wave (registered trademark), ZigBee (registered trademark), Bluetooth (registered trademark), etc. In the case of wired connection, the network also includes a network that is directly connected by a USB (Universal Serial Bus) cable or the like.

[0144] It should be noted that the computer 90 can be virtually realized by distributing all or part of each hardware configuration among multiple computers 90 and connecting them together via a network. In this way, the computer 90 is a concept that includes not only a computer 90 housed in a single housing or case, but also a virtualized computer system.

[0145] Although several embodiments of the present disclosure have been described above, these embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and modifications are within the scope of the invention and its equivalents as described in the claims, as well as the scope and spirit of the invention.

[0146] <Additional Notes> The matters described in the above embodiment will be supplemented below. (Appendix 1) 1. A program for operating a computer having a processor and a memory, the program executing the following steps: accepting a dataset including multiple combinations of experimental condition data related to three or more types of experimental conditions and experimental result data related to one or more types of experimental results; generating a prediction model using machine learning based on the dataset to set the three or more types of experimental conditions as three or more explanatory variables and the experimental result as an objective variable; compressing the dimensions of the explanatory variables into two dimensions consisting of a first variable and a second variable; inferring a value of the objective variable using the values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable and the second variable, respectively, and the prediction model; and presenting a distribution diagram of the values ​​of the objective variable inferred in the inferring step on a plane having the value of the first variable as a first axis and the value of the second variable as a second axis, in a manner that enables the magnitude of the value of the objective variable inferred in the inferring step to be recognized. (Appendix 2) 2. The program of claim 1, wherein in the presenting step, data points indicating values ​​of a first variable and a second variable corresponding to values ​​of the experimental condition data are presented on the plane. (Appendix 3) The program of claim 2, wherein in the presenting step, a predetermined identifier associated with the value of the experimental condition data is displayed in association with the data point. (Appendix 4) The program of claim 3, wherein the experimental condition data is in a predetermined order, and in the presenting step, an object is displayed indicating a line connecting two adjacent data points in the order of the experimental condition data. (Appendix 5) 2. The program according to claim 1, which executes the steps of: acquiring constraint conditions including conditions related to at least one of the values ​​of the explanatory variables and the value of the objective variable; optimizing the objective variable based on the constraint conditions; acquiring values ​​of the first variable and the second variable corresponding to the values ​​of the explanatory variables acquired in the optimizing step; and presenting data points indicating the values ​​of the first variable and the second variable acquired in the step of acquiring the values ​​of the first variable and the second variable on the plane. (Appendix 6) 6. The program according to claim 5, wherein in the optimizing step, the value of the objective variable is maximized or minimized within the range of the constraint conditions. (Appendix 7) The program of claim 5, wherein the constraint conditions include at least one of the conditions that the value of the objective variable is greater than or equal to a first specific value, greater than a first specific value, less than or equal to a second specific value, and less than a second specific value. (Appendix 8) The program of claim 5, wherein the experimental result data includes a plurality of types of experimental results, and wherein in the generating step of the predictive model, a plurality of the predictive models are generated for each type of the experimental result, and in the acquiring step, the constraint condition related to at least any one of the plurality of the predictive models is acquired, and in the optimizing step, at least any one of the values ​​of the explanatory variables and the values ​​of the target variables is optimized for each of the plurality of the predictive models based on the constraint condition related to at least any one of the plurality of the predictive models. (Appendix 9) 2. The program according to claim 1, further comprising: a step of receiving a selection of a predetermined point on the distribution map; a step of acquiring a value of the objective variable corresponding to the predetermined point; and a step of presenting to a user the value of the objective variable corresponding to the predetermined point. (Appendix 10) 1. A program for operating a computer having a processor and a memory, the program executing the following steps: accepting a dataset including a plurality of combinations of experimental condition data related to four or more types of experimental conditions and experimental result data related to one or more types of experimental results; generating a prediction model using machine learning based on the dataset to set the four or more types of experimental conditions as four or more explanatory variables and the experimental result as an objective variable; compressing the dimensions of the explanatory variables into three dimensions consisting of a first variable, a second variable, and a third variable; inferring a value of the objective variable using values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable, the second variable, and the third variable, respectively, and the predictive model; and presenting a distribution diagram of the values ​​of the objective variable inferred in the inferring step in a space having a first axis for the value of the first variable, a second axis for the value of the second variable, and a third axis for the value of the third variable, in a manner that enables the magnitude of the value of the objective variable inferred in the inferring step to be recognized. (Appendix 11) A method executed by a computer having a processor and a memory, the method comprising the steps of: receiving a dataset including a