Student psychological evaluation system based on AI reinforcement learning optimization

Abstract

The invention belongs to the field of student psychological health management, and particularly relates to a student psychological evaluation system based on AI reinforcement learning optimization. According to the system, the psychological state of the student is dynamically evaluated through multi-modal data acquisition, deep data processing and reinforcement learning optimization, and the problems of singleness and hysteresis in a traditional psychological evaluation method are solved. The system comprises a data acquisition layer, a data processing layer, a reinforcement learning optimization layer, a gamification evaluation module, a feedback and intervention layer and a user interface layer. The data acquisition layer acquires behavior, physiological and social data in multiple channels; the data processing layer performs cleaning, feature extraction and fusion on the data; the reinforcement learning optimization layer adjusts the psychological evaluation strategy in real time; the gamification evaluation module enhances the student participation degree through a task and reward mechanism; and the feedback and intervention layer provides real-time feedback and personalized intervention suggestions. The psychological evaluation accuracy, participation and intervention effect can be improved, and the method is suitable for psychological health management scenes of schools and related educational institutions.

Description

Technical Field

[0001] The present invention belongs to the technical field of student mental health management, and specifically relates to a student psychological evaluation system optimized based on AI reinforcement learning. Background Art

[0002] In recent years, with the continuous deepening of education and the increasingly prominent personalized needs of students, accurate and dynamic evaluation of students' comprehensive qualities and mental health has become a research hotspot in the education field. Traditional student quality evaluations often rely on paper or online questionnaires, or analyze through single-dimensional grades and activity records. This method lacks comprehensive consideration of multi-faceted elements such as students' emotions, stress management abilities, and social interaction levels. To solve this problem, many innovative solutions have been proposed in the industry. For example, in the "Student Comprehensive Quality Evaluation System" with patent application number 202011049371.0, an idea of using multiple evaluation indicators to comprehensively measure students' qualities is proposed, and through data collection and result analysis, relevant suggestions for students' learning and growth are provided for teachers and parents.

[0003] However, after retrieving and analyzing the "Student Comprehensive Quality Evaluation System" with patent application number 202011049371.0, it is found that this system mainly focuses on weighted statistical evaluation of students' multi-dimensional performances (such as academics, morality, sports, etc.), lacking in-depth exploration and personalized adaptation in mental health assessment; especially in data collection, more attention is paid to index statistics and linear analysis methods, while advanced technologies such as deep reinforcement learning, multi-modal data fusion, and gamification feedback have not been systematically designed and applied, making it difficult to monitor students' emotional states and potential psychological pressures in real time and perform adaptive interventions. In addition, its evaluation method is relatively traditional, and students are prone to resistance during the process of high-frequency and one-way data filling and passive acceptance of feedback, resulting in the evaluation results being difficult to fully reflect students' true psychological conditions, thereby reducing the accuracy and timeliness of the overall evaluation. Summary of the Invention

[0004] The present invention provides a student psychological evaluation system optimized based on AI reinforcement learning to solve the problems raised in the above background art.

[0005] To achieve the above object, the present invention provides the following technical solution: A student psychological evaluation system optimized based on AI reinforcement learning, the system includes;

[0006] A data collection layer for obtaining students' behavioral data, physiological data, social interaction data, and environmental context data from multiple platforms and channels;

[0007] The data processing layer, connected to the data acquisition layer, is used for preprocessing the acquired data, data cleaning, feature extraction, and multi-modal data fusion;

[0008] The reinforcement learning optimization layer, connected to the data processing layer, is used to execute the deep reinforcement learning algorithm to adaptively optimize the psychological evaluation strategy;

[0009] The gamification evaluation module, in two-way interaction with the reinforcement learning optimization layer, is used to provide students with a variety of interactive games, task challenges, reward mechanisms, and scenario simulations during the psychological assessment process, so that students can generate the data required for psychological state assessment while completing the games and tasks;

[0010] The feedback and intervention layer, connected to the gamification evaluation module and the reinforcement learning optimization layer, is used to provide instant feedback to students, teachers, and psychological counselors based on real-time evaluation results, and trigger intervention measures when potential psychological problems are detected;

[0011] The user interface layer, connected to the feedback and intervention layer, is used to provide functions of data visualization, evaluation reports, intervention prompts, and personalized settings for students, teachers, and psychological counselors;

[0012] Among them, the reinforcement learning optimization layer dynamically adjusts the task design and content of the gamification evaluation module according to the real-time status and historical data of different student groups and individuals, improving the accuracy and effectiveness of psychological evaluation.

