A mental illness early warning system and method

Abstract

The present invention discloses a mental illness early warning system and method, belonging to the technical field of mental illness prevention. The data acquisition module processes raw data to generate a target data set; the model training module analyzes and learns the target data set through a machine learning algorithm to generate a large model for mental illness prediction; the disease prediction module uses the large model to predict the mental illness situation in the target data set and generates a corresponding prediction value; the early warning interaction module receives the user's query request, and after determining the need to warn the user in combination with the prediction value, performs a psychological situation early warning; the feedback optimization module adjusts the model parameters corresponding to the large model according to the early warning accuracy report and the declaration record. By integrating the machine learning model and big data analysis, the system can effectively perform mental illness early warning and provide accurate and in-depth judgment results.

Description

Technical Field

[0001] The present invention relates to the technical field of mental illness prevention, and in particular to a mental illness early warning system and method. Background Art

[0002] In the current field of mental illness prevention technology, the main function of traditional mental illness identification is only to screen mental illness through psychological scales, and mental illness early warning systems often make a rough prediction based on relevant data. These systems have obvious limitations in dealing with college situations, including the lack of mental illness labels and prediction accuracy. With the increasing number of psychological problems in colleges and universities, and the popularization of Internet and Internet of Things technologies in colleges and universities, higher requirements have been placed on mental illness early warning technology.

[0003] In addition, traditional mental illness early warning systems usually lack the ability to capture and analyze behavioral data, and use deep learning models for prediction, but lack control over prediction accuracy. This greatly increases the cost of solving mental illness in the next step. This leads to insufficient user experience and low early warning efficiency.

[0004] With the development of artificial intelligence and machine learning technology, especially the advancement of deep learning and reinforcement learning methods, a new opportunity has emerged, that is, to use these advanced technologies to address the limitations of traditional mental illness early warning systems. An effective modern mental illness early warning system can learn from user feedback data, continuously optimize system parameters, and provide more accurate early warnings, thereby better solving the problem of difficult early warnings for mental illness in colleges and universities.

[0005] The above contents are only used to assist in understanding the technical solution of the present invention and do not constitute an admission that the above contents are prior art. Summary of the invention

[0006] The main purpose of the present invention is to provide a mental illness early warning system and method, aiming at solving the technical problem of how to accurately perform mental illness early warning.

[0007] In order to achieve the above-mentioned object, the present invention provides a mental illness early warning system, the system comprising: a data acquisition module, a model training module, a disease prediction module, an early warning interaction module and a feedback optimization module;

[0008] The data acquisition module is used to process raw data to generate a target data set, and send the target data set to the model training module, wherein the raw data includes psychological survey data, daily behavior data, and social relationship data;

[0009] The model training module is used to analyze and learn the target data set through a machine learning algorithm to generate a large model for predicting mental illness;

[0010] The disease prediction module is used to predict the mental illness conditions in the target data set using the large model and generate corresponding prediction values;

[0011] The early warning interaction module is used to receive a query request from a user, and after determining that a warning is needed for the user in combination with the predicted value, issue a psychological warning;

[0012] The warning interaction module also includes: a warning display submodule, a warning confirmation submodule and a warning reporting submodule;

[0013] The early warning display submodule is used to display data visually according to the query request of the user when performing psychological condition early warning;

[0014] The warning confirmation submodule is used to generate a warning accuracy report based on the feedback information received from the user after the psychological condition warning is performed;

[0015] The warning reporting submodule is used to report and generate reporting records for target subjects who have not received psychological warning but have been diagnosed with psychological diseases;

[0016] The feedback optimization module is used to adjust the model parameters corresponding to the large model according to the early warning accuracy report and the declaration record.

[0017] Optionally, the data collection module is further used to collect standard psychological scale survey and psychological interview data to generate the psychological survey data, and to extract psychological symptom information from the psychological interview data using text mining technology and natural language processing algorithms, wherein the psychological symptom information includes: physical discomfort symptoms, obsessive-compulsive symptoms, interpersonal relationship disorder symptoms, depression symptoms, anxiety symptoms, hostile emotion symptoms, fear symptoms, paranoia symptoms, and mental illness symptoms;

[0018] Acquire daily behavior data by presetting the first address set, the daily behavior data including: academic performance and changes, class attendance rate, physical test performance and changes, physiological function indicators and changes, campus entry and exit, canteen dining, dormitory entry and exit time distribution and dormitory time;

[0019] Acquire social relationship data by combining the preset second address set with the survey report, wherein the social relationship data includes family situation, family income situation, love situation, marriage situation and interpersonal communication situation;

[0020] The psychological survey data, the daily behavior data and the social relationship data are preprocessed to generate a target data set, wherein the preprocessing includes data cleaning, data standardization and data classification.

