Computer System and Intervention Effect Prediction Method

Abstract

To provide a system and a method for predicting an intervention effect when performing multiple kinds of interventions to a subject continuously.SOLUTION: A computing machine 100 manages a first model for calculating an output value by using time-series data including a value regarding intervention to a person, a second model for calculating a feature amount by mapping an output value of the first model to a feature amount space, and a third model for outputting a prediction value of an intervention effect from a feature amount. The time-series data include multiple data series including time of intervention, multiple factors for expressing a human state, and a type and degree of intervention. The computing machine 100 calculates a prediction value of a continuous intervention effect corresponding to time-series data by using the first model, the second model, and the third model. The second model maps an output value of the first model to a feature amount space such that a difference in distribution in the feature amount space of multiple data strings used for machine learning becomes small.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to a system and method for predicting the effect of an intervention on a person.

Background Art

[0002] In various fields such as medicine and marketing, it is necessary to predict the effect of an intervention (such as the effect of medication and the effect of exercise) on a person. A confounding factor becomes important in predicting the intervention effect. A confounding factor is a factor that affects the intervention effect and is related to a factor. When a correlation is observed in the observed data, it is necessary to distinguish whether it is due to a causal relationship or due to the influence of a confounding factor.

[0003] As an adjustment of the confounding factor, a method called a randomized controlled trial is known. Since this method requires randomly selecting subjects, there is a problem that the burden on the subjects and the experimental cost are large. Therefore, it is desired to develop a technique for performing causal inference using existing data. In contrast, the technique described in Patent Document 1 is known.

[0004] Patent Document 1 describes "In order to appropriately verify the effect of care intervention, clustering is performed based on the attributes of a plurality of subjects according to similarity, and based on the result, further divided into an intervention group and a control group, and the intervention effect is evaluated by comparing the intervention group and the control group."

[0005] In recent years, there has been a demand for a technique for predicting the intervention effect when continuously performing a plurality of types of interventions on a subject. The technique described in Patent Document 1 does not correspond to time-series data. As a system for performing prediction using time-series data, the technique described in Patent Document 2 is known.

[0006] Patent Document 2 describes "In order to bring the health state of a subject closer to an ideal health state, the measured values and target values of the health state for the past several days are learned, and then the target value of the health state to be recommended next and the expected value of achieving the target are output and presented to the user."

Prior Art Documents

Patent Documents

[0007]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0008] However, the technology described in Patent Document 2 does not consider the influence of confounding factors (attributes such as gender and age, and past intervention results).

[0009] The present invention provides a system and method for predicting the intervention effect when continuously performing multiple types of interventions on a subject in consideration of the influence of confounding factors.

Means for Solving the Problems

[0010] A typical example of the invention disclosed in the present application is as follows. That is, a computer system for predicting the effects of multiple interventions on a person, comprising at least one computer having a processor and a storage device connected to the processor, a first model for calculating an output value using time-series data including values related to the interventions performed on the person, the above-mentioned a second model for calculating feature quantities by mapping the output value of the first model to a feature quantity space, and a third model for outputting a predicted value of the effect of the intervention on the person, a fourth model that identifies the type of the intervention performed on the person from the above-mentioned feature amounts, the predicted type of the intervention output by the fourth model, the type of the intervention included in the learning data, the predicted value of the effect of the intervention, and a loss function defined from the effect value of the intervention included in the learning data which manages, wherein the time-series data includes a plurality of data sequences including the time when the intervention was performed on the person, a plurality of factors representing the state of the person, and values indicating the type and degree of the intervention performed on the person, and the processor a process of receiving the learning data including a plurality of data series including the identification information of the person, the time when the intervention was performed on the person, the values of the plurality of factors of the person, the type and degree of the intervention received by the person, and the effect value of the intervention, a process of inputting the data series into the first model and inputting the output value output from the first model into the second model, a process of calculating a predicted value of the effect of the intervention by inputting the feature amounts output from the second model into the third model, a process of calculating the predicted type of the intervention by inputting the feature amounts output from the second model into the fourth model, a process of calculating the value of the loss function using the type of the intervention and the effect value of the intervention for each of the plurality of data series and the predicted type of the intervention and the predicted value of the effect of the intervention calculated from each of the plurality of data series, and a process of updating the second model, the third model, and the fourth model using the value of the loss function, to execute machine learning, and furtherA prediction process including: a process of calculating the output value by inputting the data sequence into the first model; a process of calculating the feature amount by inputting the output value into the second model; and a process of calculating a predicted value of the continuous intervention effect corresponding to the time series data by inputting the feature amount into the third model. execute 。

