Data analysis system and method for early warning of psychological crisis in adolescents

Abstract

This application discloses a data analysis system and method for early warning of adolescent psychological crises, which uses natural language processing technology based on deep learning to perform semantic analysis on the psychological health-related data of adolescent subjects marked as psychologically healthy, extracts the core semantic features of the healthy psychological data, and uses this to establish a reference benchmark, and then based on the feature similarity between the psychological health-related data of the analyzed adolescent subjects and the reference benchmark, it realizes automatic early warning of adolescent psychological crises. In this way, the limitations of traditional evaluation methods can be effectively overcome, large-scale analysis and processing of adolescent psychological health data can be realized, and the efficiency and accuracy of psychological crisis identification can be improved.

Description

Technical Field

[0001] The present application relates to the field of data analysis, and more specifically, to a data analysis system and method for early warning of psychological crises in adolescents. Background Art

[0002] At present, early warning of adolescent psychological crisis mainly relies on professional psychological assessment and offline counseling services, but these methods are often limited by issues such as resource allocation, geographical coverage, and privacy protection. In recent years, with the development of big data and artificial intelligence technology, it has become possible to build an automated and intelligent psychological crisis early warning system by analyzing adolescents' online behaviors, social media activities, and mental health-related data. Therefore, an intelligent data analysis system and method for early warning of adolescent psychological crisis is expected. Summary of the invention

[0003] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a data analysis system and method for early warning of adolescent psychological crises, which uses natural language processing technology based on deep learning to perform semantic analysis on the mental health-related data of adolescent subjects marked as mentally healthy, extract the core semantic features of healthy psychological-related data, and use this to establish a reference benchmark, and then based on the feature similarity between the mental health-related data of the analyzed adolescent subjects and the reference benchmark, automatic early warning of adolescent psychological crises can be achieved. In this way, the limitations of traditional evaluation methods can be effectively overcome, large-scale analysis and processing of adolescent mental health data can be achieved, and the efficiency and accuracy of psychological crisis identification can be improved.

[0004] According to one aspect of the present application, a data analysis system for adolescent psychological crisis early warning is provided, which includes:

[0005] A module for acquiring mental health data to be analyzed, used to acquire mental health related data of the adolescent subject to be analyzed;

[0006] A reference mental health data acquisition module is used to acquire a collection of mental health related data of adolescent subjects marked as mentally healthy;

[0007] A mental health data semantic analysis module, used to semantically encode the mental health related data of the analyzed adolescent subject and the set of mental health related data of the adolescent subject marked as mentally healthy, respectively, to obtain a set of semantically encoded feature vectors of the mental health data of the analyzed adolescent subject and a set of semantically encoded feature vectors of the mental health data of the reference adolescent subject;

[0008] A reference information aggregation module, used for performing gated aggregation processing on the set of semantic coding feature vectors of the reference adolescent subject's mental health data to obtain a reference adolescent subject's mental health semantic coding aggregation representation vector;

[0009] The psychological crisis assessment module is used to determine whether to generate a psychological crisis warning prompt based on the feature similarity between the reference adolescent subject's psychological health semantic coding aggregation representation vector and the analyzed adolescent subject's psychological health data semantic coding feature vector.

[0010] According to another aspect of the present application, a data analysis method for adolescent psychological crisis early warning is provided, which includes:

[0011] Obtaining mental health-related data of the analyzed adolescent subjects;

[0012] Obtain a collection of mental health related data of adolescent subjects marked as mentally healthy;

[0013] Semantically encoding the mental health related data of the analyzed adolescent subject and the set of mental health related data of the adolescent subject marked as mentally healthy to obtain a set of semantically encoded feature vectors of the mental health data of the analyzed adolescent subject and a set of semantically encoded feature vectors of the mental health data of the reference adolescent subject;

[0014] Performing gated aggregation processing on the set of semantic coding feature vectors of the reference adolescent subject's mental health data to obtain a reference adolescent subject's mental health semantic coding aggregate representation vector;

[0015] Based on the feature similarity between the reference adolescent subject's mental health semantic coding aggregate representation vector and the analyzed adolescent subject's mental health data semantic coding feature vector, determine whether to generate a psychological crisis warning prompt.

