Psychological crisis comprehensive perception system and method based on AI and multi-source data

Abstract

The invention belongs to the technical field of psychological crisis comprehensive perception, and provides a psychological crisis comprehensive perception system and method based on AI and multi-source data. Comprising a multi-source data acquisition module, a data preprocessing module, a dynamic evaluation and early warning module, a notification information pushing module, an information confirmation feedback module, a self-training optimization module and a system management and configuration module. According to the invention, by integrating the daily behavior data of the students and the AI intelligent analysis technology, non-contact, non-sensitive and continuous monitoring of the psychological crisis of the students is realized, the limitation of a traditional subjective evaluation tool on time, space and crowd coverage is broken through, and the timeliness and accuracy of psychological crisis identification are remarkably improved. In the process of collecting and analyzing the student behavior data, the method strictly follows the Personal Information Protection Law of the People's Republic of China, all the data are subjected to de-identification processing, and the students know and agree, so that the privacy and legitimate rights and interests of the students are guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of comprehensive psychological crisis perception technology, and in particular to a comprehensive psychological crisis perception system and method based on AI and multi-source data. Background Art

[0002] With the development of society, students' mental health problems have become increasingly prominent, showing a new trend of multidimensionality, high frequency and concealment. However, traditional psychological assessment and screening work still mainly relies on subjective reports and periodic screening, which has obvious limitations: first, it has a strong lag, making it difficult to capture the dynamic changes in students' mental state in a timely manner; second, the coverage is narrow, and some high-risk but hidden students are not effectively identified; third, the assessment tools are not standardized enough and lack the ability to accurately distinguish different risk levels. The more prominent problem is that in actual work, many students who are truly in a state of psychological crisis do not appear on the warning list of routine psychological assessments, resulting in them not being included in the scope of key attention and intervention, missing the best time for early intervention.

[0003] Against this backdrop, achieving multidimensional assessment, accurate identification, dynamic monitoring, and scientific prevention of student psychological crises has become a core challenge for current campus mental health work. There is an urgent need to develop a scientific, real-time, and forward-looking psychological crisis monitoring system to shift psychological early warning from passive response to proactive prevention and control, thereby effectively improving the relevance and effectiveness of campus psychological services. In recent years, the advancement of big data analysis and artificial intelligence has made it possible to integrate AI with multi-source data to comprehensively perceive and monitor psychological crises. Summary of the Invention

[0004] The purpose of this invention is to provide a comprehensive psychological crisis perception system and method based on AI and multi-source data. By integrating multi-source data and combining machine learning algorithms, comprehensive perception monitoring, dynamic assessment and risk warning of psychological crises can be achieved to overcome the shortcomings of existing technologies.

[0005] In order to achieve the above-mentioned object of the invention, the present invention provides the following technical solutions:

[0006] A comprehensive psychological crisis perception system based on AI and multi-source data, including a multi-source data acquisition module, a data preprocessing module, a dynamic assessment and early warning module, a notification information push module, an information confirmation and feedback module, a self-training optimization module, and a system management and configuration module.

[0007] Preferably, the information confirmation feedback module includes the following functional units: an early warning information display unit, a disposal status marking unit, an intervention record entry unit, a timeout non-response reminder unit, a feedback data reflow unit, and a history record query unit.

[0008] Preferably, the self-training optimization module includes the following functional components: incremental learning engine, feedback-driven model tuning mechanism, sample screening and cleaning mechanism, model version control and rollback mechanism, model performance monitoring dashboard, multi-algorithm fusion mechanism, model retraining scheduler, and cross-campus / cross-grade adaptation mechanism.

[0009] Preferably, the multi-source data acquisition module is used to collect student-related multi-source heterogeneous data from various information systems of the school, such as the teaching management system, campus card system, dormitory management system, network authentication system, social platform logs, etc.

[0010] The data preprocessing module cleans, removes noise, fills in missing values, and standardizes the collected raw data, extracts key behavioral features that reflect changes in individual psychological states, and outputs them to the modeling and analysis module in a unified format;

[0011] The dynamic assessment and early warning module: establishes a psychological crisis early warning indicator system based on psychological theories and historical cases, and combines multiple machine learning algorithms such as supervised learning, semi-supervised learning, and reinforcement learning to build a psychological crisis early warning model;

[0012] The notification information push module classifies the generated warning information according to risk level and pushes it to the corresponding management personnel through the visual interface, SMS, email, mobile terminal app, etc.

[0013] The information confirmation and feedback module is used to receive confirmation and handling feedback from managers on early warning information, forming a closed-loop management system of "discovery-response-follow-up";

[0014] The self-training optimization module is used to continuously optimize the psychological crisis warning model, improve the model's prediction ability and timeliness, and form a closed-loop psychological safety management mechanism;

[0015] The system management and configuration module is used to configure system operating parameters such as data collection strategies, model parameters, warning rules, and user permissions.