Advantages of the Invention

[0011] According to the present invention, it is possible to predict the intervention effect when continuously performing a plurality of types of interventions on a subject in consideration of the influence of confounding factors. Problems, configurations, and effects other than those described above will be clarified by the following description of the embodiments.

Brief Description of the Drawings

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Modes for Carrying Out the Invention

[0013] Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the present invention is not construed as being limited to the description of the embodiments shown below. It will be easily understood by those skilled in the art that the specific configuration can be changed without departing from the spirit or gist of the present invention.

[0014] In the configuration of the invention described below, the same or similar configurations or functions are denoted by the same reference numerals, and redundant descriptions are omitted.

[0015] The notations such as "first", "second", "third", etc. in this specification and the like are attached to identify components, and do not necessarily limit the number or order.

[0016] The positions, sizes, shapes, and ranges, etc. of the respective components shown in the drawings and the like may not represent the actual positions, sizes, shapes, and ranges, etc. in order to facilitate the understanding of the invention. Therefore, in the present invention, it is not limited to the positions, sizes, shapes, and ranges, etc. disclosed in the drawings and the like.

Embodiment

[0017] FIG. 1 is a diagram showing a configuration example of the system of Embodiment 1.

[0018] 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 each other via a network 109. The network 109 is, for example, a LAN (Local Area Network) and a WAN (Wide Area Network), etc., and the connection method may be either wired or wireless.

[0019] The computer 100 executes a learning process for generating a model for predicting an intervention effect, and predicts the intervention effect on user data (input data) using the model. The computer 100 includes a CPU 101, a main memory device 102, a secondary storage device 103, a network adapter 104, an input device 105, and an output device 106. Each hardware element is connected to each other via an internal bus 108.

[0020] The CPU 101 executes a program stored in the main memory device 102. By executing the process according to the program, the CPU 101 operates as a functional unit (module) that realizes a specific function. In the following description, when the process is described with the functional unit as the subject, it indicates that the CPU 101 is executing a program that realizes the functional unit.

[0021] The main memory device 102 is a DRAM (Dynamic Random Access Memory), and stores a program executed by the CPU 101 and data used by the program. The main memory device 102 is also used as a work area.

[0022] The secondary storage device 103 is an HDD (Hard Disk Drive), an SSD (Solid State Drive), etc., and permanently stores data. The program and data stored in the main memory device 102 may be stored in the secondary storage device 103. In this case, the CPU 101 reads the program and information from the secondary storage device 103 and loads them into the main memory device 102.

[0023] The network adapter 104 is an interface for connecting to an external device via the network 109.

[0024] The input device 105 is a keyboard, a mouse, a touch panel, etc., and is a device for inputting to the computer 100.

[0025] The output device 106 is a display, a printer, etc., and is a device for outputting the processing result of the computer 100.

[0026] Note that the hardware configuration of the computer 100 is an example and is not limited thereto. For example, the computer 100 may not have the input device 105 and the output device 106.

[0027] The information terminal 110 is a terminal that performs various operations on the computer 100. For example, the information terminal 110 performs registration of learning data, registration of models, input of user data, and the like. The hardware configuration of the information terminal 110 is the same as that of the computer 100.

[0028] The external storage device 111 stores various types of information. The external storage device 111 is, for example, an external HDD or a storage system.

[0029] FIG. 2 is a diagram showing an example of the software configuration of the computer 100 according to the first embodiment.

[0030] The computer 100 includes a learning unit 200 and a prediction unit 201, and also includes a learning data DB 210 and a model DB 211. Note that the learning data DB 210 and the model DB 211 may be stored in the external storage device 111.

[0031] The learning data DB 210 is a database that stores learning data used for learning processing. The learning data DB 210 will be described with reference to FIG. 3. The model DB 211 is a database that stores information on various models.