[0013] Preferably, the reinforcement learning optimization layer adopts a reinforcement learning algorithm based on the deep Q-network, which updates the action value function Q(s,a) based on the following formula every time a task is assigned and feedback is given;

[0014]

[0015] Among them;

[0016] s represents the current psychological and behavioral state of the student, and a represents the selected gamification evaluation task;

[0017] α is the learning rate, and γ is the discount factor;

[0018] r represents the immediate reward obtained based on the student's feedback and the system goal after the execution of the gamification evaluation task;

[0019] s′ and a′ respectively represent the state and action at the next moment;

[0020] Through repeated iterative learning, the Q(s,a) approaches the optimal value function, thereby realizing adaptive task allocation and evaluation optimization for different students.

[0021] Preferably, the gamification evaluation module includes;

[0022] An emotion recognition game unit for evaluating students' emotion recognition ability and emotional expression tendency by recognizing and expressing different emotion icons, story scenarios, and role-playing scenarios;

[0023] A stress management game unit for simulating and setting various stress scenarios and time urgency challenges, measuring indicators such as students' heart rate, breathing rate, and reaction duration during task execution, and evaluating students' stress tolerance and emotion regulation skills;

[0024] A social interaction simulation unit for collecting data and evaluating students' social skills, interpersonal interaction preferences, and communication strategies through virtual social scenarios and multi-player collaborative game tasks;

[0025] A task reward mechanism unit for giving students positive feedback in the form of points, badges, and leaderboards after they complete stage goals and challenges, enhancing participation and durability;

[0026] Among them, the multi-dimensional data collected by each game unit is transmitted to the data processing layer in real time for fusion, providing an optimization basis for the reinforcement learning optimization layer.

[0027] Preferably, the stress management game unit further includes;

[0028] A scenario simulation sub-module that constructs virtual scenarios similar to real campus life and social scenarios, providing diverse and progressive task pressures;

[0029] A physiological signal acquisition sub-module that obtains students' heart rate variability, facial expression changes, and hand movement characteristics through wearable devices, cameras, and mouse trajectory recording methods;

[0030] A stress score calculation sub-module that generates a comprehensive stress score corresponding to students' task completion degree, time consumption, and error rate indicators in the scenario, and interacts with the reinforcement learning optimization layer to adaptively regulate the game difficulty and feedback mechanism.

[0031] Preferably, the data processing layer includes;

[0032] A multi-modal data fusion module for feature extraction and weighted synthesis of behavior data, physiological data, and social interaction data;

[0033] A data cleaning module for removing incomplete data, abnormal data, and noise interference, ensuring that the data input to the reinforcement learning optimization layer has high confidence;

[0034] A feature selection and dimensionality reduction module, which is used to reduce the dimensions of multi-modal features by using the methods of correlation analysis and principal component analysis, reduce redundant dimensions, and avoid the risk of overfitting;

[0035] A real-time data monitoring module, which is used to monitor and preliminarily analyze the massive data collected, and provide necessary processing instructions and risk warnings for the feedback and intervention layer.

[0036] Preferably, the feedback and intervention layer includes;

[0037] An instant feedback module, which generates a visual report and text suggestions based on the performance of students in the gamification evaluation module and the model prediction results provided by the reinforcement learning optimization layer, and displays them to students and teachers in real time;

[0038] A multi-level intervention trigger module, which provides differential support to students according to the intervention level. When the system detects that a student has significant mental health risks, it automatically sends a high-priority warning message and recommends professional psychological counseling;

[0039] A resource recommendation module, which recommends online psychological courses, offline counseling activities, and self-help training plans that match the student's stress type and emotional characteristics of social preferences;

[0040] A notification scheduling module, which timely sends risk information with a higher severity level to designated teachers and psychologists for convenient follow-up and offline intervention in a timely manner.