[0021] Optionally, the model training module is further used to classify the target data set according to the results of the psychological survey data, using a five-item scoring method to classify from six perspectives: depression, anxiety disorders, stress and stress disorders, eating disorders, sleep disorders, and addiction disorders, and marking no symptoms, mild symptoms, moderate symptoms, severe symptoms, and severe symptoms in each classification result as 0, 0.25, 0.5, 0.75, and 1, respectively;

[0022] Construct a recursive neural network RNN ​​model to extract data features F:

[0023]

[0024] Among them, R is the output result of the recurrent neural network RNN ​​model, and Represent the weight and bias of the fully connected layer respectively;

[0025] Define the loss function and use the Adam optimizer for model training;

[0026] Cross-validation is performed on a preset data set to adjust model parameters corresponding to the recursive neural network (RNN) model.

[0027] Optionally, the disease prediction module is further used to generate a mental illness prediction value through the trained large model parameters, where P is a 6-dimensional vector, and each component They represent depression, anxiety disorders, stress and stress disorders, eating disorders, sleep disorders, and addiction disorders;

[0028] Warning value The calculation formula is: ,in represents the parameter matrix of the large model, Represents the data in the target data set.

[0029] Optionally, the early warning display submodule is further used to receive query information input by the user, and determine the query type according to the keywords in the query information, and the query model includes personal psychological early warning query and group psychological early warning query;

[0030] Matching corresponding disease prediction data according to the query type;

[0031] Early warning information is generated according to the disease prediction data and the early warning information is displayed in a data visualization manner.

[0032] Optionally, the human psychological early warning query includes searching with individuals as keywords, displaying the personal search results in the form of a six-dimensional radar chart, and calculating the similarity between the user query vector q and the feature vector of each disease information in the disease prediction module;

[0033] The calculation formula for cosine similarity is:

[0034]

[0035] Wherein d represents the characteristic vector of a certain disease information in the disease prediction module.

[0036] Optionally, the group psychological early warning query includes searching with the group name as a keyword, displaying the group search results in the form of a six-dimensional radar chart, and calculating the overall data mean:

[0037]

[0038] Among them, x represents the data of all individuals of a certain indicator, and N represents the dimension of vector x.

[0039] In addition, to achieve the above-mentioned purpose, the present invention also proposes a mental illness early warning method, which is applied to the mental illness early warning system, the system includes a data acquisition module, a model training module, a disease prediction module, an early warning interaction module and a feedback optimization module, the early warning interaction module includes an early warning display submodule, an early warning confirmation submodule and an early warning reporting submodule, the method includes:

[0040] The data acquisition module processes the raw data to generate a target data set, and sends the target data set to the model training module, wherein the raw data includes psychological survey data, daily behavior data, and social relationship data;

[0041] The model training module analyzes and learns the target data set through a machine learning algorithm to generate a large model for predicting mental illness;

[0042] The disease prediction module uses the large model to predict the mental illness conditions in the target data set and generates corresponding prediction values;

[0043] The early warning interaction module receives a query request from a user, and after determining that an early warning is required for the user in combination with the predicted value, issues an early warning for the psychological condition;

[0044] When the early warning display submodule performs a psychological condition early warning, the early warning display submodule performs a data visualization display according to the query request of the user;

[0045] The early warning confirmation submodule generates an early warning accuracy report based on feedback information received from the user after the psychological condition early warning is performed;

[0046] The early warning reporting submodule reports and generates a reporting record for the target object who has not received a psychological warning but has been diagnosed with a psychological disease;

[0047] The feedback optimization module adjusts the model parameters corresponding to the large model according to the early warning accuracy report and the declaration record.

