Method for training model, and computer system

Abstract

This method for training a model for predicting the effect of intervention on a person is performed by a computer system. The computer system includes: a step for presenting an interface that designates, among training data including a plurality of factors, the details of the intervention, and the effect of the intervention, covariates included in the plurality of factors; a step for acquiring the training data; and a step for training the model by using, from among the acquired training data, the details of the intervention, the effect of the intervention, and the plurality of factors corresponding to the covariates designated via the interface.

Description

Model learning method and computer system Incorporation by Reference

[0001] This application claims priority from Japanese Patent Application No. 2023-104116, filed on June 26, 2023, the contents of which are incorporated herein by reference.

[0002] The present invention relates to training models to predict the effects of interventions on people.

[0003] In various fields, such as medicine and marketing, it is necessary to predict the effects of interventions (such as the effects of medication and exercise) on people. Covariates (confounding factors) are important when predicting the effects of an intervention. Covariates are factors that affect the intervention effect (outcome) and are related to the underlying factors.

[0004] When a correlation is observed in observed data, it is necessary to distinguish whether the correlation is due to a causal relationship or the influence of a confounding factor. In response to this, a technique described in Patent Document 1 is known.

[0005] Patent Document 1 describes that "the method includes obtaining a machine learning model trained on a subject based on a plurality of factors corresponding to the subject; obtaining expert information on the subject provided by a human indicating relationships between the plurality of factors regarding how the plurality of factors influence each other; generating a structural causal model that represents the relationships between the plurality of factors based on the expert information; identifying, as confounding factors, factors among the plurality of factors that cause confounding bias in the machine learning model based on the structural causal model; and estimating the confounding bias based on the identified confounding factors."

[0006] Japanese Patent Application Laid-Open No. 2022-56367

[0007] First of all, when training a model to predict intervention effects using input data that includes covariates and other factors, there is a problem that the prediction accuracy is low. Therefore, it is necessary to train using only covariates.

[0008] The present invention provides a system and method for learning a model that has high accuracy in predicting the effects of an intervention.

[0009] A representative example of the invention disclosed in the present application is as follows: That is, a method for training a model for predicting an effect of an intervention on a person, executed by a computer system, the computer system including at least one computer having a processor and a storage device connected to the processor, the model training method including: a first step in which the computer system presents an interface for specifying covariates included in multiple factors, in training data including multiple factors, details of an intervention, and an effect of the intervention, the covariates being included in the multiple factors, a second step in which the computer system acquires the training data, and a third step in which the computer system trains the model using the multiple factors that correspond to the details of the intervention, the effect of the intervention, and the covariates specified via the interface, which are included in the acquired training data.

[0010] According to the present invention, it is possible to realize the learning of a model with high accuracy in predicting the effect of an intervention. Problems, configurations, and effects other than those described above will become clear from the following description of the embodiments.

[0011] FIG. 1 is a diagram illustrating an example of the configuration of a system according to a first embodiment. FIG. 2 is a diagram illustrating an example of the software configuration of a computer according to a first embodiment. FIG. 3 is a diagram illustrating an example of a GUI presented by the computer according to a first embodiment. FIG. 4 is a diagram illustrating an example of the functional configuration of a prediction unit according to a first embodiment. FIG. 5 is a diagram illustrating an example of the functional configuration of a learning unit according to a first embodiment. FIG. 6 is a diagram illustrating an example of a GUI presented by a computer according to a second embodiment. FIG. 7 is a diagram illustrating a conventional problem.

[0012] Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the present invention should not be construed as being limited to the description of the embodiments shown below. Those skilled in the art will readily understand that the specific configuration can be changed without departing from the concept or spirit of the present invention.

[0013] In the configuration of the invention described below, the same or similar configurations or functions are denoted by the same reference numerals, and redundant explanations will be omitted.

[0014] In this specification, the terms "first," "second," "third," etc. are used to identify components and do not necessarily limit the number or order.

[0015] To facilitate understanding of the invention, the position, size, shape, range, etc. of each component shown in the drawings etc. may not represent the actual position, size, shape, range, etc. Therefore, the present invention is not limited to the position, size, shape, range, etc. disclosed in the drawings etc.