[0032] The learning unit 200 executes learning processing using the learning data stored in the learning data DB 210 and the models stored in the model DB 211. The prediction unit 201 predicts the intervention effect on the user data 220 using the models stored in the model DB 211 and outputs it as the predicted intervention result 221. Note that the learning data and the user data 220 in the present embodiment are time-series data.

[0033] FIG. 3 is a diagram showing an example of the learning data DB 210 according to the first embodiment.

[0034] The learning data DB 210 stores entries including an ID 301, a factor 302, a date and time 303, an intervention content 304, and an effect 305. One entry corresponds to one piece of learning data. Note that the fields included in the entry are not limited to those described above. It may not include any of the above-described fields, or may include other fields.

[0035] The ID 301 is a field for storing identification information that uniquely identifies the learning data. An identification number is stored in the ID 301 of this embodiment.

[0036] The factor 302 is a field for storing values of factors such as the state and characteristics of the person receiving the intervention. The factors are, for example, age, gender, height, and the like. In this embodiment, the types and numbers of factors included in the factor 302 are not limited.

[0037] The learning data of this embodiment is time-series data, and one piece of learning data includes a plurality of data sequences composed of the date and time 303, the intervention content 304, and the effect 305.

[0038] The date and time 303 is a field for storing the date and time. The date and time when the intervention effect was measured or the date and time when the data sequence was generated is stored in the date and time 303. Note that the present invention is not limited to the type of date and time stored in the date and time 303. Any date and time that can grasp the flow of the time series may be used.

[0039] The intervention content 304 is a group of fields for storing information indicating the content of the intervention performed on a person. The intervention content 304 includes a type and an amount as fields. The type is a field for storing the type of the intervention. Values indicating types such as medicine, treatment, and exercise are stored in the type. The amount is a field for storing a value indicating the degree of the intervention. For example, a value indicating the dosage of medicine or the exercise time is stored. In this embodiment, when no intervention is performed, 0 is stored in each of the type and the amount.

[0040] Effect 305 is a group of fields that stores values (effect prediction values) of indicators indicating the effect of an intervention. In this embodiment, the types and numbers of indicators included in Effect 305 are not limited.

[0041] FIG. 4 is a diagram showing an example of the functional configuration of the prediction unit 201 in the first embodiment.

[0042] The prediction unit 201 includes a time series data processing unit 401, an interference factor adjustment unit 402, and a predictor 403.

[0043] The time series data processing unit 401 calculates an output value using time series data. The time series data processing unit 401 is, for example, a Recurrent Neural Network (RNN). The RNN is a type of neural network and is characterized in that inputs and outputs occur at each time step. Note that the width of the time step can be arbitrarily set. The RNN can obtain an output considering the time series by using the output of the previous time step as a new input. However, the output of the RNN does not consider the influence of interference factors.

[0044] The time series data processing unit 401 in this embodiment receives, at time step t, the intervention content and factors at time step t and the effect prediction value at time step (t - 1) as inputs. When t = 0, the time series data processing unit 401 receives only the intervention content and factors at t = 0 as inputs.

[0045] In this specification, the intervention content at time step t of a person with identification information i is denoted as A t i the factor as X t i and the effect prediction value as Ŷ t i are defined. Also, a person with identification information i is described as person (i). Note that  and Ŷ correspond to A and Y with a hat symbol in mathematical formulas and drawings.

[0046] The time series data processing unit 401, at time step t, receives the intervention content A at time step t ti 、Factor X t i and the predicted value of the effect Y^ at time step (t-1) t-1 i are used to calculate the output value (feature). The output value in this embodiment is a vector.

[0047] The confounding factor adjustment unit 402 performs a process of reducing the influence of the confounding factor on the output value in order to realize a more accurate prediction of the intervention effect.

[0048] There are two types of confounding factors in this embodiment, namely the factor and the effect of the intervention performed in the past. Regarding the influence of the factor, for example, when young people often select Intervention 1 and elderly people often select Intervention 2, there is a bias in the age distribution between Intervention 1 and Intervention 2, so an accurate effect cannot be predicted. This is because it is impossible to distinguish whether the obtained effect is due to the intervention or the bias of the factor. Regarding the influence of the effect of the intervention performed in the past, for example, if side effects occur after Drug 1 is administered in the previous intervention, the selection probability of Drug 1 will be low. Therefore, there is a bias in the distribution of drug selection in the next intervention, which affects the effect prediction.