[0016] The present invention also provides a method for comprehensive perception of psychological crisis based on AI and multi-source data, comprising the following steps:

[0017] (1) Multi-source data collection and access: Acquire multi-source heterogeneous data covering the entire process of students’ learning and life from various information systems of the school;

[0018] (2) Data preprocessing and feature extraction: The collected raw data are cleaned, denoised, and normalized, and key features reflecting the changing trends of students’ psychological states are extracted based on different data types to form standardized individual behavior portraits;

[0019] (3) Construction of a psychological crisis early warning indicator system: Based on psychological theory, expert experience and historical case analysis, a multi-dimensional psychological crisis early warning indicator system is established to form quantifiable and calculable risk assessment factors;

[0020] (4) Multimodal intelligent AI model training and optimization: The extracted behavioral characteristics and early warning indicators are input into the multimodal machine learning model, and the model is trained using a combination of supervised learning, semi-supervised learning, and reinforcement learning to continuously optimize the model's ability to identify different psychological risk levels. The model supports an incremental learning mechanism and can automatically update as new data is continuously accessed, thereby improving prediction accuracy and timeliness.

[0021] (5) Dynamic assessment and risk warning generation: The system dynamically assesses students’ psychological crises based on the model output results, generates individual and group psychological crisis risk level reports, and automatically generates warning information when potential psychological crisis signals are identified;

[0022] (6) Early warning information push and intervention linkage: The generated early warning information will be pushed to relevant management personnel, such as counselors, psychological counselors, etc. according to the risk level classification, and support linkage with the school psychological intervention mechanism to promote the formation of a closed-loop management system of "discovery-response-follow-up".

[0023] Preferably, the multi-source heterogeneous data includes multi-source data such as academic performance, living habits, social interactions, and network usage.

[0024] The present invention also provides a computer-readable storage device for realizing psychological crisis monitoring and early warning functions, including a USB flash drive, a portable hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic storage disk, a solid-state hard disk or an optical disk. The computer-readable storage medium for realizing psychological crisis monitoring and early warning functions stores computer instructions, and the computer instructions are used to enable the computer to execute the method described in the claims.

[0025] The present invention also provides a computer program product, which includes a computer program stored on a computer-readable storage medium for implementing psychological crisis monitoring and early warning functions. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer executes the method described in the claims.

[0026] Beneficial effects:

[0027] By integrating students' daily behavior data with AI intelligent analysis technology, this invention achieves non-contact, non-sensory, and continuous monitoring of students' psychological crises, breaking through the limitations of traditional subjective assessment tools in time, space, and population coverage, and significantly improving the timeliness and accuracy of psychological crisis identification. In the process of collecting and analyzing student behavior data, all data is de-identified and does not involve any personal identity identification information, effectively protecting students' privacy rights and interests. The system is equipped with a complete permission control mechanism and data encryption transmission solution to ensure the security and legality of data use. In addition, all students sign an informed consent form to clarify the scope and purpose of data use and safeguard their legal rights and interests. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flowchart of a method for comprehensive psychological crisis perception based on AI and multi-source data in Example 1;

[0029] Figure 2 This is a schematic diagram of the structure of a comprehensive psychological crisis perception system based on AI and multi-source data in Example 2. DETAILED DESCRIPTION

[0030] The present invention provides a system for comprehensive perception of psychological crises based on AI and multi-source data, including a multi-source data acquisition module, a data preprocessing module, a dynamic assessment and early warning module, a notification information push module, an information confirmation and feedback module, a self-training optimization module, and a system management and configuration module. It can efficiently integrate the school's existing information resources to achieve early detection and early intervention of students' psychological crises, and is used for comprehensive perception, dynamic monitoring, risk assessment and intelligent early warning of students' psychological crises.

[0031] In the present invention, the multi-source data acquisition module is used to collect student-related multi-source heterogeneous data from various information systems of the school, such as the teaching management system, campus card system, dormitory management system, network authentication system, social platform logs, etc.; the data covers multiple dimensions such as academic performance, attendance records, work and rest patterns, consumption behavior, frequency of Internet use, frequency of social interaction, etc., to construct a comprehensive basic data set for student behavior portraits.

[0032] In this invention, the multi-source behavioral data collection module requires authorization from the school before acquiring data and uses de-identification processing to retain only unique anonymous IDs to prevent exposure of real-life identity information. The system also has a built-in data access permission hierarchy, limiting access to warning results to authorized personnel such as counselors and psychological consultants. Regular teachers and administrative staff are unable to access raw behavioral data.

[0033] In the present invention, the data preprocessing module is used to clean, denoise, fill in missing values ​​and standardize the collected raw data, extract key behavioral characteristics that reflect changes in the individual's psychological state, and output them to the modeling and analysis module in a unified format; the characteristics include but are not limited to quantifiable indicators such as irregular work and rest, the rate of decline in social activity, and the learning motivation fluctuation index.

[0034] In this invention, the dynamic assessment and early warning module is used to establish a psychological crisis early warning indicator system based on psychological theory and historical cases. It combines multiple machine learning algorithms, such as supervised learning, semi-supervised learning, and reinforcement learning, to construct a psychological crisis early warning model. This model has incremental learning capabilities, can continuously optimize prediction accuracy based on new data, and can identify different levels of mental health risks (low risk, medium risk, and high risk). The psychological crisis early warning model runs in real time or on a scheduled basis to dynamically assess students' mental health status, generating individual risk scores and group-level trend reports. When the model identifies individuals exhibiting significant abnormal behavior patterns or reaching preset early warning thresholds, a psychological crisis early warning event is automatically triggered.