[0016] FIG. 1 is a diagram illustrating an example of the configuration of a system according to a first embodiment.

[0017] The system is composed of a computer 100, an information terminal 110, and an external storage device 111. The computer 100, the information terminal 110, and the external storage device 111 are connected to one another via a network 109. The network 109 is, for example, a local area network (LAN) or a wide area network (WAN), and the connection method may be either wired or wireless.

[0018] The computer 100 executes a learning process to generate a model for predicting an intervention effect, and predicts the intervention effect on data (input data 220) using the model. The computer 100 includes a CPU 101, a main storage device 102, a secondary storage device 103, a network adapter 104, an input device 105, and an output device 106. The hardware elements are connected to each other via an internal bus 108.

[0019] The CPU 101 executes a program stored in the main memory device 102. The CPU 101 executes processing in accordance with the program, thereby operating as a functional unit (module) that realizes a specific function. In the following description, when a processing is described using a functional unit as the subject, this indicates that the CPU 101 is executing a program that realizes the functional unit.

[0020] The main memory device 102 is a dynamic random access memory (DRAM) that stores programs executed by the CPU 101 and data used by the programs. The main memory device 102 is also used as a work area.

[0021] The secondary storage device 103 is a hard disk drive (HDD) or a solid state drive (SSD), and stores data permanently. The programs and data stored in the main storage device 102 may be stored in the secondary storage device 103. In this case, the CPU 101 reads the programs and information from the secondary storage device 103 and loads them into the main storage device 102.

[0022] The network adapter 104 is an interface for connecting to an external device via a network 109 .

[0023] The input device 105 is a device such as a keyboard, a mouse, or a touch panel, and is used to input data to the computer 100 .

[0024] The output device 106 is a display, a printer, or the like, and is a device for outputting the processing results of the computer 100, etc.

[0025] The hardware configuration of the computer 100 is merely an example and is not limiting. For example, the computer 100 does not need to include the input device 105 and the output device 106.

[0026] The information terminal 110 is a terminal that performs various operations on the computer 100. For example, the information terminal 110 registers learning data, registers models, and inputs user data. The hardware configuration of the information terminal 110 is the same as that of the computer 100.

[0027] The external storage device 111 stores various types of information and is, for example, an external HDD or a storage system.

[0028] FIG. 2 is a diagram illustrating an example of the software configuration of the computer 100 according to the first embodiment.

[0029] The computer 100 includes a learning unit 200 and a prediction unit 201, as well as a training data DB 210 and a model DB 211. The training data DB 210 and the model DB 211 may be stored in the external storage device 111.

[0030] The learning data DB 210 is a database that stores learning data used in the learning process. The learning data includes, for example, factors, timestamps, intervention details, and intervention effects. The factors represent the condition and characteristics of the person receiving the intervention, such as age, gender, and blood pressure. The timestamps are the date and time when the intervention effect was measured or the date and time when the learning data was generated.

[0031] The model DB 211 is a database that stores information about a model that predicts an intervention effect.

[0032] The learning unit 200 executes a learning process using the learning data stored in the learning data DB 210 and the model stored in the model DB 211. The prediction unit 201 predicts an intervention effect for input data 220 using the model stored in the model DB 211, and outputs the predicted intervention result 221.

[0033] Next, a method for learning the model in the first embodiment will be described.

[0034] First, the learning unit 200 presents a GUI (Graphical User Interface) for receiving a covariate specification from the user. Fig. 3 is a diagram showing an example of the GUI presented by the computer 100 of the first embodiment.

[0035] The GUI 300 includes a selection table 301 and a decision button 302. The selection table 301 is a table for selecting a factor that is a covariate from among the factors, and includes a selection field and a variable field. The variable field stores the name of a variable that is included in the input data 220 and corresponds to the factor. The selection field is a field for inputting a check mark to select a covariate.

[0036] In FIG. 3, "gender," "blood pressure," "pulse rate," "number of steps," and "step distance" are specified as covariates.

[0037] The user specifies a covariate in the selection table 301 and presses the decision button 302. The learning unit 200 acquires information on the covariate and starts the learning process.

[0038] FIG. 4 is a flowchart illustrating an example of a learning process executed by the computer 100 according to the first embodiment.