[0016] Compared with the prior art, the data analysis system and method for adolescent psychological crisis warning provided by the present application uses natural language processing technology based on deep learning to perform semantic analysis on the psychological health-related data of adolescent subjects marked as psychologically healthy, extracts the core semantic features of the healthy psychological data, and thereby establishes a reference benchmark, and then based on the feature similarity between the psychological health-related data of the analyzed adolescent subjects and the reference benchmark, realizes automatic warning of adolescent psychological crises. In this way, the limitations of traditional evaluation methods can be effectively overcome, large-scale analysis and processing of adolescent psychological health data can be realized, and the efficiency and accuracy of psychological crisis identification can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0018] Figure 1 A block diagram of a data analysis system for adolescent psychological crisis warning according to an embodiment of the present application;

[0019] Figure 2 A data flow diagram of a data analysis system for early warning of psychological crisis of adolescents according to an embodiment of the present application;

[0020] Figure 3 A block diagram of the training phase of a data analysis system for adolescent psychological crisis early warning according to an embodiment of the present application;

[0021] Figure 4 The present invention is a flowchart of a data analysis method for adolescent psychological crisis warning according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0023] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0024] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0025] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0026] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0027] At present, early warning of adolescent psychological crisis mainly relies on professional psychological assessment and offline counseling services, but these methods are often limited by issues such as resource allocation, geographical coverage, and privacy protection. In recent years, with the development of big data and artificial intelligence technology, it has become possible to build an automated and intelligent psychological crisis early warning system by analyzing adolescents' online behaviors, social media activities, and mental health-related data. Therefore, an intelligent data analysis system and method for early warning of adolescent psychological crisis is expected.

[0028] In the technical solution of the present application, a data analysis system for early warning of psychological crises among adolescents is proposed. Figure 1 It is a block diagram of a data analysis system for adolescent psychological crisis warning according to an embodiment of the present application. Figure 2 FIG. 1 is a data flow diagram of a data analysis system for early warning of adolescent psychological crisis according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the embodiment of the present application, the data analysis system 300 for adolescent psychological crisis warning includes: a to-be-analyzed mental health data acquisition module 310, which is used to acquire the mental health-related data of the analyzed adolescent object; a reference mental health data acquisition module 320, which is used to acquire a set of mental health-related data of the adolescent object marked as mentally healthy; a mental health data semantic analysis module 330, which is used to semantically encode the mental health-related data of the analyzed adolescent object and the set of mental health-related data of the adolescent object marked as mentally healthy, respectively, to obtain a set of semantic encoding feature vectors of the mental health data of the analyzed adolescent object and a set of semantic encoding feature vectors of the mental health data of the reference adolescent object; a reference information aggregation module 340, which is used to perform gated aggregation processing on the set of semantic encoding feature vectors of the mental health data of the reference adolescent object to obtain a reference adolescent object mental health semantic encoding aggregate representation vector; a psychological crisis assessment module 350, which is used to determine whether to generate a psychological crisis warning prompt based on the feature similarity between the reference adolescent object mental health semantic encoding aggregate representation vector and the semantic encoding feature vector of the mental health data of the analyzed adolescent object.

[0029] In particular, the to-be-analyzed mental health data acquisition module 310 and the reference mental health data acquisition module 320 are used to acquire mental health-related data of the analyzed adolescent subject; and to acquire a collection of mental health-related data of adolescent subjects marked as mentally healthy. In the embodiments of the present application, mental health-related data of adolescent subjects can be acquired from multiple aspects and by multiple means, including but not limited to regular psychological screening activity reports and psychological test questionnaires.

[0030] In particular, the mental health data semantic analysis module 330 is used to semantically encode the mental health-related data of the analyzed adolescent subject and the set of mental health-related data of the adolescent subject labeled as mentally healthy, respectively, to obtain a set of semantically encoded feature vectors of the mental health data of the analyzed adolescent subject and a set of semantically encoded feature vectors of the mental health data of the reference adolescent subject. In a specific example of the present application, the mental health-related data of the analyzed adolescent subject is input into a mental health data semantic encoder based on the Bert model to obtain the semantically encoded feature vectors of the mental health data of the analyzed adolescent subject; and, the mental health-related data of each adolescent subject labeled as mentally healthy in the set of mental health-related data of adolescent subjects labeled as mentally healthy is input into the mental health data semantic encoder based on the Bert model to obtain the set of semantically encoded feature vectors of the mental health data of the reference adolescent subject. It should be understandable that in order to deeply understand the semantic connotation of adolescent mental health data, the present application adopts a mental health data semantic encoder based on the Bert model to semantically encode the various mental health-related data of the analyzed adolescent subjects and the adolescent subjects marked as mentally healthy, so as to utilize the powerful semantic understanding ability of the Bert model to extract the semantic meaning and contextual semantic dependency of each word unit in the mental health-related data, and realize the deep-level feature representation of the adolescent mental health data, so as to obtain a set of semantic encoding feature vectors of the mental health data of the analyzed adolescent subjects and semantic encoding feature vectors of the mental health data of the reference adolescent subjects.