[0041] Preferably, the user interface layer includes;

[0042] A student-side interface, which supports multi-platform logins, provides task entrances, personal data visualization, reward mechanism displays, and psychological self-assessment and help request channels;

[0043] A teacher-side interface, which provides a class overview of students' psychological assessment results, individual in-depth reports, intervention suggestions, and task management tools for teachers to adjust teaching and management strategies in a timely manner;

[0044] A psychologist-side interface, which provides professional visual charts of psychological assessment data, multi-dimensional comparative analysis, intervention record management, and student case tracking functions;

[0045] A management background interface, which allows system administrators to perform operations such as permission allocation, policy adjustment, data backup and recovery, and privacy compliance configuration.

[0046] Preferably, a student psychological evaluation method based on AI reinforcement learning optimization includes the following steps;

[0047] Step S1; Data collection, obtaining multi-modal data of students from mobile applications, web pages, and third-party platforms;

[0048] Step S2; Data preprocessing and fusion, cleaning, normalizing, and feature extracting the collected data, and fusing to obtain the current comprehensive psychological state features of the student;

[0049] Step S3; Gamification task assignment, adaptively selecting and generating personalized gamification evaluation tasks by the reinforcement learning optimization layer according to the comprehensive psychological state features of the student;

[0050] Step S4; Task execution and behavior collection, the student completes the assigned tasks in the gamification evaluation module, and simultaneously collects their behavior performance and physiological signals in real time;

[0051] Step S5; Reinforcement learning policy update, based on real-time feedback and the collected data, iteratively updating the task assignment policy through a deep reinforcement learning algorithm;

[0052] Step S6; Result feedback and intervention, generating a psychological assessment report for the student, automatically triggering an intervention mechanism for the detected high-risk students, and pushing the corresponding information to teachers and psychological counselors.

[0053] Preferably, when fusing multi-modal data, the method adopts a fusion formula based on weighted coefficients;

[0054] F fusion = ω 1 F behavior + ω 2 F physiology + ω 3 F social

[0055] Where;

[0056] F behavior represents the operation behavior feature vector of the student in the gamification task;

[0057] F physiology represents the feature vector of the student's physiological data;

[0058] F social represents the data feature vector related to the student's social interaction;

[0059] ω 1 , ω 2 , ω 3 , are respectively the weighted coefficients used to balance the weights of each feature vector in the fusion result, satisfying ω 1 + ω 2 + ω 3 = 1;

[0060] By dynamically adjusting ω 1 , ω 2, ω 3 , realizing the flexible capture and evaluation of students' psychological changes at different times.

[0061] Preferably, it also includes;

[0062] Privacy protection: Anonymize students' personal identities during data collection and transmission, and use encryption algorithms to ensure the security of data transmission between the server and the client;

[0063] Access control: Set different permissions for users at all levels of the system to ensure that only authorized teachers, psychologists, and administrators can access students' detailed psychological data;

[0064] Data storage: Use a distributed storage method for data backup and redundancy, and perform secondary encrypted storage on sensitive data to avoid data leakage;

[0065] Compliance audit: Regularly audit and evaluate the data process of the system to ensure that the utilization and processing of students' psychological data comply with relevant laws and regulations as well as the compliance requirements of schools and educational institutions.

[0066] The beneficial effects of the present invention are;

[0067] 1. The present invention adopts a deep reinforcement learning algorithm to perform personalized optimization configuration on the psychological evaluation task according to students' historical data and real-time behavior characteristics, enabling the system to dynamically adapt to different students' psychological states and behavior characteristics, improving the accuracy and effectiveness of psychological evaluation results. Compared with traditional static evaluation methods, the present invention can update strategies in real time, avoiding the problem of insufficient accuracy caused by a single or lagging evaluation model.

[0068] 2. The present invention conducts psychological evaluation tasks through a gamified evaluation module in ways such as interactive games, reward mechanisms, and scenario simulations, effectively enhancing students' interest and initiative in participating in the evaluation. At the same time, the gamified design reduces students' psychological defenses and stress responses in traditional questionnaire evaluations, making the multi-modal behavior data collected more authentic and representative, providing a reliable data basis for subsequent in-depth analysis.