[0039] The learning unit 200 acquires learning data from the learning data DB 210 (step S101).

[0040] The learning unit 200 selects factors to be input to the model from the factors included in the learning data based on the covariate information (step S102).

[0041] The learning unit 200 executes a model learning process using the intervention details, the factors selected in step S102, and the intervention effects included in the learning data (step S103). The present invention is not limited to a model learning method.

[0042] High prediction accuracy can be achieved by training the model using only the covariates included in the factors.

[0043] The learning unit 200 determines whether to end learning (step S104). For example, when processing has been completed for all learning data, the learning unit 200 determines to end learning. Furthermore, when the number of updates is greater than a threshold, the learning unit 200 determines to end learning. Furthermore, when the prediction accuracy of the intervention effect of the evaluation data is greater than a threshold, the learning unit 200 determines to end learning.

[0044] If it is determined not to end the learning, the learning unit 200 returns to step S101 and executes the same process. If it is determined to end the learning, the learning unit 200 ends the process.

[0045] FIG. 5 is a flowchart illustrating an example of the learning process executed by the computer 100 according to the first embodiment.

[0046] In this method, the model is a neural network consisting of multiple layers containing multiple neurons. The input layer of the neural network contains neurons corresponding to the factors and interventions contained in the training data.

[0047] The learning unit 200 acquires learning data from the learning data DB 210 (step S201).

[0048] The learning unit 200 transforms the model structure based on the covariate information (step S202). Specifically, the learning unit 200 masks connections from neurons in the input layer corresponding to factors other than the covariates to the next layer. In other words, the weights are set to 0.

[0049] The learning unit 200 executes a model learning process using the learning data (step S203).

[0050] Since factors other than covariates do not affect the calculations within the model, it is possible to generate a model with high predictive accuracy.

[0051] The learning unit 200 determines whether or not to end the learning (step S204).

[0052] If it is determined not to end the learning, the learning unit 200 returns to step S201 and executes the same process. If it is determined to end the learning, the learning unit 200 ends the process.

[0053] According to the present invention, it is possible to generate a model with high prediction accuracy. In addition, since the number of variables used for learning is reduced, the cost required for learning can also be reduced.

[0054] Here, the conventional problem will be explained using Figures 7A and 7B. Figures 7A and 7B are examples of the output of a model that predicts the degree of arteriosclerosis and the effects of various types of exercise based on blood pressure pulse waves. Figure 7A shows the output of a model trained by selecting only covariates, while Figure 7B shows the output of a model trained without selecting any particular factors. Figure 7A shows the result that exercise intervention improves the degree of arteriosclerosis. On the other hand, Figure 7B shows no correlation between the degree of arteriosclerosis and exercise, resulting in a prediction result that deviates from common sense. For this reason, it is important to use only covariates during learning.

[0055] In the first embodiment, it is assumed that the user has knowledge of covariates. However, it is often difficult to determine in advance which of multiple factors is a covariate. In the second embodiment, the computer 100 makes a prediction using a model trained based on the covariates selected by the user, and displays the prediction result. The user checks the prediction result and selects a covariate again. The second embodiment will be described below, focusing on the differences from the first embodiment.

[0056] The system configuration of the second embodiment is the same as that of the first embodiment. The software configuration of the computer 100 of the second embodiment is the same as that of the first embodiment.

[0057] In the second embodiment, the GUI presented by the computer 100 is partially different. Fig. 6 is a diagram showing an example of the GUI presented by the computer 100 of the second embodiment.

[0058] The GUI 300 includes a selection table 301, a decision button 302, a prediction button 303, and a prediction result field 304. The selection table 301 is a table for selecting a factor that is a covariate from among the factors, and includes a selection field and a variable field. The variable field stores the name of a variable that is included in the input data 220 and corresponds to the factor. The selection field is a field for inputting a check mark to select a covariate.