[0031] In particular, the reference information aggregation module 340 is used to perform gated aggregation processing on the set of semantically encoded feature vectors of the reference adolescent subject's mental health data to obtain a reference adolescent subject's mental health semantically encoded aggregated representation vector. In a specific example of the present application, the set of semantically encoded feature vectors of the reference adolescent subject's mental health data is input into a feature aggregation network based on feature local energy distribution gating to obtain the reference adolescent subject's mental health semantically encoded aggregated representation vector. It should be understood that in order to comprehensively consider the feature diversity of the mental health-related data of adolescent subjects marked as mentally healthy, the set of semantically encoded feature vectors of the reference adolescent subject's mental health data is further feature integrated. At the same time, in order to improve the accuracy of feature expression, the present application uses a feature aggregation network based on feature local energy distribution gating to process the set of semantically encoded feature vectors of the reference adolescent subject's mental health data. Specifically, the feature aggregation network calculates its feature energy level coefficient based on the semantic feature statistics of each mental health-related data, and quantitatively represents its feature energy distribution and energy level. Next, the local neighborhood significance of the semantic coding feature vectors describing the mental health data of each reference adolescent subject is quantified by calculating the mean of the feature energy level coefficients within the local neighborhood, and the weights are screened and assigned according to their feature significance through gated masking technology. The set of semantic coding feature vectors of the mental health data of the reference adolescent subjects is adaptively weighted aggregated to enhance the expression of key information while suppressing the influence of noise information, thereby constructing a comprehensive and stable reference benchmark and obtaining a more representative reference adolescent subject mental health semantic coding aggregate representation vector for subsequent feature comparison and analysis.

[0032] In an embodiment of the present application, a set of semantically coded feature vectors of mental health data of reference adolescent subjects is gated and aggregated to obtain a semantically coded aggregate representation vector of mental health data of reference adolescent subjects, including: based on the maximum value, minimum value, mean value and variance of each semantically coded feature vector of mental health data of reference adolescent subjects in the set of semantically coded feature vectors of mental health data of reference adolescent subjects, determining the characteristic energy level coefficient of each semantically coded feature vector of mental health data of reference adolescent subjects to obtain a set of semantic characteristic energy level coefficients of mental health reference data; determining the scale of the local neighborhood, calculating the characteristic energy level coefficient of each semantic characteristic energy level coefficient of the ... The method comprises the following steps: obtaining a set of mental health reference data semantic feature energy level coefficients using local neighborhood significance descriptors of the mental health reference data semantic feature energy level coefficients to obtain a set of mental health reference data semantic feature energy level significance descriptors; inputting the set of mental health reference data semantic feature energy level significance descriptors into a mask module based on a gating function to obtain a set of mask-gated probabilistic mental health reference data semantic feature energy level significance descriptors; using the set of mask-gated probabilistic mental health reference data semantic feature energy level significance descriptors as a set of weights, calculating the position-weighted sum of the set of reference adolescent subject mental health data semantic encoding feature vectors to obtain the reference adolescent subject mental health semantic encoding aggregate representation vector.

[0033] Among them, based on the maximum value, minimum value, mean value and variance of each reference adolescent subject's mental health data semantic coding feature vector in the set of reference adolescent subject's mental health data semantic coding feature vectors, the process of determining the characteristic energy level coefficient of each reference adolescent subject's mental health data semantic coding feature vector to obtain a set of mental health reference data semantic characteristic energy level coefficients includes: calculating the variance and the sum of the bias terms of the reference adolescent subject's mental health data semantic coding feature vector to obtain a first intermediate term, and calculating the square of the difference between the characteristic maximum value and the mean of the reference adolescent subject's mental health data semantic coding feature vector, the square of the difference between the characteristic minimum value and the mean of the reference adolescent subject's mental health data semantic coding feature vector, and the sum of the bias terms to obtain a second intermediate term; calculating the first intermediate term divided by the second intermediate term to obtain the mental health reference data semantic characteristic energy level coefficient.