[0035] In the present invention, the notification information push module is used to classify the generated early warning information according to risk level, and push it to corresponding management personnel, such as counselors, psychological counselors, etc., through a visual interface, text messages, emails, mobile terminal apps, etc.; at the same time, it supports docking with the school psychological intervention system to realize the linkage closed loop between early warning information and subsequent psychological intervention processes.

[0036] In the present invention, the information confirmation and feedback module is used to receive confirmation and handling feedback of early warning information from management personnel, forming a closed-loop management system of "discovery-response-follow-up".

[0037] In the present invention, the self-training optimization module is used to continuously optimize the psychological crisis warning model, improve the model's prediction ability and timeliness, and form a closed-loop psychological safety management mechanism.

[0038] In the present invention, the system management and configuration module is used to configure system operating parameters such as data collection strategies, model parameters, warning rules, and user permissions, and supports flexible adjustment of warning models and indicator systems according to conditions such as different schools, stages, and grades, thereby improving the adaptability and scalability of the system.

[0039] The present invention also provides a method for comprehensive perception of psychological crises based on AI and multi-source data, comprising the following steps:

[0040] (1) Multi-source data collection and access: Acquire multi-source heterogeneous data covering the entire process of students’ learning and life from various information systems of the school, including academic performance, living habits, social interactions, network usage, etc.

[0041] (2) Data preprocessing and feature extraction: The collected raw data are cleaned, denoised, and normalized, and key features reflecting the changing trends of students’ psychological states are extracted based on different data types to form standardized individual behavior portraits;

[0042] (3) Construction of a psychological crisis early warning indicator system: Based on psychological theory, expert experience and historical case analysis, a multi-dimensional psychological crisis early warning indicator system is established to form quantifiable and calculable risk assessment factors;

[0043] (4) Multimodal intelligent AI model training and optimization: The extracted behavioral characteristics and early warning indicators are input into the multimodal machine learning model, and the model is trained using a combination of supervised learning, semi-supervised learning, and reinforcement learning to continuously optimize the model's ability to identify different psychological risk levels (low risk, medium risk, and high risk). The model supports an incremental learning mechanism and can be automatically updated as new data is continuously accessed, thereby improving prediction accuracy and timeliness.

[0044] (5) Dynamic assessment and risk warning generation: The system dynamically assesses students’ psychological crises based on the model output results, generates individual and group psychological crisis risk level reports, and automatically generates warning information when potential psychological crisis signals are identified;

[0045] (6) Early warning information push and intervention linkage: The generated early warning information will be pushed to relevant management personnel, such as counselors, psychological counselors, etc. according to the risk level classification, and support linkage with the school psychological intervention mechanism to promote the formation of a closed-loop management system of "discovery-response-follow-up".

[0046] The present invention also provides a computer-readable storage device for implementing psychological crisis monitoring and early warning functions. The computer-readable storage medium can be any form of non-volatile or removable storage medium, including, but not limited to, a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic storage disk, a solid-state drive, or an optical disk. The storage device stores executable program code. When the program is called and executed by a computing device, it can drive the computing device to execute the method described above. This implements operations related to big data-based psychological crisis identification, dynamic assessment, and early warning generation. The computer-readable storage medium for implementing psychological crisis monitoring and early warning functions stores computer instructions that cause the computer to execute the method described above.

[0047] The present invention also provides a computer program product. This product comprises a set of software modules executable on an electronic processing unit. When loaded into a device with data processing capabilities and executed, the software modules enable the device to continuously and comprehensively perceive and intelligently warn of psychological crises in a student population, in accordance with the technical solutions described in any embodiment of the present invention. The computer program product comprises a computer program stored on a computer-readable storage medium for implementing psychological crisis monitoring and early warning functions. The computer program comprises program instructions that, when executed by a computer, cause the computer to perform the described method.

[0048] The technical solutions provided by the present invention are described in detail below with reference to the embodiments, but they should not be construed as limiting the scope of protection of the present invention.

[0049] Example 1:

[0050] Figure 1 This is a flowchart of a method for comprehensive psychological crisis perception based on AI and multi-source data in Example 1 of the present invention.

[0051] Step S101: Multi-source data collection and access, which involves acquiring multi-source heterogeneous data covering the entire student learning and life process from various information systems of the school. The data includes multi-source data such as academic performance, living habits, social interactions, and network usage;

[0052] Specifically, in step S101, the school's information system includes but is not limited to the student information system, library management system, campus card consumption record system, dormitory access control system, network access control system, etc. Through API interface calls or database synchronization, information such as students' academic performance, attendance, book borrowing habits, consumption behavior, work and rest time, social platform activities and network usage habits is obtained. These data are uniformly stored in the data warehouse and preliminarily processed according to predetermined data cleaning rules. User identifiers of different systems are uniformly mapped to the standard student ID system to prepare for subsequent analysis. During the data collection process, the ETL process is used for data extraction, conversion and loading operations, and a data quality monitoring mechanism is designed to regularly check and repair data errors or inconsistencies to ensure data integrity and consistency.