[0048] Optionally, the step of processing the raw data to generate a target data set and sending the target data set to the model training module includes:

[0049] Collecting standard psychological scale survey and psychological interview data to generate the psychological survey data, and extracting psychological symptom information from the psychological interview data using text mining technology and natural language processing algorithms, wherein the psychological symptom information includes: physical discomfort symptoms, obsessive-compulsive symptoms, interpersonal relationship disorder symptoms, depression symptoms, anxiety symptoms, hostile emotion symptoms, fear symptoms, paranoia symptoms, and mental illness symptoms;

[0050] Acquire daily behavior data by presetting the first address set, the daily behavior data including: academic performance and changes, class attendance rate, physical test performance and changes, physiological function indicators and changes, campus entry and exit, canteen dining, dormitory entry and exit time distribution and dormitory time;

[0051] Acquire social relationship data by combining the preset second address set with the survey report, wherein the social relationship data includes family situation, family income situation, love situation, marriage situation and interpersonal communication situation;

[0052] The psychological survey data, the daily behavior data and the social relationship data are preprocessed to generate a target data set, wherein the preprocessing includes data cleaning, data standardization and data classification.

[0053] Optionally, the step of analyzing and learning the target data set by a machine learning algorithm to generate a large model for predicting mental illness includes: classifying the target data set according to the results of the psychological survey data, using a five-item scoring method to classify from six perspectives: depression, anxiety disorders, stress and stress disorders, eating disorders, sleep disorders, and addiction disorders, and marking no symptoms, mild symptoms, moderate symptoms, severe symptoms, and severe symptoms in each classification result as 0, 0.25, 0.5, 0.75, and 1, respectively;

[0054] Construct a recursive neural network RNN ​​model to extract data features F:

[0055]

[0056] Among them, R is the output result of the recurrent neural network RNN ​​model, and Represent the weight and bias of the fully connected layer respectively;

[0057] Define the loss function and use the Adam optimizer for model training;

[0058] Cross-validation is performed on a preset data set to adjust model parameters corresponding to the recursive neural network (RNN) model.

[0059] In the present invention, the data acquisition module processes the raw data to generate a target data set; the model training module analyzes and learns the target data set through a machine learning algorithm to generate a large model for predicting mental illness; the disease prediction module uses the large model to predict the mental illness conditions in the target data set and generates corresponding prediction values; the early warning interaction module receives the user's query request, and after determining the need to warn the user in combination with the prediction value, it performs a psychological condition warning; the feedback optimization module adjusts the model parameters corresponding to the large model according to the early warning accuracy report and the declaration record. By integrating the machine learning model and big data analysis, the system can effectively perform early warning of mental illness and provide accurate and in-depth judgment results. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 is a structural block diagram of a first embodiment of a mental illness early warning system of the present invention;

[0061] Figure 2 It is a flowchart diagram of a first embodiment of a method for early warning of mental illness according to the present invention.

[0062] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0063] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0064] Reference Figure 1 , an embodiment of the present invention provides a mental illness early warning system, Figure 1 This is a structural block diagram of a first embodiment of a mental illness early warning system of the present invention.

[0065] The mental illness early warning system in this embodiment is composed of a data collection module 10, a model training module 20, a disease prediction module 30, an early warning interaction module 40 and a feedback optimization module 50;

[0066] The data collection module 10 is used to process the raw data to generate a target data set, and send the target data set to the model training module 20. The raw data includes psychological survey data, daily behavior data, and social relationship data;

[0067] A model training module 20, for analyzing and learning a target data set through a machine learning algorithm to generate a large model for predicting mental illness;

[0068] The disease prediction module 30 is used to predict the mental illness conditions in the target data set using the large model and generate corresponding prediction values;

[0069] The early warning interaction module 40 is used to receive the query request of the user, and after determining the need to warn the user in combination with the predicted value, issue a psychological warning;

[0070] The warning interaction module 40 also includes: a warning display submodule 401, a warning confirmation submodule 402 and a warning reporting submodule 403;

[0071] The early warning display submodule 401 is used to display data visually according to the user's query request when performing psychological condition early warning;

[0072] The early warning confirmation submodule 402 is used to generate an early warning accuracy report based on the feedback information received from the user after the psychological condition early warning is performed;

[0073] The early warning reporting submodule 403 is used to report and generate a reporting record for the target object who has not received a psychological warning but has been diagnosed with a psychological disease;

[0074] The feedback optimization module 50 is used to adjust the model parameters corresponding to the large model according to the early warning accuracy report and the declaration record.