[0034] In summary, in the above embodiment, the set of semantically encoded feature vectors of the reference adolescent subject's mental health data is gated and aggregated to obtain a reference adolescent subject's mental health semantically encoded aggregate representation vector, including: processing the set of semantically encoded feature vectors of the reference adolescent subject's mental health data with the following feature aggregation formula to obtain the reference adolescent subject's mental health semantically encoded aggregate representation vector, wherein the feature aggregation formula is:

[0035]

[0036]

[0037]

[0038]

[0039]

[0040]

[0041]

[0042] in, represents a set of semantically encoded feature vectors of the mental health data of the reference adolescent subject, The value of is the number of feature vectors in the set of semantically encoded feature vectors of the reference adolescent subject's mental health data, , , and Respectively represent the first, second, and third in the set of semantic encoding feature vectors of the reference adolescent subject's mental health data and A semantic encoding feature vector of the psychological health data of reference adolescent subjects, For the said The feature mean of the semantic encoding feature vector of the psychological health data of the reference adolescent subjects, For the said The feature variance of the semantic encoding feature vector of the reference adolescent subject mental health data, and Respectively represent the The maximum and minimum values ​​of each eigenvalue of the semantic encoding feature vector of the reference adolescent subject mental health data, is the bias term, Indicates the The mental health reference data semantic feature energy level coefficient of the semantic encoding feature vector of the mental health data of the reference adolescent subjects, Indicates The local neighborhood centered at The semantic feature energy level coefficient of mental health reference data, represents the number of semantic feature energy level coefficients of mental health reference data in the local neighborhood, For the The semantic feature energy level saliency descriptors of mental health reference data, Indicates probabilistic mental health reference data semantic feature energy level significance descriptor, Indicates mask processing, Indicates the preset mask threshold, Indicates mask-gated probabilistic mental health reference data semantic feature energy level saliency descriptors, An aggregate representation vector is provided for the mental health semantic encoding of the reference adolescent subject.

[0043] In particular, the psychological crisis assessment module 350 is used to determine whether to generate a psychological crisis warning prompt based on the feature similarity between the reference adolescent object's psychological health semantic coding aggregate representation vector and the analyzed adolescent object's psychological health data semantic coding feature vector. In a specific example of the present application, the cosine similarity between the reference adolescent object's psychological health semantic coding aggregate representation vector and the analyzed adolescent object's psychological health data semantic coding feature vector is first calculated; here, in order to quantitatively evaluate the information similarity between the analyzed adolescent object and the reference adolescent object's psychological health data, the present application further calculates the cosine similarity between the reference adolescent object's psychological health semantic coding aggregate representation vector and the analyzed adolescent object's psychological health data semantic coding feature vector to characterize the degree of deviation of the analyzed adolescent object's psychological characteristics from the health benchmark. It should be understood that cosine similarity is a commonly used measurement method that can effectively measure the directional consistency of two vectors in a high-dimensional space, thereby reflecting the difference between the psychological characteristics of the analyzed adolescent object and the average health status of the reference adolescent object. Further, based on the comparison between the cosine similarity and a preset threshold, it is determined whether to generate the psychological crisis warning prompt. Specifically, after obtaining the cosine similarity, a threshold is set to determine whether the cosine similarity exceeds a preset psychological health threshold range. If the cosine similarity is lower than the preset threshold, it indicates that there is a significant difference between the psychological characteristics of the analyzed adolescent object and the average health status of the reference adolescent object, and there may be a risk of psychological crisis. On the contrary, if the cosine similarity is within the preset threshold range, the psychological characteristics of the analyzed adolescent object can be considered to be consistent with the health status of the reference adolescent object, and the warning is not triggered temporarily.

[0044] It should be understood that before using the above-mentioned neural network model for inference, it is necessary to train the mental health data semantic encoder based on the Bert model and the feature aggregation network based on feature local energy distribution gating. That is to say, according to the data analysis system 300 for adolescent psychological crisis warning of the present application, it also includes a training phase for training the mental health data semantic encoder based on the Bert model and the feature aggregation network based on feature local energy distribution gating.