plurality of combinations of experimental condition data relating to three or more types of experimental conditions and experimental result data relating to one or more types of experimental results, performing machine learning based on the dataset to generate a prediction model in which the three or more types of experimental conditions are three or more explanatory variables and the experimental result is an objective variable, compressing the dimensions of the explanatory variables into two dimensions consisting of a first variable and a second variable, inferring a value of the objective variable using the values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable and the second variable, respectively, and the prediction model, and presenting a distribution diagram of the values ​​of the objective variable inferred in the inferring step on a plane having the value of the first variable as a first axis and the value of the second variable as a second axis, in a manner that enables the magnitude of the value of the objective variable inferred in the inferring step to be recognized. (Appendix 12) 1. An information processing device comprising a control unit and a storage unit, wherein the control unit executes the following steps: accepting a dataset including a plurality of combinations of experimental condition data relating to three or more types of experimental conditions and experimental result data relating to one or more types of experimental results; generating a prediction model in which the three or more types of experimental conditions are three or more explanatory variables and the experimental result is an objective variable by performing machine learning based on the dataset; compressing the dimensions of the explanatory variables into two dimensions consisting of a first variable and a second variable; inferring a value of the objective variable using the values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable and the second variable, respectively, and the prediction model; and presenting a distribution diagram of the values ​​of the objective variable inferred in the inferring step on a plane having the value of the first variable as a first axis and the value of the second variable as a second axis, in a manner in which the magnitude of the value of the objective variable inferred in the inferring step can be recognized. (Appendix 13) a step of receiving a dataset including a plurality of combinations of experimental condition data relating to three or more types of experimental conditions and experimental result data relating to one or more types of experimental results; a step of generating a prediction model in which the three or more types of experimental conditions are three or more explanatory variables and the experimental result is an objective variable by performing machine learning based on the dataset; a step of compressing the dimensionality of the explanatory variables into two dimensions consisting of a first variable and a second variable; a step of inferring a value of the objective variable using the values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable and the second variable, respectively, and the predictive model; and a step of presenting a distribution diagram of the values ​​of the objective variable inferred in the inferring step on a plane having the value of the first variable as a first axis and the value of the second variable as a second axis, in a manner in which the magnitude of the value of the objective variable inferred in the inferring step can be recognized. (Appendix 14) 11. The program of claim 10, wherein in the presenting step, data points indicating values ​​of a first variable, a second variable, and a third variable corresponding to values ​​of the experimental condition data are presented in the space. (Appendix 15) 15. The program of claim 14, wherein in the presenting step, a predetermined identifier associated with the value of the experimental condition data is displayed in association with the data point. (Appendix 16) The program described in Appendix 15, wherein the experimental condition data is in a predetermined order, and in the presenting step, an object is displayed indicating a line connecting two adjacent data points in the order of the experimental condition data. (Appendix 17) 11. The program according to claim 10, which executes the steps of: acquiring constraint conditions including conditions related to at least one of the values ​​of the explanatory variables and the value of the objective variable; optimizing the objective variable based on the constraint conditions; acquiring values ​​of the first variable, the second variable, and the third variable corresponding to the values ​​of the explanatory variables acquired in the optimizing step; and presenting, on the space, data points indicating the values ​​of the first variable, the second variable, and the third variable acquired in the steps of acquiring the values ​​of the first variable, the second variable, and the third variable. (Appendix 18) 18. The program according to claim 17, wherein in the optimizing step, the value of the objective variable is maximized or minimized within the range of the constraint conditions. (Appendix 19) The program of claim 17, wherein the constraint conditions include at least one of the conditions that the value of the objective variable is greater than or equal to a first specific value, greater than a first specific value, less than or equal to a second specific value, and less than a second specific value. (Appendix 20) The program according to claim 17, wherein the experimental result data includes a plurality of types of experimental results, and wherein in the generating step of the predictive model, a plurality of the predictive models are generated for each type of the experimental result, and in the acquiring step, the constraint condition related to at least any one of the plurality of the predictive models is acquired, and in the optimizing step, at least any one of the values ​​of the explanatory variables and the values ​​of the target variables is optimized for each of the plurality of the predictive models based on the constraint condition related to at least any one of the plurality of the predictive models. (Appendix 21) 11. The program according to claim 10, further comprising: a step of accepting a selection of a predetermined point on the distribution map; a step of acquiring a value of the objective variable corresponding to the predetermined point; and a step of presenting to a user the value of the objective variable corresponding to the predetermined point. [Explanation of symbols]