[0069] 3. The present invention can quickly generate warning signals when detecting potential psychological risks by real-time monitoring of students' psychological state changes and combining with the dynamic analysis ability of the reinforcement learning model, and provide personalized intervention suggestions according to the specific situation of students. Compared with the lag of traditional evaluation methods, the present invention can trigger intervention at the budding stage of psychological problems, helping schools and parents formulate support strategies early and reducing the possibility of mental health problems escalating. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] Figure 1 is a block diagram of the working process of the present invention;

[0071] Figure 2 This is the overall system architecture block diagram of the present invention;

[0072] Figure 3 This is the block diagram of the data acquisition layer of the present invention;

[0073] Figure 4 This is the block diagram of the data processing layer of the present invention;

[0074] Figure 5 This is the block diagram of the reinforcement learning optimization layer and the gamification evaluation module of the present invention;

[0075] Figure 6 This is the block diagram of the feedback and intervention layer and the user interface layer of the present invention

[0076] Figure 7 This is the flow chart of the feedback and intervention layer and the user interface layer of the present invention. Detailed implementation manners

[0077] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0078] See Figures 1 to 7 , a student psychological evaluation system based on AI reinforcement learning optimization provided by the present invention aims to achieve dynamic, accurate and personalized assessment of students' psychological states through multi-modal data acquisition and fusion, deep reinforcement learning strategy optimization, and gamification evaluation methods. This system mainly consists of the following parts: data acquisition layer, data processing layer, reinforcement learning optimization layer, gamification evaluation module, feedback and intervention layer, user interface layer, and system physical layout.

[0079] This system can be deployed on the private servers of schools or educational institutions, or can be deployed with the help of cloud platforms. Students log in and interact through mobile terminals, PC terminals or campus self-service devices, and all data is encrypted and transmitted to the server side for calculation and analysis. Teachers and psychological counselors can obtain the assessment results of students' psychological states and intervention suggestions through a dedicated management interface or APP.

[0080] See Figures 1 to 7 , in some practical applications, the system components and function divisions are as follows;

[0081] 1. Data acquisition layer; collect students' behavioral data, physiological data, social interaction data, and external environment data, etc. through various channels to provide raw data support for subsequent analysis.

[0082] 2. Data processing layer; including data cleaning module, multi-modal data fusion module, feature selection and dimensionality reduction module, etc., which standardize, remove noise and reduce the dimensionality of high-dimensional data from the original data collected in the previous step, so as to obtain high-confidence feature vectors as the input of the reinforcement learning algorithm and gamification task design.

[0083] 3. Reinforcement learning optimization layer; adopting the deep reinforcement learning algorithm to adaptively adjust the task assignment and evaluation strategies for different students. This layer generates immediate rewards or punishments according to the performance of students in the gamification evaluation module and historical behavior records, and updates the policy parameters to continuously optimize the task assignment and evaluation process.

[0084] 4. Gamification evaluation module; including multiple functional units, such as emotion recognition game unit, stress management game unit, social interaction simulation unit and task reward mechanism unit. By testing students' coping styles, facial expression changes, response speed and accuracy, etc. in the game scenario, multi-dimensional psychological state information is collected.

[0085] 5. Feedback and intervention layer; based on the collected evaluation data and the prediction results of reinforcement learning, generate personalized immediate feedback reports for students; when detecting high-risk psychological states, immediately trigger intervention measures and notify teachers, psychologists or parents to take corresponding psychological counseling means.

[0086] 6. User interface layer; provide friendly interfaces for students, teachers and psychologists to log in and use respectively. The student side can view personal growth curves and reward lists; the teacher side can browse the overall mental health map of the class and master the psychological warning information of each student; the psychologist side provides functions of in-depth data visualization and intervention tracking records.

[0087] Example 1

[0088] See Figures 1 to 7 , in some practical applications, the system process and operation method are as follows;

[0089] 1. Data collection and preprocessing

[0090] Students first log in to this system using their registered accounts and grant a certain degree of data reading permissions on the mobile or PC side. The system performs data collection and preprocessing through the following steps;

[0091] Data collection: Record students' click behaviors, mouse movement trajectories, answering times, operation accuracies, etc. during gamified tasks. Obtain students' heart rate changes, facial expression key points, and skin conductance responses, etc. through the camera of the student device, wearable bracelet, or other sensors. Statistically analyze students' language communications, opinion interactions, and friendship relationship structures, etc. in virtual social scenarios or group tasks.

