CN-122025162-A - Dynamic intervention method and system for digital therapy of depression based on robust optimization and causal inference

CN122025162ACN 122025162 ACN122025162 ACN 122025162ACN-122025162-A

Abstract

The application relates to a dynamic intervention method and a system for digital therapy of depression, wherein the method comprises the steps of obtaining multi-source original data of a target user in a current decision period and related to psychological states, determining state data for representing the psychological states of the target user according to the multi-source original data, and determining an uncertainty set for quantifying measurement errors of the state data; based on the group characteristics of the target users, determining expected causal effect values of each intervention scheme in the intervention scheme library to the target users by using a causal effect analysis model, constructing a robust optimization model based on the state data and the uncertainty set, taking the expected causal effect values of each intervention scheme to the target users as an optimization target, solving the robust optimization model to determine the target intervention scheme from the intervention scheme library, and pushing the target intervention scheme to the target users. The application can dynamically push accurate personalized intervention strategies for depression to users under the condition of data uncertainty.

Inventors

LI DAI
LI YUANHUI
LIU XIANG
CHEN LU
HAO GUODONG

Assignees

阿呆科技（北京）有限公司

Dates

Publication Date: 20260512
Application Date: 20251224

Claims (10)

1. A method of dynamic intervention in digital therapy of depression, the method comprising: Acquiring multi-source original data related to the psychological state of a target user in a current decision period, determining state data used for representing the psychological state of the target user according to the multi-source original data, and determining an uncertainty set used for quantifying measurement errors of the state data; based on the group characteristics of the target user, determining expected causal effect values of each intervention scheme in an intervention scheme library on the target user by using a causal effect analysis model; constructing a robust optimization model based on the state data and the uncertainty set, taking expected causal effect values of each intervention scheme on the target user as an optimization target, and solving the robust optimization model to determine a target intervention scheme from the intervention scheme library; Pushing the target intervention scheme to the target user.
2. The method of claim 1, wherein the multi-source raw data comprises sensor objective data from a user terminal, and subjective assessment data from a user's active input; the determining state data for representing the psychological state of the target user according to the multi-source original data comprises the following steps: preprocessing and fusing the sensor objective data and the subjective evaluation data to generate a multidimensional state vector, wherein the multidimensional state vector comprises quantitative indexes of emotion, behavior liveness and circadian rhythm of a user; the multi-dimensional state vector is taken as the state data.
3. The method of claim 2, wherein preprocessing and fusing the sensor objective data and subjective assessment data to generate a multi-dimensional state vector comprises: cleaning the objective data of the sensor to remove noise and abnormal values, and performing interpolation processing on the missing values; performing standardization processing on the subjective evaluation data to convert the subjective evaluation data into a unified dimension and scoring range; extracting objective characteristic indexes from the processed sensor objective data, and extracting subjective characteristic indexes from the processed subjective evaluation data; And fusing the objective characteristic index and the subjective characteristic index to obtain the multidimensional state vector.
4. The method of claim 1, wherein the determining, based on the population characteristics of the target user, the expected causal effect value for each intervention plan in the library of intervention plans for the target user using a causal effect analysis model, comprises: acquiring group characteristics of the target user; Inputting the population characteristics into a pre-trained causal effect analysis model; And determining the expected intervention effect value of each intervention scheme for the target user according to the output data of the causal effect analysis model.
5. The method of claim 4, wherein the training process of the causal effect analysis model comprises: Constructing a training data set based on historical multi-source original data of a plurality of users, an actually executed intervention scheme and intervention effect feedback data; And selecting a mobile terminal behavior index which is strongly related to intervention distribution and is irrelevant to potential confounding factors as a tool variable, and training a tool variable forest model which is conditioned on the tool variable by using the training data set to obtain a causal effect analysis model for evaluating the causal effect of the intervention scheme on the heterogeneity of the user group with different characteristics.
6. The method of claim 1, wherein the constructing a robust optimization model based on the state data and the uncertainty set, taking expected causal effect values of individual intervention scenarios on the target user as optimization targets, solving the robust optimization model to determine a target intervention scenario from the intervention scenario library, comprises: Establishing a robust optimization model by taking the state data as an input parameter and defining a parameter fluctuation range by the uncertainty set; Solving the robust optimization model with a goal of minimizing a cost function based on the expected intervention effect value; and determining a target intervention scheme from the intervention scheme library according to the solving result.
7. The method according to claim 1, wherein the method further comprises: Acquiring behavior mode indexes of the multi-source original data of the target user in a preset historical time window; Performing similarity matching on the behavior pattern index and a pre-constructed group feature template library, wherein the group feature template library is established by clustering historical user data in advance and labeling typical behavior patterns of each user group based on clustering results; And acquiring group characteristics, of which the similarity with the behavior mode index is larger than a preset similarity threshold value, in the group characteristic template library as group characteristics of the target user.
8. A digital therapy dynamic intervention system for depression, the system comprising: The system comprises a user data determining module, a state data processing module and a state data processing module, wherein the user data determining module is used for obtaining multi-source original data related to the psychological state of a target user in a current decision period, determining state data used for representing the psychological state of the target user according to the multi-source original data, and determining an uncertainty set used for quantifying measurement errors of the state data; The expected intervention effect determining module is used for determining expected causal effect values of each intervention scheme in the intervention scheme library for the target user by utilizing a causal effect analysis model based on the group characteristics of the target user; The target intervention scheme determining module is used for constructing a robust optimization model based on the state data and the uncertainty set, taking expected causal effect values of each intervention scheme on the target user as an optimization target, and solving the robust optimization model to determine a target intervention scheme from the intervention scheme library; And the intervention scheme pushing module is used for pushing the target intervention scheme to the target user.
9. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method according to any one of claims 1 to 7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps of the method according to any one of claims 1 to 7 when the computer program is executed by the processor.