[0075] It should be noted that the data collection module is also used to collect standard psychological scale surveys and psychological interview data to generate psychological survey data, and uses text mining technology and natural language processing algorithms to extract psychological symptom information from psychological interview data. The psychological symptom information includes: physical discomfort symptoms, obsessive-compulsive symptoms, interpersonal relationship disorder symptoms, depression symptoms, anxiety symptoms, hostile emotion symptoms, fear symptoms, paranoia symptoms and mental illness symptoms; daily behavior data is obtained through the preset first address set, and the daily behavior data includes: academic performance and changes, class attendance rate, physical test scores and changes, physiological function indicators and changes, in and out of the campus, dining in the cafeteria, in and out of the dormitory time distribution and length of time in bed; social relationship data is obtained through the preset second address set combined with the survey report, and the social relationship data includes family situation, family income situation, love situation, marriage situation and interpersonal communication situation; data preprocessing is performed on the psychological survey data, daily behavior data and social relationship data to generate a target data set, and the preprocessing includes data cleaning, data standardization and data classification.

[0076] It should be noted that, in this embodiment, the first address set refers to daily behavior data obtained through the campus Internet of Things system and the educational affairs system, and the daily behavior data mainly corresponds to the behavior data of students.

[0077] It can be understood that the second address set in this embodiment refers to the student registration system and the student file system. The step of obtaining social relationship data by combining the preset second address set with the survey report refers to obtaining social relationship data about students through the student registration system and the student file system, while using survey interviews, combined with data and text processing systems, including family situation, family income situation, love situation, marriage situation and interpersonal communication situation.

[0078] It can be understood that the model training module is also used to classify the target data set according to the results of the psychological survey data, using a five-item scoring method to classify from six perspectives: depression, anxiety disorders, stress and stress disorders, eating disorders, sleep disorders, and addiction disorders. Asymptomatic, mild symptoms, moderate symptoms, severe symptoms, and severe symptoms are marked as 0, 0.25, 0.5, 0.75, and 1 in each classification result; a recursive neural network RNN ​​model is constructed to extract data features F:

[0079]

[0080] Among them, R is the output result of the recurrent neural network RNN ​​model, and Represent the weight and bias of the fully connected layer respectively;

[0081] Define the loss function and use the Adam optimizer for model training;

[0082] Cross-validation is performed on the preset data set to adjust the model parameters corresponding to the recurrent neural network RNN ​​model.

[0083] In the specific implementation, the model training module is also used for structured data processing (RNN part):

[0084] Use recursive neural networks (RNNs) to process structured data such as psychological survey data, daily behavior data, and social relationship data. RNNs can process sequence data through their loop structure and retain the time information of the data. For each time step t, the calculation formula of RNN is:

[0085]

[0086] in is the input at time step t, is the hidden state at time step t, , and b are network parameters. The final state Used to represent the information of the entire input sequence.

[0087] Loss function and optimization algorithm:

[0088] Loss function: Combining the cross-extraction loss (classification task) and the mean square error loss (regression task), for the classification task, the cross-extraction loss calculation formula is:

[0089]

[0090] in is the true label, is the predicted probability. For regression tasks, the mean square error loss is calculated as:

[0091]

[0092] in is the true value, is the predicted value.

[0093] Optimization algorithm: Use the Adam optimizer, which combines the advantages of momentum and adaptive learning rate. The update formula of the Adam optimizer is:

[0094]

[0095] Where θ is the parameter, η is the learning rate, and are the first-order and second-order moment estimates, corresponding to the mean and variance of the gradient, respectively, and ε is a small constant to prevent division by zero.

[0096] It should be noted that the early warning display submodule is also used to receive query information input by the user and determine the query type according to the keywords in the query information. The query model includes personal psychological early warning query and group psychological early warning query;

[0097] Match the corresponding disease prediction data according to the query type;

[0098] Generate early warning information based on disease prediction data and display the early warning information in a data visualization manner.

[0099] It should be noted that the early warning confirmation submodule specifically includes the user's investigation of the actual situation, feedback in the system, marking of the predicted results, confirmation of identifiable mental illness problems, and refutation of incorrect mental illness problems, which are recorded by the system.

[0100] The early warning reporting submodule specifically includes users reporting on college students who have not received early warnings but have been diagnosed with mental illness or suspected of having mental illness, and the system records the reporting status.

[0101] It can be understood that the human psychological early warning query includes a search with the individual as a keyword, presenting the individual search results in the form of a six-dimensional radar chart, and calculating the similarity between the user query vector q and the feature vector of each disease information in the disease prediction module;

[0102] The calculation formula for cosine similarity is:

[0103]

[0104] Where d represents the feature vector of a disease information in the disease prediction module.