[0049] The confounding factor adjustment unit 402 performs a process of reducing the difference in distribution so that all interventions can be selected evenly, and generates a feature with a balanced distribution. Specifically, the confounding factor adjustment unit 402 calculates the feature by mapping the output value (vector) calculated by the time series data processing unit 401 to a feature space of an arbitrary dimension.

[0050] The predictor 403 calculates the predicted value of the intervention effect using the feature calculated by the confounding factor adjustment unit 402. The predictor 403 is, for example, a Neural Network or a linear regression model.

[0051] Figure 5 is a diagram showing an example of the functional configuration of the learning unit 200 of Example 1.

[0052] The learning unit 200 includes a time-series data processing unit 401, an interference factor adjustment unit 402, a predictor 403, a discriminator 501, a calculator 502, and a calculator 503. The time-series data processing unit 401, the interference factor adjustment unit 402, and the predictor 403 are the same as those included in the prediction unit 201. The learning unit 200 uses a method such as Adversarial Learning to train the interference factor adjustment unit 402, the predictor 403, and the discriminator 501.

[0053] The discriminator 501 receives, as input, the feature amount calculated by the interference factor adjustment unit 402, and predicts the intervention content A^ t+1 i to be performed on person (i) at the next time step (t + 1). The discriminator 501 is defined as a model such as a neural network.

[0054] The calculator 503 calculates an Imbalance loss that evaluates the error between the predicted intervention content A^ t+1 i and the actual intervention content A t+1 i . The Imbalance loss function for calculating the Imbalance loss is defined by Equation (1).

[0055]

Equation

[0056] Here, G g is a function representing the output of the interference factor adjustment unit 402, and G d is a function representing the output of the discriminator 501. n represents the number of fields (number of samples) of the factor 302. II represents an indicator function, κ represents a threshold value, ε represents an error tolerance, and N represents the number of samples within the range of (A t+1 j +ε) centered on κ.

[0057] To handle multiple types of interventions, it is necessary to ensure continuity. Therefore, the intervention content A t+1 i and the intervention content A t+1j When the difference is less than or equal to the threshold κ or greater than or equal to the threshold κ, different weights are multiplied by the intervention content prediction error to calculate the Imbalance loss. Here, the intervention content prediction error represents the logarithmic term in Equation (1).

[0058] The learning unit 200 performs learning of the discriminator 501 so as to increase the prediction accuracy, while performing learning of the interference factor adjustment unit 402 so that the discriminator 501 cannot identify it.

[0059] The calculator 502 calculates the Factual loss for evaluating the error between the effect prediction value Y^ t-1 i calculated by the predictor 403 and the actual intervention effect Y t-1 i The Factual loss function for calculating the Factual loss is defined by Equation (2).

[0060]

Number

[0061] Here, G y is a function representing the output of the predictor 403.

[0062] As shown in Equation (3), the learning unit 200 performs learning of each model so that the loss function defined from the sum of the Imbalance loss over all time steps and the sum of the Factual loss over all time steps is minimized. In this learning, the learning unit 200 updates the discriminator 501 so that the prediction accuracy of the intervention content based on the feature amount is increased, and updates the interference factor adjustment unit 402 so that the discriminator 501 cannot predict the intervention content.

[0063]

Number

[0064] Here, α is a parameter for adjusting the Factual loss and the Imbalance loss.

[0065] By learning using the loss function, the difference in the distribution of the feature amounts generated by the interference factor adjustment unit 402 can be reduced. That is, the influence of the interference factor can be reduced. Thereby, the intervention effect can be predicted accurately.

[0066] FIG. 6 is a flowchart for explaining an example of the learning process executed by the learning unit 200 of the first embodiment.

[0067] When the learning unit 200 receives a learning execution instruction via the information terminal 110 or the input device 105, the learning unit 200 executes a learning process.

[0068] The learning unit 200 acquires learning data from the learning data DB 210 (step S101). Here, it is assumed that a learning data set composed of a plurality of learning data is acquired.