[0092] Data cleaning: Delete duplicate, distorted, or severely missing data records. Calibrate timestamps, sensor acquisition frequencies, etc. to ensure the temporal consistency between data.

[0093] Multimodal fusion and dimensionality reduction: Synthesize features of each dimension through a weighted fusion formula. For example;

[0094] F fusion = ω 1 F behavior + ω 2 F physiology + ω 3 F social

[0095] where ω 1 + ω 2 + ω 3 = 1. The system can dynamically adjust each ω value according to the psychological focus points of different students or different stages. Adopt principal component analysis or other dimensionality reduction methods for high-dimensional features to reduce feature redundancy and improve calculation efficiency.

[0096] 2. Gamification evaluation and data interaction

[0097] Gamification module startup: After the student logs in, the system automatically matches appropriate gamified tasks for them, including emotion recognition, stress management, and social interaction simulation, etc. The system provides a short new user guide for new users to reduce the threshold and psychological pressure for the first use.

[0098] During the game process: For the emotion recognition game unit; the student needs to quickly identify the facial expressions of characters in the picture or simulate the expression methods of different emotions in the virtual scene, and the system records according to the recognition accuracy and response duration. For the stress management game unit; the system sets up various scenarios, such as countdown exams, emergency task handling, etc., and real-time collects the students' heart rates, facial expressions, and tone changes; and generates a stress score according to the task completion situation. For the social interaction simulation unit; introduce a virtual social environment, let the student interact with virtual characters or other students in specific topics or cooperative tasks, and evaluate the social ability and emotion management level through language expression, opinion exchange, and teamwork.

[0099] Task Reward Mechanism: When students complete tasks at different stages, they will receive points, badges, or leaderboard rankings as positive incentives. The reward mechanism can maintain students' interest while enabling them to voluntarily and actively participate in more assessment tasks, creating conditions for the system to collect more comprehensive data.

[0100] 3. Optimization of Reinforcement Learning Strategy

[0101] State Definition: The comprehensive definition of the student's mental state, historical task completion, current game scenario, and system evaluation model is used as the state s for reinforcement learning.

[0102] Action Selection: The actions a that the system can choose include specific forms such as assigning emotion recognition tasks, stress management tasks, or social interaction simulation tasks.

[0103] Reward Mechanism: When the student completes the task, the system will give an immediate reward r based on their performance, subsequent feedback, and target optimization metrics.

[0104] Value Function Update: During the learning process, the system uses a deep Q-network to update the value function Q(s, a), and the formula is as follows;

[0105]

[0106] where α is the learning rate, γ is the discount factor, s ′ is the next state, and a ′ is the available action at the next moment.

[0107] Adaptive Optimization: After multiple rounds of iteration, the system gradually converges, enabling it to adaptively adjust the assigned task types and difficulties according to the personalities, mental states, and participation willingness of different students, in order to improve the efficiency and accuracy of assessment.

[0108] 4. Feedback and Intervention Strategies

[0109] Immediate Feedback: After completing the game task, the student can obtain a personal performance report, including emotion recognition accuracy, stress management scores, and social interaction evaluations; the interface will give friendly tips and improvement suggestions.

[0110] Multi-level Intervention: When the system detects that a student may have psychological abnormalities, it will send warnings to teachers or counselors at different levels. In case of serious situations, the system will issue a red alert and recommend one-on-one offline counseling or further medical evaluation.

[0111] Resource Recommendation: The system can recommend some stress-relieving mini-games, stress reduction skill training, or activities to participate in with peers according to the student's stress type, interests, or social preferences. For students who need regular psychological counseling, a psychological learning plan can be automatically generated and tracked and adjusted in combination with the progress feedback from the system.

[0112] 5. User Interface Design

[0113] Student side; Provide simple and intuitive task entrances, clear goal guidance, and visual emotion and stress curves so that students can independently understand their psychological change trends; The integral, badges, and leaderboards of the reward system can be visually displayed to motivate students to actively participate; Provide "psychological self-test" and "help-seeking" entrances to guide students to conduct psychological counseling in a timely manner or seek help from teachers.

[0114] Teacher side; Can view the overall mental health status map of the class at a glance, and view the mental score trends and intervention records of each student; When it is detected that a certain student shows an abnormal trend, their past data can be quickly retrieved for targeted analysis, and individual conversations or referrals to psychological counseling institutions can be triggered.