[0105] In the specific implementation, the feedback optimization module specifically includes data feedback of early warning confirmation or declaration, monitoring system performance indicators, including query response time and error rate, identifying model parameters that affect query accuracy and response speed, and adjusting model parameters based on feedback, as follows:

[0106] If user feedback shows that the query results are not relevant enough, adjust the model weight update rules or increase the amount of training data;

[0107] Adjust the learning rate η: If the accuracy of the query results does not improve significantly, you can reduce the learning rate to avoid oscillations caused by too large a step size during the optimization process. Conversely, if the model converges too slowly, you can increase the learning rate appropriately.

[0108] Optimize the number of layers and neurons: Adjust the complexity of the model by increasing or decreasing the number of network layers and / or neurons. Too many layers or neurons may lead to overfitting, while too few may lead to underfitting.

[0109] Adjust activation function: According to model performance and training results, select appropriate activation function (such as ReLU, Sigmoid, Tanh, etc.) to improve the nonlinear fitting ability of the model.

[0110] Response speed optimization: Based on system performance data, especially response time, adjust the model to improve efficiency. For example, for slower response time, the inference speed can be accelerated by reducing model complexity or applying model compression technology.

[0111] Implement batch processing and parallel processing techniques to optimize the system's processing capabilities and reduce the average response time of a single query.

[0112] It also includes a user interface module, specifically including:

[0113] Intuitive interface layout:

[0114] Design a clear interface layout to facilitate users to quickly understand and use. Main functions such as search box, query button, result display area, etc. should be intuitively displayed on the main interface. For complex query results, such as images, tables or text, use appropriate layout and visualization techniques to make it easy for users to interpret. Design a responsive search box that supports natural language input. For query results, provide click to view details and swipe to browse more; integrate advanced search options to allow users to accurately customize queries, including filtering by class, college or relevance, and allow users to comment or provide feedback on query results.

[0115] The data acquisition module of this embodiment processes the raw data to generate a target data set; the model training module analyzes and learns the target data set through a machine learning algorithm to generate a large model for predicting mental illness; the disease prediction module uses the large model to predict the mental illness conditions in the target data set and generate corresponding prediction values; the early warning interaction module receives the user's query request, and after determining the need to warn the user in combination with the prediction value, it issues a psychological condition warning; the feedback optimization module adjusts the model parameters corresponding to the large model according to the early warning accuracy report and the declaration record. By integrating the machine learning model and big data analysis, the system can effectively provide early warnings for mental illnesses and provide accurate and in-depth judgment results.

[0116] Furthermore, the embodiment of the present invention provides a method for early warning of mental illness, referring to Figure 2 , Figure 2 It is a flow chart of a first embodiment of a method for early warning of mental illness according to the present invention.

[0117] In this embodiment, the mental illness early warning method includes the following steps:

[0118] Step S10: The data acquisition module processes the raw data to generate a target data set, and sends the target data set to the model training module. The raw data includes psychological survey data, daily behavior data, and social relationship data.

[0119] Step S20: The model training module analyzes and learns the target data set through a machine learning algorithm to generate a large model for predicting mental illness.

[0120] Step S30: The disease prediction module uses the large model to predict the mental illness conditions in the target data set and generates corresponding prediction values.

[0121] Step S40: the early warning interaction module receives the user's query request, and after determining that the user needs to be warned in combination with the predicted value, issues a psychological condition warning.

[0122] Step S50: When the early warning display submodule performs a psychological condition early warning, it performs a data visualization display according to the user's query request.

[0123] Step S60: The warning confirmation submodule generates a warning accuracy report based on the feedback information received from the user after the psychological condition warning is performed.

[0124] Step S70: The early warning reporting submodule reports and generates a reporting record for the target object who has not received a psychological warning but has been diagnosed with a mental illness.

[0125] Step S80: The feedback optimization module adjusts the model parameters corresponding to the large model according to the early warning accuracy report and the declaration record.