[0053] Step S102: Data preprocessing and feature extraction: cleaning, denoising, and normalizing the collected raw data. Key features that reflect the changing trends of students' psychological states are extracted based on different data types to form standardized individual behavior profiles.

[0054] Specifically, in step S102, the original data is first cleaned to remove obviously erroneous or incomplete data entries. Then the noise and redundant information in the data are removed to improve the accuracy and reliability of the data and ensure the quality of the data. Next, for different types of variables (such as continuous and categorical types), appropriate normalization methods are applied, such as Min-Max scaling or Z-score standardization, so that all features have the same scale. Finally, based on psychological research and domain knowledge, key features that can reflect the changing trends of students' psychological states are extracted from the data, such as mood fluctuation indicators, learning efficiency index, social interaction frequency, etc. These features are quantified into numerical indicators for subsequent analysis. All features are output to the modeling and analysis module in a unified format to ensure the consistency of model training and form a personalized standardized behavioral portrait for each student.

[0055] Step S103: constructing a psychological crisis early warning indicator system. Based on psychological theory, expert experience, and historical case analysis, a multi-dimensional psychological crisis early warning indicator system is established to form quantifiable and calculable risk assessment factors.

[0056] Specifically, in step S103, a multi-dimensional psychological crisis early warning indicator system is designed based on psychological theoretical frameworks (such as the stress-coping model), expert experience, and historical case analysis results. This system not only covers internal factors such as cognitive function (attention concentration, memory ability) and emotional response (proportion of positive emotions, anxiety level), but also considers external environmental factors (quality of family relationships, level of peer support). Under each dimension, several specific quantifiable assessment factors are set, and the weight coefficient of each factor is determined through statistical methods to facilitate the calculation of a comprehensive risk score.

[0057] Step S104: Training and optimizing the psychological crisis warning model. The extracted behavioral characteristics and warning indicators are input into a multimodal machine learning model. The model is trained using a combination of supervised learning, semi-supervised learning, and reinforcement learning to continuously optimize the model's ability to identify different psychological risk levels (low risk, medium risk, and high risk). The model supports an incremental learning mechanism and can automatically update with the continuous access of new data, improving prediction accuracy and timeliness.

[0058] Specifically, in step S104, the collected student behavioral characteristics and psychological crisis warning indicators are used as input to select an appropriate machine learning algorithm (such as random forest, deep neural network, etc.) for model training. During the training process, in addition to using labeled historical data, a semi-supervised learning strategy is introduced to fully utilize unlabeled data resources. At the same time, in order to improve the adaptability and prediction accuracy of the model, a reinforcement learning mechanism is used to continuously adjust the parameter configuration. In addition, by setting a regular update plan, the model can automatically absorb the latest data to maintain its recognition ability and timeliness.

[0059] Step S105: Dynamic Assessment and Risk Warning Generation: The system dynamically assesses students' mental health status based on the model output, generates individual and group mental health risk level reports, and automatically generates warning information when potential psychological crisis signals are identified;

[0060] Specifically, in step S105, the system runs regularly, automatically generating individual and group-level risk reports based on the latest student behavioral data and psychological status assessment results. For high-risk individuals, the system immediately triggers an early warning signal, accompanied by detailed background analysis, helping administrators quickly understand the situation. For groups, macro-level trend analysis charts are provided, demonstrating changes in mental health status over a specific time period, facilitating the development of targeted preventive measures.

[0061] Step S106: Early warning information push and intervention linkage, the generated early warning information will be pushed to relevant management personnel, such as counselors, psychological counselors, etc. according to risk level classification, and support linkage with the school's psychological intervention mechanism to promote the formation of a "discovery-response-follow-up" closed-loop management system.

[0062] Specifically, in step S106, the warning information is sent to designated recipients through multiple channels (SMS, email, APP notification) to ensure the timeliness and accuracy of information transmission. Once the warning is received, relevant personnel can take appropriate intervention measures according to the guidance manual, such as arranging one-on-one psychological counseling sessions or organizing group counseling activities. The entire process emphasizes closed-loop management, that is, from problem discovery, immediate response to the final effect tracking, forming a complete work chain to ensure that every student in need can receive appropriate attention and support.

[0063] Example 2

[0064] Figure 2 This is a structural diagram of a comprehensive psychological crisis perception system based on AI and multi-source data in Example 2 of the present invention.

[0065] Step S201, multi-source data acquisition module: used to collect multi-source heterogeneous data related to students from various information systems of the school, such as the teaching management system, campus card system, dormitory management system, network authentication system, social platform logs, etc.; the data covers multiple dimensions such as academic performance, attendance records, work and rest patterns, consumption behavior, frequency of Internet use, frequency of social interaction, etc., to build a comprehensive basic data set for student behavior portraits.