[0045] Figure 3 FIG. 1 is a block diagram of the training phase of the data analysis system for adolescent psychological crisis early warning according to an embodiment of the present application. Figure 3 As shown, according to the data analysis system 300 for adolescent psychological crisis warning according to the embodiment of the present application, it includes: a training module 400, including: a training data acquisition unit 410, used to acquire training data, wherein the training data includes training mental health related data of the analyzed adolescent object, a set of training mental health related data of the adolescent object marked as mentally healthy, and the true value of cosine similarity; an analyzed object mental health data semantic encoding unit 420, used to input the training mental health related data of the analyzed adolescent object into the mental health data semantic encoder based on the Bert model to obtain the training mental health data semantic encoding feature vector of the analyzed adolescent object; a reference object mental health data semantic encoding unit 430, used to input the training mental health related data of each adolescent object marked as mentally healthy in the set of training mental health related data of the adolescent object marked as mentally healthy into the mental health data semantic encoder based on the Bert model to obtain a set of training reference adolescent object mental health data semantic encoding feature vectors; a training data feature aggregation unit 440, used to input the training mental health related data of the adolescent object marked as mentally healthy into the mental health data semantic encoder based on the Bert model to obtain a set of training reference adolescent object mental health data semantic encoding feature vectors. The set of semantic coding feature vectors of the training reference adolescent subject's mental health data is input into the feature aggregation network based on feature local energy distribution gating to obtain the training reference adolescent subject's mental health semantic coding aggregation representation vector; the training cosine similarity measurement unit 450 is used to calculate the training cosine similarity between the training reference adolescent subject's mental health semantic coding aggregation representation vector and the training analyzed adolescent subject's mental health data semantic coding feature vector; the cosine similarity loss calculation unit 460 is used to calculate the difference between the training cosine similarity and the true value of the cosine similarity to obtain the cosine similarity loss value; the constraint loss function value calculation unit 470 is used to calculate the constraint loss function value of the training reference adolescent subject's mental health semantic coding aggregation representation vector and the training analyzed adolescent subject's mental health data semantic coding feature vector; the model training unit 480 is used to train the Bert model-based mental health data semantic encoder and the feature aggregation network based on feature local energy distribution gating with the weighted sum of the cosine similarity loss value and the constraint loss function value as the loss function value.

[0046] Preferably, considering that the training reference adolescent subject mental health semantic coding aggregate representation vector and the training analyzed adolescent subject mental health data semantic coding feature vector respectively represent the feature aggregate representation of the textual semantic features of the set of training mental health-related data of adolescent subjects labeled as mentally healthy and the textual semantic features of the training mental health-related data of the analyzed adolescent subjects, the training reference adolescent subject mental health semantic coding aggregate representation vector and the training analyzed adolescent subject mental health data semantic coding feature vector may also cause an imbalance in the proportion of correspondence in the fine-grained distribution of image semantics due to the difference in the sample size of textual semantic features, resulting in a mapping offset relative to the common cosine similarity domain. To this end, the applicant of the present application hopes to improve the common domain mapping offset of the training reference adolescent subject mental health semantic coding aggregate representation vector and the training analyzed adolescent subject mental health data semantic coding feature vector under the imbalance in the proportion of correspondence in the fine-grained distribution of image semantics, thereby improving the calculation accuracy of cosine similarity.

[0047] Based on this, the applicant of the present application introduces a constrained loss function for the training reference adolescent subject mental health semantic coding aggregate representation vector and the training analyzed adolescent subject mental health data semantic coding feature vector, including the steps of:

[0048] Calculate the first norm and second norm of the mean feature vector of the training reference adolescent subject psychological health semantic coding aggregate representation vector and the training analyzed adolescent subject psychological health data semantic coding feature vector;

[0049] Calculate the weighted sum of the inverse of the square root of the second norm and the first norm, and perform point multiplication with the difference result of the training reference adolescent subject mental health semantic coding aggregate representation vector and the training analyzed adolescent subject mental health data semantic coding feature vector, and sum all eigenvalues ​​of the point product result feature vector to obtain the first loss term;

[0050] Perform dot-wise addition of the dotted feature vector of the training reference adolescent subject mental health semantic coding aggregate representation vector and the training analyzed adolescent subject mental health data semantic coding feature vector and the square root of the scale of the upper gastrointestinal tract state difference feature vector, and sum all eigenvalues ​​of the dot-wise added feature vector to obtain a second loss term; and

[0051] The loss function is obtained by calculating a difference between the first loss term and the second loss term multiplied by a weight as a hyperparameter.

[0052] Wherein, the constraint loss function is expressed as:

[0053]

[0054] is the mean feature vector of the aggregate representation vector of the training reference adolescent subject mental health semantic coding and the feature vector of the training analyzed adolescent subject mental health data semantic coding, and denote the first norm and second norm of the mean eigenvector respectively, and They are respectively the training reference adolescent subject mental health semantic coding aggregate representation vector and the training analyzed adolescent subject mental health data semantic coding feature vector, is the scale of the training reference adolescent subject mental health semantic coding aggregate representation vector and the training analyzed adolescent subject mental health data semantic coding feature vector, and and is the weighted sum weight as a hyperparameter, is the constraint loss function value.