[0069] Next, the learning unit 200 starts loop processing of the data series included in the learning data (step S102). The learning unit 200 selects the data series in chronological order and repeatedly executes the following processing.

[0070] The learning unit 200 calculates a feature amount using the data series (step S103). Specifically, the learning unit 200 inputs the intervention content and factors corresponding to the data series and the effect prediction value using the data series one time series earlier to the time series data processing unit 401, and also inputs the output value calculated by the time series data processing unit 401 to the interference factor adjustment unit 402. The learning unit 200 stores the feature amount in the work area in association with the time series.

[0071] The learning unit 200 inputs the feature amount to the discriminator 501, and the predicted intervention content  t+1 i output from the discriminator 501 and the intervention content A t+1 i of the data series one time series later, and calculates an Imbalance loss based on them (step S104). The learning unit 200 stores the Imbalance loss in the work area in association with the time series.

[0072] The learning unit 200 updates the discriminator 501 and the interference factor adjustment unit 402 based on the backpropagation method using the Imbalance loss function or the like, and updates the feature amount using the updated interference factor adjustment unit 402 (step S105).

[0073] The learning unit 200 inputs the updated feature amount to the predictor 403, and the predicted value of the intervention effect Y^ t i and the intervention effect Y of the data series t i Based on this, the Factual loss is calculated (step S106). The learning unit 200 stores the Factual loss in the work area in association with the time series.

[0074] The learning unit 200 determines whether the processing has been completed for all data series included in the learning data (step S107).

[0075] If the processing has not been completed for all data series included in the learning data, the learning unit 200 returns to step S102 and executes the same processing.

[0076] If the processing has been completed for all data series included in the learning data, the learning unit 200 calculates the value of the loss function shown in Equation (3) (step S108).

[0077] The learning unit 200 updates the interference factor adjustment unit 402, the predictor 403, and the discriminator 501 based on the value of the loss function (step S109).

[0078] The learning unit 200 determines whether to end the learning (step S110). For example, when the processing is completed for all the learning data in the learning dataset, the learning unit 200 determines to end the learning. Also, when the number of update times is greater than the threshold value, the learning unit 200 determines to end the learning. Further, when the prediction accuracy of the intervention effect of the user data 220 for evaluation is higher than the threshold value, the learning unit 200 determines to end the learning.

[0079] When it is determined not to end the learning, the learning unit 200 returns to step S101 and executes the same processing.

[0080] When it is determined to end the learning, the learning unit 200 ends the processing.

[0081] FIG. 7 is a flowchart for explaining an example of the prediction process executed by the prediction unit 201 of the first embodiment.

[0082] When the prediction unit 201 receives a prediction execution instruction including the user data 220 via the information terminal 110 or the input device 105, the prediction unit 201 executes a prediction process.

[0083] The prediction unit 201 acquires the models of the time series data processing unit 401, the interference factor adjustment unit 402, and the predictor 403 from the model DB 211 (step S201).

[0084] The prediction unit 201 starts loop processing of the data series included in the user data 220 (step S202). The prediction unit 201 selects the data series in time series order and repeatedly executes the following processing.

[0085] The prediction unit 201 calculates a feature amount using the data series (step S203). Specifically, the prediction unit 201 inputs the intervention content and factors corresponding to the data series and the effect prediction value using the data series one time series earlier to the time series data processing unit 401, and also inputs the output value calculated by the time series data processing unit 401 to the interference factor adjustment unit 402. The prediction unit 201 stores the feature amount in the work area in association with the time series.

[0086] The prediction unit 201 calculates an intervention effect prediction value by inputting the feature quantity into the predictor 403 (step S204).

[0087] The prediction unit 201 determines whether the processing has been completed for all data sequences included in the user data 220 (step S205).

[0088] If the processing has not been completed for all data sequences included in the user data 220, the prediction unit 201 returns to step S202 and executes the same processing.

[0089] If the processing has been completed for all data sequences included in the user data 220, the prediction unit 201 generates and outputs a predicted intervention result 221 including the intervention effect prediction value corresponding to each data sequence (step S206). Thereafter, the prediction unit 201 ends the prediction process.