[0126] The data acquisition module of this embodiment processes the raw data to generate a target data set; the model training module analyzes and learns the target data set through a machine learning algorithm to generate a large model for predicting mental illness; the disease prediction module uses the large model to predict the mental illness conditions in the target data set and generate corresponding prediction values; the early warning interaction module receives the user's query request, and after determining the need to warn the user in combination with the prediction value, it issues a psychological condition warning; the feedback optimization module adjusts the model parameters corresponding to the large model according to the early warning accuracy report and the declaration record. By integrating the machine learning model and big data analysis, the system can effectively provide early warnings for mental illnesses and provide accurate and in-depth judgment results.

[0127] It should be understood that the above is only an example and does not constitute any limitation on the technical solution of the present invention. In specific applications, technicians in this field can make settings as needed, and the present invention does not limit this.

[0128] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of the present invention. In practical applications, technicians in this field can select part or all of them according to actual needs to achieve the purpose of the present embodiment, and no limitation is made here.

[0129] In addition, for technical details not fully described in this embodiment, reference can be made to the mental illness early warning system and method provided in any embodiment of the present invention, and will not be repeated here.

[0130] In addition, it should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0131] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0132] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as a read-only memory (ROM) / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0133] The above are only preferred embodiments of the present invention, and the protection scope of the present invention is not limited thereto. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the protection scope of the present invention.

Claims

1. A mental illness early warning system, characterized in that: The system includes: a data acquisition module, a model training module, a disease prediction module, an early warning interaction module and a feedback optimization module; The data acquisition module is used to process raw data to generate a target data set, and send the target data set to the model training module, wherein the raw data includes psychological survey data, daily behavior data, and social relationship data; The model training module is used to analyze and learn the target data set through a machine learning algorithm to generate a large model for predicting mental illness; The disease prediction module is used to predict the mental illness conditions in the target data set using the large model and generate corresponding prediction values; The early warning interaction module is used to receive a query request from a user, and after determining that a warning is needed for the user in combination with the predicted value, issue a psychological warning; The warning interaction module also includes: a warning display submodule, a warning confirmation submodule and a warning reporting submodule; The early warning display submodule is used to display data visually according to the query request of the user when performing psychological condition early warning; The warning confirmation submodule is used to generate a warning accuracy report based on the feedback information received from the user after the psychological condition warning is performed; The warning reporting submodule is used to report and generate reporting records for target subjects who have not received psychological warning but have been diagnosed with psychological diseases; The feedback optimization module is used to adjust the model parameters corresponding to the large model according to the early warning accuracy report and the declaration record; The model training module is also used to classify the target data set according to the results of the psychological survey data, using a five-item scoring method to classify from six perspectives: depression, anxiety disorders, stress and stress disorders, eating disorders, sleep disorders, and addiction disorders, and marking no symptoms, mild symptoms, moderate symptoms, severe symptoms, and severe symptoms in each classification result as 0, 0.25, 0.5, 0.75, and 1 respectively; Construct a recursive neural network RNN ​​model to extract data features F Among them, R is the output result of the recurrent neural network RNN ​​model, and Represent the weight and bias of the fully connected layer respectively; Define the loss function and use the Adam optimizer for model training; Performing cross validation on a preset data set to adjust model parameters corresponding to the recurrent neural network (RNN) model; The disease prediction module is also used to generate a mental illness prediction value through the trained large model parameters, where P is a 6-dimensional vector, and each component They represent depression, anxiety disorders, stress and stress disorders, eating disorders, sleep disorders, and addiction disorders; Warning value The calculation formula is: ,in represents the parameter matrix of the large model, Represents the data in the target data set.

2. The mental illness early warning system according to claim 1, characterized in that: The data collection module is further used to collect standard psychological scale survey and psychological interview data to generate the psychological survey data, and to extract psychological symptom information from the psychological interview data using text mining technology and natural language processing algorithms, wherein the psychological symptom information includes: physical discomfort symptoms, obsessive-compulsive symptoms, interpersonal relationship disorder symptoms, depression symptoms, anxiety symptoms, hostile emotion symptoms, fear symptoms, paranoia symptoms, and mental illness symptoms; Acquire daily behavior data by presetting the first address set, the daily behavior data including: academic performance and changes, class attendance rate, physical test performance and changes, physiological function indicators and changes, campus entry and exit, canteen dining, dormitory entry and exit time distribution and dormitory time; Acquire social relationship data by combining the preset second address set with the survey report, wherein the social relationship data includes family situation, family income situation, love situation, marriage situation and interpersonal communication situation; The psychological survey data, the daily behavior data and the social relationship data are preprocessed to generate a target data set, wherein the preprocessing includes data cleaning, data standardization and data classification.