[0055] Therefore, based on the mean feature vector norm between the training reference adolescent subject's mental health semantic coding aggregate representation vector and the training analyzed adolescent subject's mental health data semantic coding feature vector, the constrained representation of the structured foreground and background distinction of the supermanifold of the training reference adolescent subject's mental health semantic coding aggregate representation vector and the training analyzed adolescent subject's mental health data semantic coding feature vector is used to model the feature-level key differential correspondence between the feature vectors of the training reference adolescent subject's mental health semantic coding aggregate representation vector and the training analyzed adolescent subject's mental health data semantic coding feature vector, and adjust the global differential association relationship between the corresponding features, so as to make positive fine-grained differential correspondence suggestions through the unbalanced ratio control between the corresponding eigenvalues ​​of the training reference adolescent subject's mental health semantic coding aggregate representation vector and the training analyzed adolescent subject's mental health data semantic coding feature vector, so as to avoid the imbalance of the feature vector-level image semantic distribution correspondence ratio of the training reference adolescent subject's mental health semantic coding aggregate representation vector and the training analyzed adolescent subject's mental health data semantic coding feature vector by focusing on the focus, improve its mapping offset relative to the common cosine similarity domain, and thus improve the accuracy of the obtained cosine similarity.

[0056] As described above, the data analysis system 300 for adolescent psychological crisis warning according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a data analysis algorithm for adolescent psychological crisis warning. In a possible implementation, the data analysis system 300 for adolescent psychological crisis warning according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or a hardware module. For example, the data analysis system 300 for adolescent psychological crisis warning can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the data analysis system 300 for adolescent psychological crisis warning can also be one of the many hardware modules of the wireless terminal.

[0057] Alternatively, in another example, the data analysis system 300 for adolescent psychological crisis warning and the wireless terminal may be separate devices, and the data analysis system 300 for adolescent psychological crisis warning may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0058] Furthermore, a data analysis method for early warning of psychological crises among adolescents is also provided.

[0059] Figure 4 Flow chart of a data analysis method for adolescent psychological crisis warning according to an embodiment of the present application. Figure 4 As shown, according to the embodiment of the present application, the data analysis method for adolescent psychological crisis warning includes the following steps: S1, obtaining the mental health related data of the analyzed adolescent object; S2, obtaining a set of mental health related data of the adolescent object marked as mentally healthy; S3, semantically encoding the mental health related data of the analyzed adolescent object and the set of mental health related data of the adolescent object marked as mentally healthy, respectively, to obtain a set of semantic encoding feature vectors of the mental health data of the analyzed adolescent object and a set of semantic encoding feature vectors of the mental health data of the reference adolescent object; S4, performing gated aggregation processing on the set of semantic encoding feature vectors of the mental health data of the reference adolescent object to obtain a reference adolescent object mental health semantic encoding aggregate representation vector; S5, determining whether to generate a psychological crisis warning prompt based on the feature similarity between the reference adolescent object mental health semantic encoding aggregate representation vector and the semantic encoding feature vector of the mental health data of the analyzed adolescent object.

[0060] In summary, the data analysis method for adolescent psychological crisis warning according to the embodiment of the present application is explained, which uses natural language processing technology based on deep learning to perform semantic analysis on the psychological health-related data of adolescent subjects marked as psychologically healthy, extracts the core semantic features of healthy psychological data, and uses this to establish a reference benchmark, and then based on the feature similarity between the psychological health-related data of the analyzed adolescent subjects and the reference benchmark, it can realize automatic warning of adolescent psychological crises. In this way, the limitations of traditional evaluation methods can be effectively overcome, large-scale analysis and processing of adolescent psychological health data can be realized, and the efficiency and accuracy of psychological crisis identification can be improved.