[0090] Here, the screen presented by the prediction unit 201 will be described. FIG. 8 is a diagram showing an example of the screen presented by the prediction unit 201 of the first embodiment.

[0091] The prediction unit 201 presents a screen 800 to the user. The screen 800 includes an intervention content input field 801 and an intervention effect display field 802.

[0092] In the intervention content input field 801, a pattern setting field 810 for inputting an intervention pattern is displayed in a tabular format. The pattern setting field 810 includes a setting table 811, an add button 812, and a predict button 813. The setting table 811 is a table for setting the intervention content and stores entries including the type, amount, and timing of the intervention. The add button 812 is an operation button for adding an entry to the setting table 811. The predict button 813 is an operation button for instructing the execution of the prediction process. When the predict button 813 is operated, the user data 220 including the time-series data up to the present and the information in the setting table 811 is input to the prediction unit 201.

[0093] In addition, the intervention content may be set in a format different from the pattern setting field 810 in FIG. 8. For example, a method of displaying the type of intervention in a pull-down format and displaying a control bar for adjusting the amount and timing can be considered.

[0094] The intervention effect display column 802 is a column that displays the transition of the effect from the present to the past and the transition of the intervention effect. In the intervention effect display column 802, a graph showing the transition of the intervention effect is displayed for each intervention pattern. In FIG. 8, the transition of the intervention effect of each of the intervention pattern 1 in which intervention 1 is performed at time t1, the intervention pattern 2 in which intervention 2 is performed at the present and time t2, and the intervention pattern 3 in which no intervention is performed is displayed.

[0095] The system of Example 1 can highly accurately predict the effect when the influence of interfering factors is reduced and a plurality of types of interventions are continuously performed on a person.

[0096] When the input content in the intervention content input field 801 is updated, the display content in the intervention effect display column 802 is also updated. Note that only the effect transition of a specific intervention pattern may be displayed.

Example

[0097] When the predicted value of the intervention effect is corrected in the system of Example 2, the prediction is performed again. Hereinafter, Example 2 will be described centering on the differences from Example 1.

[0098] The configuration of the system of Example 2 is the same as that of Example 1. The functional configurations of the learning unit 200 and the prediction unit 201 of Example 2 are the same as those of Example 1. The process executed by the learning unit 200 of Example 2 is the same as that of Example 1.

[0099] In Example 2, the prediction process executed by the prediction unit 201 is partially different. FIG. 9 is a flowchart for explaining an example of the prediction process executed by the prediction unit 201 of Example 2. FIG. 10 is a diagram showing an example of the screen presented by the prediction unit 201 of Example 2.

[0100] The processes from step S201 to step S206 are the same as those in the first embodiment. After the process of step S206, the prediction unit 201 presents a screen and accepts an operation from the user (step S251). Here, the screen presented by the prediction unit 201 will be described with reference to FIG. 10.

[0101] The prediction unit 201 presents the screen 1000 to the user. The screen 1000 includes a correction field 1001 and an intervention effect display field 1002. The intervention effect display field 1002 is the same as the intervention effect display field 802.

[0102] The correction field 1001 includes a correction setting table 1011, an add button 1012, a prediction button 1013, and an end button 1014. The correction setting table 1011 is a table for setting the correction content of the intervention effect prediction value, and stores entries including timing and effects. The add button 1012 is an operation button for adding an entry to the correction setting table 1011. The prediction button 1013 is an operation button for instructing the execution of the prediction process again. When the prediction button 1013 is operated, the correction content is input to the prediction unit 201. The end button 1014 is an operation button for ending the prediction process.

[0103] Note that the correction content may be set in a format different from the correction field 1001 in FIG. 10. For example, a correction button is displayed, and when the correction button is operated, the points that can be corrected are displayed on the graph displayed in the intervention effect display field 1002. The user corrects the intervention effect prediction value by operating the points displayed on the graph with a mouse or the like.

[0104] The above is the description of the screen. Return to the description of FIG. 9.

[0105] The prediction unit 201 determines whether the operation received via the screen 1000 is a correction operation.

[0106] If the received operation is an end operation, the prediction unit 201 ends the prediction process.