[0066] Specifically, in step S201, the system connects to the school's various information systems through integrated interfaces, including but not limited to the teaching management system, campus card system, dormitory management system, network authentication system, and social platform logs. These systems provide students with raw data on academic performance, attendance records, daily routines, consumption behavior, frequency of Internet use, and frequency of social interactions. To ensure data integrity and consistency, the ETL (Extract, Transform, Load) process is used for data extraction, transformation, and loading operations. In addition, a data quality monitoring mechanism is designed to regularly check and fix data errors or inconsistencies to build a comprehensive basic dataset for student behavior portraits.

[0067] Furthermore, the module connects to the aforementioned subsystems through standardized interfaces (such as RESTful APIs, database connections, and log file parsing), and supports both scheduled and event-triggered data collection modes. For example, daily grade data from the teaching management system can be synchronized once a day via a scheduled task, while access control clock-in records are uploaded in real time using an event-triggered mechanism.

[0068] Furthermore, the module includes data quality monitoring capabilities, including outlier detection, duplicate record identification, and field integrity verification, ensuring the accuracy and consistency of collected data. All collected data is stored in a data warehouse or data lake and indexed by student ID, providing a high-quality input source for subsequent modeling and analysis.

[0069] Step S202, data preprocessing module: cleans, denoises, fills in missing values, and standardizes the collected raw data, extracts key behavioral features that reflect changes in the individual's psychological state, and outputs them to the modeling and analysis module in a unified format; the features include but are not limited to quantifiable indicators such as irregularity of work and rest, rate of decline in social activity, and learning motivation fluctuation index.

[0070] Specifically, in step S202, a series of preprocessing steps are first performed on the original data, including cleaning, denoising and missing value filling. The cleaning process involves deleting duplicates and correcting obvious data entry errors; denoising uses statistical methods to identify and exclude outliers; for missing values, average value filling, nearest neighbor interpolation or other appropriate strategies are adopted according to the field type. Next, based on the guidance of psychological theory, key behavioral features are extracted from the preprocessed data, such as irregular work and rest, social activity decline rate and learning motivation fluctuation index. These features are quantified into numerical indicators for subsequent analysis. All features are output to the modeling and analysis module in a unified format to ensure the consistency of model training. Finally, all features are converted into standardized vector form and stored in the feature database after adding timestamp information for dynamic evaluation and early warning module to call.

[0071] Step S203, Dynamic Assessment and Early Warning Module: A psychological crisis early warning indicator system is established based on psychological theory and historical cases. A multimodal intelligent AI analysis model is constructed by combining various machine learning algorithms, including supervised learning, semi-supervised learning, and reinforcement learning. This model has incremental learning capabilities, can continuously optimize prediction accuracy based on new data, and can identify different levels of mental health risks (low risk, medium risk, and high risk). The psychological crisis early warning model is run in real time or on a scheduled basis to dynamically assess students' mental health status, generating individual-level risk scores and group-level trend reports. When the model identifies an individual exhibiting significant abnormal behavior patterns or reaching a preset early warning threshold, a psychological crisis early warning event is automatically triggered.

[0072] Specifically, in step S203, based on a broad psychological theoretical framework and historical case studies, a psychological crisis early warning indicator system with multiple dimensions was developed. This system not only takes into account individual behavioral characteristics, but also includes the influence of environmental and social factors. Combining multiple machine learning techniques such as supervised learning (such as support vector machines), semi-supervised learning (such as label propagation algorithms) and reinforcement learning (such as Q-learning), a multimodal intelligent AI analysis model was developed. This model has incremental learning capabilities and can automatically adjust parameters and optimize prediction performance as new data is continuously added. At the same time, the model can identify different levels of mental health risks (low risk, medium risk, high risk) and provide corresponding risk scores.

[0073] The psychological crisis warning model runs regularly to continuously monitor and dynamically assess students' mental health. It not only generates individual risk scores but also provides group-level trend reports, helping education administrators understand overall mental health trends. When the model detects significant abnormal behavior patterns or reaches a preset warning threshold, it automatically generates a psychological crisis warning event and passes the relevant information to the next step, the warning information push and management module. This process emphasizes the importance of real-time responsiveness and personalized warning services.

[0074] Step S204, notification information push module: classify the generated warning information according to risk level, and push it to corresponding management personnel, such as counselors, psychological counselors, etc. through visual interface, SMS, email, mobile terminal App, etc.; at the same time, it supports docking with the school psychological intervention system to realize the linkage closed loop between warning information and subsequent psychological intervention process.

[0075] Specifically, in step S204, based on risk level classification, early warning information is accurately pushed to relevant management personnel, such as counselors and psychological consultants. Push information is delivered through various channels, including visual interfaces, text messages, emails, and mobile terminal apps, ensuring efficient and timely information delivery.

[0076] In the specific implementation, this module sets different push strategies according to the warning level. For example:

[0077] High-risk warning: Immediately sent to counselors and psychological consultants via SMS, app pop-up window, and email;

[0078] Medium-risk warning: This warning is sent to the counselor via app message or email, allowing them to confirm the response within a specified timeframe.

[0079] Low-risk warning: only displayed in the system background for regular screening and summary analysis.

[0080] The push content includes key information such as student basic information, risk level, behavioral characteristics summary, recommended intervention measures, etc., to help administrators quickly determine whether intervention is needed.