3. The mental illness early warning system according to claim 1, characterized in that: The early warning display submodule is also used to receive query information input by the user and determine the query type according to the keywords in the query information, and the query type includes personal psychological early warning query and group psychological early warning query; Matching corresponding disease prediction data according to the query type; Early warning information is generated according to the disease prediction data and the early warning information is displayed in a data visualization manner.

4. The mental illness early warning system according to claim 3, characterized in that: The human psychological early warning query includes searching with individuals as keywords, displaying the personal search results in the form of a six-dimensional radar chart, and calculating the similarity between the user query vector q and the feature vector of each disease information in the disease prediction module; The calculation formula of cosine similarity is: Wherein d represents the characteristic vector of a certain disease information in the disease prediction module.

5. The mental illness early warning system according to claim 3, characterized in that: The group psychological early warning query includes searching with group name as keyword, displaying group search results in the form of six-dimensional radar chart, and calculating the overall data mean. Among them, x represents the data of all individuals of a certain indicator, and N represents the dimension of vector x.

6. A method for early warning of mental illness, characterized in that: The method is applied to the mental illness early warning system of claim 1, the system includes a data acquisition module, a model training module, a disease prediction module, an early warning interaction module and a feedback optimization module, the early warning interaction module includes an early warning display submodule, an early warning confirmation submodule and an early warning reporting submodule, the method includes: The data acquisition module processes the raw data to generate a target data set, and sends the target data set to the model training module, wherein the raw data includes psychological survey data, daily behavior data, and social relationship data; The model training module analyzes and learns the target data set through a machine learning algorithm to generate a large model for predicting mental illness; The disease prediction module uses the large model to predict the mental illness conditions in the target data set and generates corresponding prediction values; The early warning interaction module receives a query request from a user, and after determining that an early warning is required for the user in combination with the predicted value, issues an early warning for the psychological condition; When the early warning display submodule performs a psychological condition early warning, the early warning display submodule performs a data visualization display according to the query request of the user; The early warning confirmation submodule generates an early warning accuracy report based on feedback information received from the user after the psychological condition early warning is performed; The early warning reporting submodule reports and generates a reporting record for the target object who has not received a psychological warning but has been diagnosed with a psychological disease; The feedback optimization module adjusts the model parameters corresponding to the large model according to the early warning accuracy report and the declaration record.

7. The method for early warning of mental illness according to claim 6, characterized in that: The step of processing the raw data to generate a target data set and sending the target data set to the model training module comprises: Collecting standard psychological scale survey and psychological interview data to generate the psychological survey data, and extracting psychological symptom information from the psychological interview data using text mining technology and natural language processing algorithms, wherein the psychological symptom information includes: physical discomfort symptoms, obsessive-compulsive symptoms, interpersonal relationship disorder symptoms, depression symptoms, anxiety symptoms, hostile emotion symptoms, fear symptoms, paranoia symptoms, and mental illness symptoms; Acquire daily behavior data by presetting the first address set, the daily behavior data including: academic performance and changes, class attendance rate, physical test performance and changes, physiological function indicators and changes, campus entry and exit, canteen dining, dormitory entry and exit time distribution and dormitory time; Acquire social relationship data by combining the preset second address set with the survey report, wherein the social relationship data includes family situation, family income situation, love situation, marriage situation and interpersonal communication situation; The psychological survey data, the daily behavior data and the social relationship data are preprocessed to generate a target data set, wherein the preprocessing includes data cleaning, data standardization and data classification.

8. The method for early warning of mental illness according to claim 6, characterized in that: The step of analyzing and learning the target data set by a machine learning algorithm to generate a large model for predicting mental illness includes: classifying the target data set according to the results of the psychological survey data, using a five-item scoring method to classify from six perspectives: depression, anxiety disorders, stress and stress disorders, eating disorders, sleep disorders, and addiction disorders, and marking no symptoms, mild symptoms, moderate symptoms, severe symptoms, and severe symptoms in each classification result as 0, 0.25, 0.5, 0.75, and 1, respectively; Construct a recursive neural network RNN ​​model to extract data features F Among them, R is the output result of the recurrent neural network RNN ​​model, and Represent the weight and bias of the fully connected layer respectively; Define the loss function and use the Adam optimizer for model training; Cross-validation is performed on a preset data set to adjust model parameters corresponding to the recursive neural network (RNN) model.

Images