[0107] When the received operation is a correction operation, the prediction unit 201 generates a data sequence to be used for a new prediction (step S253), and then returns to step S202. For example, the prediction unit 201 generates a data sequence that reflects the corrected intervention effect prediction value.

[0108] Note that the present invention is not limited to the above-described embodiments, and includes various modifications. Further, for example, the above-described embodiments have been described in detail for the sake of easy understanding of the present invention, and are not necessarily limited to those having all the configurations described. Also, it is possible to add, delete, or replace a part of the configuration of each embodiment with another configuration.

[0109] In addition, each of the above configurations, functions, processing units, processing means, etc. may be realized in hardware by designing a part or all of them, for example, by using an integrated circuit. Further, the present invention can also be realized by a program code of software that realizes the functions of the embodiments. In this case, a storage medium recording the program code 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 realizes the functions of the above-described embodiments, and the program code itself and the storage medium storing the same constitute the present invention. As a storage medium for supplying such a program code, for example, a flexible disk, a CD-ROM, a DVD-ROM, a hard disk, an SSD (Solid State Drive), an optical disk, a magneto-optical disk, a CD-R, a magnetic tape, a non-volatile memory card, a ROM, etc. are used.

[0110] Also, the program code for realizing the functions described in the present invention can be implemented in a wide range of programs or script languages such as assembler, C / C++, perl, Shell, PHP, Python, Java (registered trademark), etc.

[0111] Furthermore, the program code of the software that realizes the functions of the embodiments may be distributed via a network, stored in a storage means such as a hard disk or memory of a computer, or a storage medium such as a CD-RW or CD-R, and the processor provided in the computer may read and execute the program code stored in the storage means or the storage medium.

[0112] In the above-described embodiments, the control lines and information lines show those considered necessary for explanation, and not necessarily all the control lines and information lines on the product. All the components may be interconnected.

Description of Reference Numerals

[0113] 100 Computer 101 CPU 102 Main Memory Device 103 Auxiliary Storage Device 104 Network Adapter 105 Input Device 106 Output Device 108 Internal Bus 109 Network 110 Information Terminal 111 External Storage Device 200 Learning Unit 201 Prediction Unit 210 Learning Data DB 211 Model DB 220 User Data 221 Prediction Intervention Result 401 Time-Series Data Processing Unit 402 Interference Factor Adjustment Unit 403 Predictor 501 Discriminator 502, 503 Arithmetic Unit

Claims

1. A computer system for predicting the effects of multiple interventions on a person, comprising: at least one computer having a processor and a storage device connected to the processor, a first model that calculates an output value using time series data including values related to interventions performed on a person, a second model that calculates features by mapping the output value of the first model to a feature space, a third model that outputs a predicted value of the effect of the intervention on the person from the features, a fourth model that identifies the type of the intervention performed on the person from the features, and a loss function defined from the predicted type of the intervention output by the fourth model, the type of the intervention included in the learning data, the predicted value of the effect of the intervention, and the effect value of the intervention included in the learning data, and manages them; The time series data includes a plurality of data sequences including the time when the intervention was performed on the person, a plurality of factors representing the state of the person, and values indicating the type and degree of the intervention performed on the person. The processor: a process of receiving the learning data including a plurality of data sequences including the identification information of the person, the time when the intervention was performed on the person, the values of the plurality of factors of the person, the type and degree of the intervention received by the person, and the effect value of the intervention; a process of inputting the data sequence into the first model and inputting the output value output from the first model into the second model; a process of calculating a predicted value of the effect of the intervention by inputting the feature output from the second model into the third model; a process of calculating the predicted type of the intervention by inputting the feature output from the second model into the fourth model; a process of calculating the value of the loss function using the type and effect value of the intervention for each of the plurality of data sequences and the predicted type and predicted value of the effect of the intervention calculated from each of the plurality of data sequences; a process of updating the second model, the third model, and the fourth model using the value of the loss function; executes machine learning including; Furthermore, a process of calculating the output value by inputting the data sequence into the first model; a process of calculating the feature by inputting the output value into the second model; a process of calculating a predicted value of the continuous effect of the intervention corresponding to the time series data by inputting the feature into the third model; A computer system characterized by executing prediction processing including

2. The computer system according to claim 1, wherein the loss function is a function defined by a first loss function that evaluates the total error between the effect value of the intervention included in the data series and the predicted value of the effect of the intervention calculated from the data series, and a second loss function that evaluates the total error between the type of the intervention included in the data series and the predicted type of the intervention calculated from the data series. A computer system characterized by that.