[0115] Psychologist side; Support viewing multi-dimensional data summaries and visual analysis, and can track the intervention effects; The system retains the time series data of multiple psychological evaluations of students, providing data support for psychologists to deeply analyze the regular patterns of students' phased psychological changes; Through the intervention record module, the gamification task content, intervention time period, and follow-up methods of students can be adjusted to facilitate case management.

[0116] System management background; Allow administrators to configure data access permissions and implement multi-level role management; Administrators can update and maintain the task library in the gamification evaluation module, or appropriately adjust the hyperparameters of the reinforcement learning algorithm.

[0117] Example Two

[0118] See Figures 1 to 7 In some practical applications, the deep fusion and adaptive intervention of multi-modal data are as follows;

[0119] To further illustrate the application effect of the present invention in a complex teaching environment, the following takes a typical scenario as an example; In the first grade of a certain high school, the student group includes both academic students with excellent grades but low social interests, students with volatile emotions and weak stress management abilities, and even students at risk of developing psychological problems under specific emotional or family environments. The application steps of this system in this scenario include full registration and basic data collection, the system conducts a preliminary emotional state classification based on multi-modal data, students start to conduct gamification task evaluations, and the reinforcement learning optimization layer assigns tasks according to the current state and interest characteristics of each student, and teachers and psychologists regularly view the class psychological reports automatically generated by the system.

[0120] 1. Deep Fusion Strategy

[0121] Real-time fusion: When students are performing gamified tasks, real-time acquisition of heart rate fluctuations, changes in facial expression key points, and operation behavior data, and short-term prediction and early warning are carried out in combination with historical social network data.

[0122] Dynamic weight adjustment: Automatically adjust ω according to the psychological characteristics of students in different stages 1 , ω 2 , ω 3 ,, strengthening the attention to stress management or emotion recognition.

[0123] Multi-source data matching: The system can be docked with the learning management platform to obtain students' academic data (such as exam scores, homework completion), and conduct a comprehensive analysis of the psychological stress dimension and learning stress dimension.

[0124] 2. Example of adaptive intervention

[0125] When the system determines that a certain student has large mood swings before an exam and has made many incorrect operations in the gamified stress scenario, it will automatically increase the priority of this student to participate in the "stress relief mini-game" this week, and push invitations to online courses or offline salons on "stress management skills". For students detected with signs of serious social withdrawal, the system will remind the head teacher or psychological teacher in advance to conduct individual talks or group activity interventions.

[0126] Example three

[0127] See Figures 1 to 7 , in some practical applications, to ensure students' privacy and data security, the system has made relatively strict technical designs in the following aspects;

[0128] 1. Data anonymization processing

[0129] The present invention uses a unique encrypted identification code to replace personal information such as students' names and student numbers, ensuring that the real identity of students cannot be directly deduced during data collection and analysis. For feedback data that needs to be provided to third-party services or parents, the system performs desensitization processing or high-level overview output again to avoid leakage of sensitive information.

[0130] 2. Multi-layer encryption and access control

[0131] In the data transmission stage, the SSL / TLS protocol is used to encrypt the communication to prevent network attacks and data interception. The data storage layer uses a distributed database, and each node is provided with an independent encryption key to perform secondary encryption on sensitive data segments; user roles are separated from access levels; students can only view their personal data, teachers can view the data of the classes they are responsible for, psychological counselors can view the data of authorized students, and administrators have the highest level of data management authority but have no right to access specific student privacy details.

[0132] 3. Compliance audit

[0133] The system regularly audits and archives the data operation logs to prevent unauthorized access or illegal operations. When the school or education authorities need to check the usage of students' psychological data, they can retrieve the corresponding records in the system management background to ensure the compliance and transparency of data usage.