[0147] 1: System 10: Terminal device 20: Server 203: Control section 2031: Receiving control module 2032: Transmission control module 2033: Learning Module 2034: Dimensionality reduction module 2035: Inverse conversion module 2036: Inference module 2037: Distribution map generation module 2038: Optimization module 2039: Presentation control module

Claims

1. A program for operating a computer having a processor and a memory, receiving a data set including a plurality of combinations of experimental condition data for three or more experimental conditions and experimental result data for one or more experimental results; generating a prediction model by performing machine learning based on the dataset, the prediction model having the three or more types of experimental conditions as three or more explanatory variables and the experimental results as objective variables; compressing the dimensionality of the explanatory variables into two dimensions consisting of a first variable and a second variable; inferring a value of the objective variable by using the values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable and the second variable, respectively, and the prediction model; presenting a distribution map of the values ​​of the objective variable inferred in the inferring step on a plane having the value of the first variable as a first axis and the value of the second variable as a second axis in such a manner that the magnitude of the value of the objective variable inferred in the inferring step can be recognized; A program to execute.

2. 2. The program according to claim 1, wherein in the presenting step, data points indicating values ​​of a first variable and a second variable corresponding to values ​​of the experimental condition data are presented on the plane.

3. 3. The program according to claim 2, wherein in said presenting step, a predetermined identifier associated with a value of said experimental condition data is displayed in association with said data point.

4. The experimental condition data is in a predetermined order, 4. The program according to claim 3, wherein in said presenting step, an object showing a line connecting two of said data points adjacent in order of said experimental condition data is displayed.

5. obtaining a constraint condition including a condition regarding at least one of the values ​​of the explanatory variables and the values ​​of the objective variables; optimizing the objective variable based on the constraint condition; acquiring values ​​of the first variable and the second variable corresponding to the values ​​of the explanatory variables acquired in the optimizing step; presenting data points on the plane indicative of the values ​​of the first variable and the values ​​of the second variable obtained in the step of obtaining the values ​​of the first variable and the values ​​of the second variable; The program according to claim 1 .

6. 6. The program according to claim 5, wherein in the optimizing step, the value of the objective variable is maximized or minimized within the range of the constraint conditions.

7. The constraint is:

6. The program according to claim 5, further comprising at least one of the conditions that the value of the objective variable is equal to or greater than a first specific value, is greater than the first specific value, is equal to or less than a second specific value, and is smaller than the second specific value.

8. The experiment result data includes a plurality of types of experiment results, In the step of generating the prediction model, a plurality of the prediction models are generated for each type of the experimental result; In the step of acquiring the constraint condition, the constraint condition is acquired for at least one of the plurality of the prediction models; In the optimizing step, at least one of the explanatory variable value and the target variable value is optimized for each of the plurality of prediction models based on the constraint condition for at least one of the plurality of prediction models. The program according to claim 5.

9. accepting a selection of a given point in the distribution map; obtaining a value of the objective variable corresponding to the predetermined point; presenting to a user a value of the objective variable corresponding to the predetermined point; The program according to claim 1 .

10. A program for operating a computer having a processor and a memory, receiving a data set including a plurality of combinations of experimental condition data relating to four or more experimental conditions and experimental result data relating to one or more experimental results; A step of generating a prediction model by performing machine learning based on the dataset, the prediction model having the four or more types of experimental conditions as four or more explanatory variables and the experimental results as objective variables; compressing the dimensions of the explanatory variables into three dimensions consisting of a first variable, a second variable, and a third variable; inferring a value of the objective variable using the values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable, the second variable, and the third variable, respectively, and the prediction model; presenting a distribution map of the values ​​of the objective variable inferred in the inferring step in a space having the value of the first variable as a first axis, the value of the second variable as a second axis, and the value of the third variable as a third axis, in a manner that allows the magnitude of the value of the objective variable inferred in the inferring step to be recognized; A program to execute.

11. 1. A computer-implemented method comprising a processor and a memory, the processor comprising: receiving a data set including a plurality of combinations of experimental condition data for three or more experimental conditions and experimental result data for one or more experimental results; generating a prediction model by performing machine learning based on the dataset, the prediction model having the three or more types of experimental conditions as three or more explanatory variables and the experimental results as objective variables; compressing the dimensionality of the explanatory variables into two dimensions consisting of a first variable and a second variable; inferring a value of the objective variable by using the values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable and the second variable, respectively, and the prediction model; presenting a distribution map of the values ​​of the objective variable inferred in the inferring step on a plane having the value of the first variable as a first axis and the value of the second variable as a second axis in such a manner that the magnitude of the value of the objective variable inferred in the inferring step can be recognized; How to do it.

12. An information processing device including a control unit and a storage unit, receiving a data set including a plurality of combinations of experimental condition data for three or more experimental conditions and experimental result data for one or more experimental results; generating a prediction model by performing machine learning based on the dataset, the prediction model having the three or more types of experimental conditions as three or more explanatory variables and the experimental results as objective variables; compressing the dimensionality of the explanatory variables into two dimensions consisting of a first variable and a second variable; inferring a value of the objective variable by using the values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable and the second variable, respectively, and the prediction model; presenting a distribution map of the values ​​of the objective variable inferred in the inferring step on a plane having the value of the first variable as a first axis and the value of the second variable as a second axis in such a manner that the magnitude of the value of the objective variable inferred in the inferring step can be recognized; An information processing device that executes the above.

13. A system including a computer having a processor and a memory, receiving a data set including a plurality of combinations of experimental condition data for three or more experimental conditions and experimental result data for one or more experimental results; generating a prediction model by performing machine learning based on the dataset, the prediction model having the three or more types of experimental conditions as three or more explanatory variables and the experimental results as objective variables; compressing the dimensionality of the explanatory variables into two dimensions consisting of a first variable and a second variable; inferring a value of the objective variable by using the values ​​of the explanatory variables restored by substituting predetermined values ​​for the first variable and the second variable, respectively, and the prediction model; presenting a distribution map of the values ​​of the objective variable inferred in the inferring step on a plane having the value of the first variable as a first axis and the value of the second variable as a second axis in such a manner that the magnitude of the value of the objective variable inferred in the inferring step can be recognized; A system that runs

Images