Description

Dynamic intervention method and system for digital therapy of depression based on robust optimization and causal inference Technical Field The embodiment of the application relates to the technical field of computers, in particular to a dynamic intervention method and a system for digital therapy of depression based on robust optimization and causal inference. Background The prior digital therapy software for depression, such as CBT-I, generally adopts a solidified intervention flow, namely a system pushes tasks with the same intensity and the same type to all users according to a preset sequence or a simple questionnaire threshold value, and can not be dynamically adjusted according to real-time emotion, behavior or physiological change. When the user is in a valley of emotion, a high load job may still be allocated, instead burdening, reducing compliance, and even inducing a negative effect. Meanwhile, the prior art mainly designs intervention based on group average effect or correlation analysis, and is difficult to effectively identify and evaluate heterogeneous treatment effects of the same intervention scheme on different user subgroups, and lacks a technical foundation for realizing true personalized accurate treatment. Disclosure of Invention The present application addresses the above-described shortcomings or drawbacks by providing a method and system for dynamic intervention of digital therapy for depression based on robust optimization and causal inference. The application realizes dynamic optimization and accurate pushing of the personalized intervention strategy of the depression under the uncertainty of data through closed-loop fusion of robust optimization and causal inference. The present application provides, according to a first aspect, a method of dynamic intervention in digital therapy for depression, the method comprising: Acquiring multi-source original data related to the psychological state of a target user in a current decision period, determining state data used for representing the psychological state of the target user according to the multi-source original data, and determining an uncertainty set of measurement errors used for quantifying the state data; based on the group characteristics of the target users, determining expected causal effect values of each intervention scheme in the intervention scheme library on the target users by using a causal effect analysis model; Constructing a robust optimization model based on the state data and the uncertainty set, taking expected causal effect values of each intervention scheme on the target user as an optimization target, and solving the robust optimization model to determine a target intervention scheme from an intervention scheme library; pushing a target intervention scheme to a target user. In some embodiments, the multi-source raw data includes sensor objective data from the user terminal and subjective assessment data from the user's active inputs, and determining state data for characterizing the psychological state of the target user from the multi-source raw data includes: preprocessing and fusing the sensor objective data and the subjective evaluation data to generate a multi-dimensional state vector, wherein the multi-dimensional state vector comprises quantitative indexes of emotion, behavior liveness and circadian rhythm of a user; The multidimensional state vector is used as state data. In some embodiments, preprocessing and fusing the sensor objective data and the subjective assessment data to generate a multi-dimensional state vector includes: Cleaning objective data of the sensor to remove noise and abnormal values, and performing interpolation processing on the missing values; Carrying out standardization processing on subjective evaluation data so as to convert the subjective evaluation data into a unified dimension and scoring range; extracting objective characteristic indexes from the processed sensor objective data, and extracting subjective characteristic indexes from the processed subjective evaluation data; And fusing the objective characteristic index and the subjective characteristic index to obtain the multidimensional state vector. In some embodiments, determining an expected causal effect value for each intervention plan in the intervention plan library for the target user based on the population characteristics of the target user using a causal effect analysis model, comprises: Acquiring group characteristics of target users; inputting group characteristics into a pre-trained causal effect analysis model; and determining the expected intervention effect value of each intervention scheme on the target user according to the output data of the causal effect analysis model. In some embodiments, the training process of the causal effect analysis model comprises: Constructing a training data set based on historical multi-source original data of a plurality of users, an actually executed intervention scheme