[0059] The user specifies a covariate in the selection table 301 and presses the decision button 302. The learning unit 200 acquires information about the covariate and starts the learning process. Note that the learning process in Example 2 is the same as that in Example 1. After learning is complete, the prediction button 303 is enabled. When the prediction button 303 is pressed, the prediction unit 201 makes a prediction using the model generated by the learning process and displays the prediction result in the prediction result field 304. At this time, the computer 100 manages the covariate, the model, and the prediction result in association with each other. This makes it possible to compare prediction results of models with different covariates. The user can refer to multiple prediction results and select the optimal covariate.

[0060] The present invention is not limited to the above-described embodiments, but includes various modifications. For example, the above-described embodiments are provided to explain the present invention in detail, and the present invention is not necessarily limited to those including all of the described configurations. Furthermore, some of the configurations of each embodiment can be added to, deleted from, or replaced with other configurations.

[0061] Furthermore, some or all of the above-described configurations, functions, processing units, processing means, etc. may be implemented in hardware, for example, by designing them as integrated circuits. The present invention can also be realized by software program code that implements the functions of the embodiments. In this case, a storage medium on which the program code is recorded is provided to a computer, and a processor included in the computer reads the program code stored in the storage medium. In this case, the program code itself read from the storage medium implements the functions of the above-described embodiments, and the program code itself and the storage medium on which it is stored constitute the present invention. Examples of storage media for providing such program code include flexible disks, CD-ROMs, DVD-ROMs, hard disks, solid-state drives (SSDs), optical disks, magneto-optical disks, CD-Rs, magnetic tape, non-volatile memory cards, and ROMs.

[0062] Additionally, program code for realizing the functionality described in the present invention can be implemented in a wide range of programming or scripting languages, such as, for example, Assembler, C / C++, Perl, Shell, PHP, Python, Java, and the like.

[0063] Furthermore, the program code of the software that realizes the functions of the embodiments may be distributed via a network and stored in a storage means such as a computer's hard disk or memory, or in a storage medium such as a CD-RW or CD-R, and the processor of the computer may read and execute the program code stored in the storage means or the storage medium.

[0064] In the above-described embodiment, the control lines and information lines are those that are considered necessary for the explanation, and not all control lines and information lines are necessarily shown in the product. All components may be interconnected.

Claims

1. A method for training a model for predicting the effect of an intervention on a human, executed by a computer system, the computer system including at least one computer having a processor and a storage device connected to the processor, the model training method comprising: a first step in which the computer system presents an interface for specifying covariates included in multiple factors in training data including multiple factors, content of the intervention, and effect of the intervention; a second step in which the computer system acquires the training data; and a third step in which the computer system trains the model using the multiple factors included in the acquired training data, which correspond to the content of the intervention, the effect of the intervention, and the covariates specified via the interface.

2. A method for training a model as described in claim 1, wherein the model is a neural network including an input layer composed of neurons corresponding to the plurality of factors and the content of the intervention, and the third step includes a step in which the computer system transforms the structure of the model by masking connections from the plurality of factors to the next layer other than covariates specified via the interface, and a step in which the computer system trains the model whose structure has been transformed using the training data.

3. A model training method as described in claim 1, characterized in that the third step includes: a step by the computer system selecting from the training data the content of the intervention, the effect of the intervention, and the plurality of factors corresponding to the covariates specified via the interface; and a step by the computer system inputting the content of the intervention and the selected plurality of factors from the training data into the model.

4. A computer system comprising at least one computer having a processor and a storage device connected to the processor, the computer system managing a model for predicting the effect of an intervention on a human, presenting an interface for specifying covariates contained in multiple factors in learning data including multiple factors, the content of the intervention, and the effect of the intervention, acquiring the learning data, and training the model using the multiple factors contained in the acquired learning data that correspond to the content of the intervention, the effect of the intervention, and the covariates specified via the interface.

5. A computer system as described in claim 4, wherein the model is a neural network including an input layer composed of neurons corresponding to the plurality of factors and the content of the intervention, the computer system transforms the structure of the model by masking connections from the plurality of factors to the next layer other than covariates specified via the interface, and the computer system trains the model whose structure has been transformed using the training data.

6. A computer system as described in claim 4, characterized in that the computer system selects from the learning data the content of the intervention, the effect of the intervention, and the plurality of factors corresponding to covariates specified via the interface, and inputs the content of the intervention and the selected plurality of factors from the learning data into the model.

Images