[0061] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A data analysis system for early warning of adolescent psychological crisis, characterized in that: include: A module for acquiring mental health data to be analyzed, used to acquire mental health related data of the adolescent subject to be analyzed; A reference mental health data acquisition module is used to acquire a collection of mental health related data of adolescent subjects marked as mentally healthy; A mental health data semantic analysis module, used for semantically encoding the mental health related data of the analyzed adolescent subject and the set of mental health related data of the adolescent subject marked as mentally healthy, respectively, to obtain a set of semantically encoded feature vectors of the mental health data of the analyzed adolescent subject and a set of semantically encoded feature vectors of the mental health data of the reference adolescent subject; A reference information aggregation module, used for performing gated aggregation processing on the set of semantic coding feature vectors of the reference adolescent subject's mental health data to obtain a reference adolescent subject's mental health semantic coding aggregation representation vector; A psychological crisis assessment module, used to determine whether to generate a psychological crisis warning prompt based on the feature similarity between the reference adolescent subject's psychological health semantic coding aggregate representation vector and the analyzed adolescent subject's psychological health data semantic coding feature vector; Wherein, the mental health data semantic analysis module is used to: Inputting the mental health related data of the analyzed adolescent subject into a mental health data semantic encoder based on the Bert model to obtain a semantic encoding feature vector of the mental health data of the analyzed adolescent subject; Inputting the mental health related data of each adolescent subject marked as mentally healthy in the set of mental health related data of adolescent subjects marked as mentally healthy into the mental health data semantic encoder based on the Bert model to obtain a set of semantic encoding feature vectors of mental health data of reference adolescent subjects; Wherein, the reference information aggregation module is used to: Inputting the set of semantically encoded feature vectors of the psychological health data of the reference adolescent subject into a feature aggregation network based on feature local energy distribution gating to obtain a semantically encoded aggregated representation vector of the psychological health of the reference adolescent subject; Wherein, the reference information aggregation module includes: A characteristic energy level calculation unit, used to determine the characteristic energy level coefficients of the semantic coding feature vectors of the mental health data of each reference adolescent subject based on the maximum value, minimum value, mean value and variance of the semantic coding feature vectors of the mental health data of each reference adolescent subject in the set of semantic coding feature vectors of the mental health data of the reference adolescent subject, so as to obtain a set of semantic characteristic energy level coefficients of the mental health reference data; A local neighborhood significance description unit, used to determine the scale of the local neighborhood, calculate the local neighborhood significance descriptor of each mental health reference data semantic feature energy level coefficient in the mental health reference data semantic feature energy level coefficient to obtain a set of mental health reference data semantic feature energy level significance descriptors; A gated screening unit, used for inputting the set of mental health reference data semantic feature energy level significance descriptors into a mask module based on a gating function to obtain a set of masked gated probabilistic mental health reference data semantic feature energy level significance descriptors; A feature aggregation unit, used to calculate the position-weighted sum of the set of semantic coding feature vectors of the reference adolescent subject's mental health data using the set of semantic feature energy level significance descriptors of the mask-gated probabilistic mental health reference data as a set of weights to obtain a semantic coding aggregation representation vector of the reference adolescent subject's mental health; Wherein, the characteristic energy level calculation unit is used to: Calculate the sum of the variance and bias term of the semantic coding feature vector of the reference adolescent subject's mental health data to obtain a first intermediate term, and calculate the square of the difference between the maximum feature value and the mean value of the semantic coding feature vector of the reference adolescent subject's mental health data, the square of the difference between the minimum feature value and the mean value of the semantic coding feature vector of the reference adolescent subject's mental health data, and the sum of the bias term to obtain a second intermediate term; Calculate the first intermediate term and divide it by the second intermediate term to obtain the semantic feature energy level coefficient of the mental health reference data; The gated aggregation processing is performed on the set of semantic coding feature vectors of the mental health data of the reference adolescent subject to obtain the reference adolescent subject mental health semantic coding aggregate representation vector, including: the set of semantic coding feature vectors of the mental health data of the reference adolescent subject is processed by the following feature aggregation formula to obtain the reference adolescent subject mental health semantic coding aggregate representation vector, wherein the feature aggregation formula is: X={χ1,χ2,...,χ n } w si =mask(w i ) Wherein, X represents the set of semantically coded feature vectors of the reference adolescent subject's mental health data, the value of n is the number of feature vectors in the set of semantically coded feature vectors of the reference adolescent subject's mental health data, and χ1, χ2, χ j and χ n Respectively represent the first, second, i-th and n-th reference adolescent subject psychological health data semantic coding feature vectors in the set of reference adolescent subject psychological health data semantic coding feature vectors, μ i is the feature mean of the semantic encoding feature vector of the mental health data of the i-th reference adolescent subject, σ 2 i is the feature variance of the semantic encoding feature vector of the mental health data of the i-th reference adolescent subject, max(χ i ) and min(χ i ) represent the maximum and minimum values ​​of each eigenvalue of the semantic encoding feature vector of the mental health data of the i-th reference adolescent subject, ∈ is the bias term, e i The mental health reference data semantic feature energy level coefficient, e, represents the semantic encoding feature vector of the mental health data of the i-th reference adolescent object. η Indicated by e i The qth mental health reference data semantic feature energy level coefficient in the local neighborhood centered at r represents the number of semantic feature energy level coefficients of mental health reference data in the local neighborhood, e ci is the energy level saliency descriptor of the semantic feature of the i-th mental health reference data, w i represents the energy level saliency descriptor of the semantic feature of the i-th probabilistic mental health reference data, mask(·) represents mask processing, γ represents the preset mask threshold, and w si represents the i-th mask-gated probabilistic mental health reference data semantic feature energy level significance descriptor, and v is the reference adolescent subject's mental health semantic encoding aggregate representation vector.