[0081] Furthermore, the module seamlessly integrates with the school's existing psychological intervention system, ensuring smooth flow of early warning information and fostering a closed-loop management mechanism of "discovery-response-follow-up," from early warning information to actual intervention measures. This facilitates rapid response to potential psychological crises and improves intervention effectiveness.

[0082] Step S205: Information confirmation feedback module

[0083] Specifically, the information confirmation and feedback module described in step S205 is used to receive feedback from managers on the processing status of the psychological crisis warning information generated by the system, and input the feedback data back to the model optimization module, thereby realizing a closed-loop management mechanism from warning to intervention to model update.

[0084] Specifically, the information confirmation and feedback module includes the following functional units:

[0085] Warning information display unit: Displays each warning message in a list or card format in a visual interface, including basic student information (such as name, student ID, and college), risk level, warning time, behavioral characteristics summary, recommended intervention measures, etc.

[0086] Disposal status marking unit: provides interactive buttons for managers to mark the status of warning information, including but not limited to options such as "viewed", "following up", "intervention completed", and "no action required";

[0087] Intervention record entry unit: supports administrators to upload relevant records of the intervention process, including psychological interview minutes, intervention measures details, student feedback, records of referral to the psychological counseling center, and other text or attachment materials;

[0088] Timeout and non-response reminder unit: Set a default response time limit (e.g., within 48 hours). If a high-risk warning message is not confirmed or processed, an escalation reminder mechanism will be automatically triggered, notifying the superior person in charge via SMS, email, or app;

[0089] Feedback data reflow unit: All confirmation status, intervention records, and treatment results are stored in a unified and structured manner, and serve as a supplementary data source for subsequent model training to improve model recognition accuracy;

[0090] Historical record query unit: supports retrieval of past warning and intervention records by multiple dimensions such as students, time range, risk level, and processing status, facilitating retrospective analysis and quality control by the school's mental health management department.

[0091] Furthermore, the information confirmation and feedback module also realizes data interoperability with third-party platforms such as the school's academic affairs system, student work management system, and psychological counseling case management system to ensure smooth flow of early warning information and a closed-loop controllable intervention process.

[0092] Furthermore, the system sets up a multi-role permission mechanism, and personnel in different positions (such as counselors, psychological counselors, etc.) can access the corresponding early warning information according to their scope of responsibilities, and fill in the corresponding handling opinions in the system to avoid information silos and shirking of responsibility.

[0093] Step S206: Self-training optimization module

[0094] Specifically, the self-training optimization module described in step S206 is used to continuously iterate and optimize the performance of the psychological crisis warning model, so that it can adapt to new changes in student behavior patterns, improve the accuracy and timeliness of model predictions, and form a dynamically evolving intelligent warning capability.

[0095] Specifically, the self-training optimization module includes the following functional components:

[0096] Incremental learning engine: Using an online learning mechanism, when the system receives new student behavior data, it automatically incorporates it into the model training set and performs partial updates based on the original model parameters, avoiding the waste of computing resources caused by retraining;

[0097] Feedback-driven model tuning mechanism: Utilizing the disposal records and intervention results collected in the information confirmation and feedback module, the feature weights in the model are re-evaluated. For example, if a certain behavioral feature (such as frequent nighttime internet use) is verified as an effective early warning signal in actual intervention, its importance in the model will be increased accordingly in the next round of training.

[0098] Sample screening and cleaning mechanism: Before each model update, new samples are tested for outliers and noise filtered to remove obvious false positives or invalid data, preventing the model from being misled by incorrect labels.

[0099] Model version control and rollback mechanism: A new version is generated for each model update, and historical versions are retained for comparison and analysis. If the new version performs poorly in the test environment, a one-click rollback to the previous stable version is supported;

[0100] Model performance monitoring dashboard: displays key evaluation indicators of the model in real time, including accuracy, recall, F1 score, AUC curve, false positive rate, false negative rate, etc., to help technicians understand the model operation status;

[0101] Multi-algorithm fusion mechanism: Integrates multiple machine learning algorithms (such as random forest, XGBoost, LSTM neural network, etc.) and adopts ensemble learning strategy for model fusion to improve overall prediction robustness;

[0102] Model retraining scheduler: Set periodic tasks (such as weekly / monthly) to automatically perform full model training, introducing the latest large-scale datasets to cope with seasonal changes in behavior patterns or changes in student population structure;

[0103] Cross-campus / cross-grade adaptation mechanism: Based on the differences in psychological characteristics of students from different schools and grades, it supports personalized configuration and transfer learning of model parameters, enhancing the system's generalization ability and applicability.

[0104] Furthermore, the self-training optimization module also has the ability to connect with external expert knowledge bases, allowing psychology experts to input new theoretical models or intervention experiences as auxiliary basis for model optimization, thereby realizing human-computer collaborative optimization.

[0105] Furthermore, the system has a built-in model interpretation tool that can visualize the basis for each warning decision, facilitating manual review and improving model transparency, thereby enhancing users' trust in AI decisions.