3. The computer system according to claim 1, wherein the processor presents a first user interface for adjusting the type and degree of the intervention and the timing of the intervention for at least one of the data series included in the time series data, A computer system characterized by executing the prediction processing using the time series data including the data series input via the first user interface.

4. The computer system according to claim 1, wherein the processor presents a second user interface for displaying the predicted value of the effect of the intervention calculated from each of the plurality of data series, receives the correction content of the predicted value of the effect of the intervention via the second user interface, A computer system characterized by executing the prediction processing using the time series data including the data series in which the correction content of the predicted value of the effect of the intervention input via the second user interface is reflected.

5. An intervention effect prediction method for predicting the effects of a plurality of interventions on a person executed by a computer system, wherein the computer system includes at least one computer having a processor and a storage device connected to the processor, a first model that calculates an output value using time series data including values related to an intervention performed on a person, a second model that calculates a feature amount by mapping the output value of the first model to a feature amount space, and the feature amount And a third model that outputs a predicted value of the effect of the intervention on the person, a fourth model that identifies the type of the intervention performed on the person from the feature amount, the predicted type of the intervention output by the fourth model, and the intervention included in the learning data And a loss function defined from the type, the predicted value of the effect of the intervention, and the effect value of the intervention included in the learning data. The time series data includes a plurality of data sequences including the time when the intervention was performed on the person, a plurality of factors representing the state of the person, and the type and degree of the intervention performed on the person. The intervention effect prediction method is as follows. The step in which the processor receives the learning data including a plurality of data sequences including the identification information of the person, the time when the intervention was performed on the person, the values of the plurality of factors of the person, the type and degree of the intervention received by the person, and the effect value of the intervention. The step in which the processor inputs the data sequence into the first model and inputs the output value output from the first model into the second model. The step in which the processor calculates a predicted value of the effect of the intervention by inputting the feature amount output from the second model into the third model. The step in which the processor calculates a predicted intervention type by inputting the feature amount output from the second model into the fourth model. The step in which the processor calculates the value of the loss function by using the type of the intervention and the effect value of the intervention in each of the plurality of data sequences, and the predicted intervention type and the predicted value of the effect of the intervention calculated from each of the plurality of data sequences. The step in which the processor updates the second model, the third model, and the fourth model by using the value of the loss function. Performing machine learning including the above steps. Furthermore, The step in which the processor calculates the output value by inputting the data sequence into the first model. The step in which the processor calculates the feature amount by inputting the output value into the second model. The step in which the processor calculates a predicted value of the continuous effect of the intervention corresponding to the time series data by inputting the feature amount into the third model. An intervention effect prediction method characterized by performing prediction processing including the above steps.

6. The intervention effect prediction method according to claim 5, The loss function is a function defined by a first loss function that evaluates the total error between the effect value of the intervention included in the data sequence and the predicted value of the effect of the intervention calculated from the data sequence, and a second loss function that evaluates the total error between the type of the intervention included in the data sequence and the predicted intervention type calculated from the data sequence. **Claim 7**: An intervention effect prediction method according to claim 5, comprising: a step in which the processor presents a first user interface for adjusting the type and degree of the intervention and the timing of performing the intervention for at least one of the data sequences included in the time series data; a step in which the processor executes the prediction process using the time series data including the data sequence input via the first user interface. The intervention effect prediction method is characterized by including these steps. **Claim 8**: An intervention effect prediction method according to claim 5, comprising: a step in which the processor presents a second user interface for displaying predicted values of the effects of the intervention calculated from each of the plurality of data sequences; a step in which the processor receives the content of correction of the predicted value of the effect of the intervention via the second user interface; a step in which the processor executes the prediction process using the time series data including the data sequence in which the content of correction of the predicted value of the effect of the intervention input via the second user interface is reflected. The intervention effect prediction method is characterized by including these steps.

Images