2. The data analysis system for early warning of psychological crisis of adolescents according to claim 1 is characterized in that: The psychological crisis assessment module includes: A similarity evaluation unit, used to calculate the cosine similarity between the reference adolescent subject's mental health semantic coding aggregate representation vector and the analyzed adolescent subject's mental health data semantic coding feature vector; The early warning prompt generating unit is used to determine whether to generate the psychological crisis early warning prompt based on the comparison between the cosine similarity and a preset threshold.

3. The data analysis system for early warning of psychological crisis of adolescents according to claim 2 is characterized in that: It also includes a training module for training the mental health data semantic encoder based on the Bert model and the feature aggregation network based on feature local energy distribution gating.

4. The data analysis system for early warning of psychological crisis of adolescents according to claim 3 is characterized in that: The training module comprises: A training data acquisition unit, used to acquire training data, wherein the training data includes training mental health related data of the analyzed adolescent subject, a set of training mental health related data of the adolescent subject marked as mentally healthy, and a true value of the cosine similarity; The analyzed object mental health data semantic encoding unit is used to input the training mental health related data of the analyzed adolescent object into the mental health data semantic encoder based on the Bert model to obtain the training semantic encoding feature vector of the mental health data of the analyzed adolescent object; A reference subject mental health data semantic encoding unit, used to input the training mental health related data of each adolescent subject marked as mentally healthy in the set of training mental health related data of adolescent subjects marked as mentally healthy into the mental health data semantic encoder based on the Bert model to obtain a set of semantic encoding feature vectors of the training reference adolescent subject mental health data; A training data feature aggregation unit, used for inputting the set of semantic coding feature vectors of the training reference adolescent subject mental health data into the feature aggregation network based on feature local energy distribution gating to obtain a training reference adolescent subject mental health semantic coding aggregation representation vector; A training cosine similarity measurement unit, used to calculate the training cosine similarity between the training reference adolescent subject psychological health semantic coding aggregate representation vector and the training analyzed adolescent subject psychological health data semantic coding feature vector; A cosine similarity loss calculation unit, used for calculating the difference between the training cosine similarity and the true value of the cosine similarity to obtain a cosine similarity loss value; A constraint loss function value calculation unit, used to calculate the constraint loss function value of the training reference adolescent subject mental health semantic coding aggregate representation vector and the training analyzed adolescent subject mental health data semantic coding feature vector; A model training unit is used to train the mental health data semantic encoder based on the Bert model and the feature aggregation network based on feature local energy distribution gating by using the weighted sum of the cosine similarity loss value and the constraint loss function value as the loss function value.

5. A data analysis method for adolescent psychological crisis early warning, using the data analysis system for adolescent psychological crisis early warning according to claim 1, characterized in that: include: Obtaining mental health-related data of the analyzed adolescent subjects; Obtain a collection of mental health related data of adolescent subjects marked as mentally healthy; Semantically encoding the mental health related data of the analyzed adolescent subject and the set of mental health related data of the adolescent subject marked as mentally healthy to obtain a set of semantically encoded feature vectors of the mental health data of the analyzed adolescent subject and a set of semantically encoded feature vectors of the mental health data of the reference adolescent subject; Performing gated aggregation processing on the set of semantic coding feature vectors of the reference adolescent subject's mental health data to obtain a reference adolescent subject's mental health semantic coding aggregate representation vector; Based on the feature similarity between the reference adolescent subject's mental health semantic coding aggregate representation vector and the analyzed adolescent subject's mental health data semantic coding feature vector, determine whether to generate a psychological crisis warning prompt.

Images