[0106] Step S207, system management and configuration module: used to configure system operating parameters such as data collection strategies, model parameters, warning rules, user permissions, etc., supporting flexible adjustment of warning models and indicator systems according to different schools, stages, grades and other conditions, to improve the adaptability and scalability of the system.

[0107] Specifically, in step S207, a system management and configuration module is specifically designed to enhance the system's flexibility and adaptability. This module allows administrators to configure key settings such as data collection strategies, model parameters, warning rules, and user permissions based on specific needs. For example, the warning model and its associated indicator system can be flexibly adjusted based on the characteristics of different schools, academic stages, or grade levels, ensuring the effectiveness and relevance of the warning mechanism. Furthermore, this module supports system performance monitoring and optimization, ensuring the stable operation and service quality of the entire comprehensive psychological crisis perception system.

[0108] The system management and configuration module is used to configure system operating parameters such as data collection strategies, model parameters, warning rules, and user permissions. It supports flexible adjustment of warning models and indicator systems based on different schools, stages, grades, and other conditions, thereby improving the system's adaptability and scalability.

[0109] Specifically, this module provides a graphical configuration interface, allowing administrators to freely define policies such as data collection frequency (such as real-time, daily, or weekly), collection field range, and data source access permissions. In terms of model parameter configuration, it supports changing the algorithm type, adjusting feature weights, modifying warning thresholds, and enabling / disabling incremental learning.

[0110] The module also supports a multi-role permission management mechanism, which can assign different levels of access control permissions to system administrators, college leaders, counselors, psychological counselors, etc. to ensure the security of sensitive data and compliance of use.

[0111] In addition, the module also has log auditing and operation recording functions, recording all key configuration changes and operation behaviors to facilitate subsequent review and responsibility tracing.

[0112] Example 3

[0113] This embodiment provides a computer-readable storage device for implementing psychological crisis monitoring and early warning functions. This storage device stores executable program code. When the program is called and executed by a computing device, it can drive the computing device to complete operations related to psychological crisis identification, dynamic assessment, and early warning generation based on big data.

[0114] Specifically, the computer-readable storage medium can be any form of non-volatile or removable storage carrier, such as but not limited to: USB flash drive, portable hard disk, read-only memory (ROM), random access memory (RAM), magnetic storage disk, solid-state hard disk or optical disk, etc., which can all serve as physical carriers for carrying the program instructions.

[0115] Example 4

[0116] This embodiment also provides a computer program product. This product comprises a set of software modules executable on an electronic processing unit. When these software modules are loaded into and executed on a device with data processing capabilities, the device can perform continuous, comprehensive perception and intelligent early warning of psychological crises in a student population, in accordance with the technical solutions described in any embodiment of the present invention.

[0117] The computer program product of the present invention includes multiple software modules, such as a multi-source data acquisition and access module, a data preprocessing and feature engineering module, a psychological crisis warning model module, a dynamic assessment and warning generation module, a warning information push and management module, a system management and configuration module, etc.

[0118] Each module has a clear division of functions. For example, the multi-source data acquisition and access module is used to collect multi-source heterogeneous student-related data from various information systems of the school, such as the teaching management system, campus card system, dormitory management system, network authentication system, social platform logs, etc. The data covers academic performance, attendance records, work and rest patterns, consumption behavior, frequency of Internet use, frequency of social interaction, and other dimensions to build a comprehensive basic data set for student behavior portraits.

[0119] The data preprocessing and feature engineering module cleans, denoises, fills in missing values, and standardizes the collected raw data, extracts key behavioral features that reflect changes in an individual's psychological state, and outputs them to the modeling and analysis module in a unified format; the features include but are not limited to quantifiable indicators such as irregular work and rest schedules, the rate of decline in social activity, and the learning motivation fluctuation index.

[0120] The psychological crisis warning model module establishes a psychological crisis warning indicator system based on psychological theories and historical cases, and combines multiple machine learning algorithms such as supervised learning, semi-supervised learning and reinforcement learning to build a multimodal intelligent AI analysis model; this model has incremental learning capabilities, can continuously optimize prediction accuracy based on new data, and can identify different levels of mental health risks (low risk, medium risk, high risk).

[0121] The dynamic assessment and warning generation module runs the psychological crisis warning model in real time or periodically, conducts dynamic assessments of students' mental health status, and generates individual-level risk scores and group-level trend reports; when the model identifies an individual with significant abnormal behavior patterns or reaches a preset warning threshold, it automatically triggers a psychological crisis warning event.

[0122] The early warning information push and management module classifies the generated early warning information according to risk levels, and pushes it to relevant management personnel, such as counselors, psychological counselors, etc., through visual interfaces, text messages, emails, mobile terminal apps, etc.; at the same time, it supports docking with the school's psychological intervention system to achieve a closed-loop linkage between early warning information and subsequent psychological intervention processes.

[0123] The system management and configuration module is used to configure system operating parameters such as data collection strategies, model parameters, warning rules, and user permissions. It supports flexible adjustment of warning models and indicator systems based on different schools, stages, grades, and other conditions, thereby improving the system's adaptability and scalability.

[0124] Program products can be deployed by distributing them to target devices in the form of installation packages. After installation, system administrators can configure various parameters as needed. Furthermore, program products support online updates and maintenance to ensure system up-to-dateness and security.

[0125] It can be seen from the above embodiments that the present invention provides a comprehensive psychological crisis perception system and method based on AI and multi-source data. By integrating students' daily behavior data with AI intelligent analysis technology, it realizes non-contact, non-sensitive and continuous monitoring of students' psychological crises, breaking through the limitations of traditional subjective evaluation tools in time, space and population coverage, and significantly improving the timeliness and accuracy of psychological crisis identification.

[0126] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A comprehensive psychological crisis perception system based on AI and multi-source data, characterized by: It includes multi-source data acquisition module, data preprocessing module, dynamic evaluation and early warning module, notification information push module, information confirmation feedback module, self-training optimization module, and system management and configuration module.

2. The psychological crisis comprehensive perception system based on AI and multi-source data according to claim 1 is characterized in that: The information confirmation and feedback module includes the following functional units: an early warning information display unit, a disposal status marking unit, an intervention record entry unit, a timeout and no response reminder unit, a feedback data reflow unit, and a history record query unit.

3. The psychological crisis comprehensive perception system based on AI and multi-source data according to claim 2 is characterized in that: The self-training optimization module includes the following functional components: incremental learning engine, feedback-driven model tuning mechanism, sample screening and cleaning mechanism, model version control and rollback mechanism, model performance monitoring dashboard, multi-algorithm fusion mechanism, model retraining scheduler, and cross-campus / cross-grade adaptation mechanism.

4. The psychological crisis comprehensive perception system based on AI and multi-source data according to claim 1 is characterized in that: The multi-source data acquisition module is used to collect student-related multi-source heterogeneous data from various information systems of the school, such as the teaching management system, campus card system, dormitory management system, network authentication system, social platform logs, etc. The data preprocessing module cleans, removes noise, fills in missing values, and standardizes the collected raw data, extracts key behavioral features that reflect changes in individual psychological states, and outputs them to the modeling and analysis module in a unified format; The dynamic assessment and early warning module: establishes a psychological crisis early warning indicator system based on psychological theories and historical cases, and combines multiple machine learning algorithms such as supervised learning, semi-supervised learning, and reinforcement learning to build a psychological crisis early warning model; The notification information push module classifies the generated warning information according to risk level and pushes it to the corresponding management personnel through the visual interface, SMS, email, mobile terminal app, etc. The information confirmation and feedback module is used to receive confirmation and handling feedback from managers on warning information, forming a closed-loop management system of "discovery-response-follow-up"; The self-training optimization module is used to continuously optimize the psychological crisis warning model, improve the model's prediction ability and timeliness, and form a closed-loop psychological safety management mechanism; The system management and configuration module is used to configure system operating parameters such as data collection strategies, model parameters, warning rules, and user permissions.

5. A comprehensive psychological crisis perception method based on AI and multi-source data, characterized by: The following steps are involved: (1) Multi-source data collection and access: Acquire multi-source heterogeneous data covering the entire process of students’ learning and life from various information systems of the school; (2) Data preprocessing and feature extraction: The collected raw data are cleaned, denoised, and normalized, and key features reflecting the changing trends of students’ psychological states are extracted based on different data types to form standardized individual behavior portraits; (3) Construction of a psychological crisis early warning indicator system: Based on psychological theory, expert experience and historical case analysis, a multi-dimensional psychological crisis early warning indicator system is established to form quantifiable and calculable risk assessment factors; (4) Multimodal intelligent AI model training and optimization: The extracted behavioral characteristics and early warning indicators are input into the multimodal machine learning model, and the model is trained using a combination of supervised learning, semi-supervised learning, and reinforcement learning to continuously optimize the model's ability to identify different psychological risk levels. The model supports an incremental learning mechanism and can automatically update as new data is continuously accessed, thereby improving prediction accuracy and timeliness. (5) Dynamic assessment and risk warning generation: The system dynamically assesses students’ psychological crises based on the model output results, generates individual and group psychological crisis risk level reports, and automatically generates warning information when potential psychological crisis signals are identified; (6) Early warning information push and intervention linkage: The generated early warning information will be pushed to relevant management personnel, such as counselors, psychological consultants, etc. according to the risk level classification, and support linkage with the school's psychological intervention mechanism to promote the formation of a closed-loop management system of "discovery-response-follow-up".

6. The method for comprehensive psychological crisis perception based on AI and multi-source data according to claim 5 is characterized in that: The multi-source heterogeneous data includes multi-source data such as academic status, living habits, social interactions, and network usage.

7. A computer-readable storage device for implementing psychological crisis monitoring and early warning functions, characterized in that: The computer-readable storage medium for implementing the psychological crisis monitoring and early warning function includes a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic storage disk, a solid-state hard drive or an optical disk. The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable the computer to execute the method described in any one of claims 1 to 6.

8. A computer program product, characterized in that The computer program product includes a computer program stored on a computer-readable storage medium for implementing psychological crisis monitoring and early warning functions. The computer program includes program instructions. When the program instructions are executed by a computer, the computer executes the method according to any one of claims 1 to 6.

Images