[0134] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0135] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A student psychological evaluation system based on AI reinforcement learning optimization, characterized in that: The system comprises: The data collection layer is used to obtain students' behavioral data, physiological data, social interaction data, and environmental context data from multiple platforms and channels; A data processing layer, connected to the data acquisition layer, for preprocessing, data cleaning, feature extraction and multimodal data fusion of the acquired data; A reinforcement learning optimization layer, connected to the data processing layer, for executing a deep reinforcement learning algorithm to adaptively optimize the psychological evaluation strategy; A gamification evaluation module interacts bidirectionally with the reinforcement learning optimization layer to provide students with a variety of interactive games, task challenges, reward mechanisms, and situational simulations during the psychological assessment process, so that students can generate data required for psychological state assessment while completing games and tasks; A feedback and intervention layer, connected to the gamification evaluation module and the reinforcement learning optimization layer, for providing immediate feedback to students, teachers and psychological counselors based on real-time evaluation results and triggering intervention measures when potential psychological problems are detected; A user interface layer, connected to the feedback and intervention layer, for providing data visualization, evaluation reports, intervention prompts and personalized settings for students, teachers and psychological counselors; Among them, the reinforcement learning optimization layer dynamically adjusts the task design and content of the gamification evaluation module according to the real-time status and historical data of different student groups and individuals, thereby improving the accuracy and effectiveness of psychological evaluation.

2. The student psychological evaluation system based on AI reinforcement learning optimization according to claim 1 is characterized in that: The reinforcement learning optimization layer adopts a reinforcement learning algorithm based on a deep Q network, which updates the action value function Q(s,a) based on the following formula at each task assignment and feedback: in; s represents the student’s current psychological and behavioral state, and a represents the selected gamified assessment task; α is the learning rate, γ is the discount factor; r represents the immediate reward obtained based on student feedback and system goals after the gamification evaluation task is completed; s ′ and a ′ Respectively represent the state and action at the next moment; Through repeated iterative learning, the Q(s,a) is made to approach the optimal value function, thereby achieving adaptive task allocation and evaluation optimization for different students.

3. The student psychological evaluation system based on AI reinforcement learning optimization according to claim 1 is characterized in that: The gamification evaluation module includes: Emotion recognition game unit, which is used to assess students' emotion recognition ability and emotional expression tendency by identifying and expressing different emotion icons, story situations, and role-playing scenarios; Stress management game unit, which is used to simulate and set up various stressful scenarios and time-critical challenges, measure students' heart rate, breathing rate and reaction time during task execution, and evaluate students' stress tolerance and emotion regulation skills; The social interaction simulation unit is used to collect and evaluate students’ social skills, interpersonal interaction preferences, and communication strategies through virtual social scenarios and multiplayer collaborative game tasks; The task reward mechanism unit is used to give students positive feedback in the form of points, badges and leaderboards after they complete stage goals and challenges, thereby enhancing their participation and persistence; Among them, the multi-dimensional data collected by each game unit is transmitted to the data processing layer in real time for fusion, providing an optimization basis for the reinforcement learning optimization layer.

4. The student psychological evaluation system based on AI reinforcement learning optimization according to claim 3 is characterized in that: The stress management game unit also includes: The situation simulation submodule constructs virtual scenarios similar to real campus life and social situations, providing diverse and progressive task pressures; The physiological signal acquisition submodule obtains students’ heart rate variability, expression changes, and hand movement characteristics through wearable devices, cameras, and mouse trajectory recording; The stress score calculation submodule combines the students' task completion, time consumption and error rate indicators in the context to generate a comprehensive stress score, and interacts with the reinforcement learning optimization layer to adaptively control the game difficulty and feedback mechanism.

5. The student psychological evaluation system based on AI reinforcement learning optimization according to claim 1 is characterized in that: The data processing layer comprises: Multimodal data fusion module, used for feature extraction and weighted synthesis of behavioral data, physiological data, and social interaction data; A data cleaning module is used to remove incomplete data, abnormal data and noise interference to ensure that the data input to the reinforcement learning optimization layer has high confidence; The feature selection and dimensionality reduction module is used to reduce the dimensionality of multimodal features by using correlation analysis and principal component analysis to reduce redundant dimensions and avoid overfitting risks; The real-time data monitoring module is used to conduct real-time monitoring and preliminary analysis of the massive amount of collected data, and provide necessary processing instructions and risk warnings for the feedback and intervention layer.

6. The student psychological evaluation system based on AI reinforcement learning optimization according to claim 1 is characterized in that: The feedback and intervention layer includes: The instant feedback module generates visual reports and text suggestions based on students’ performance in the gamification evaluation module and the model prediction results provided by the reinforcement learning optimization layer, and displays them to students and teachers in real time; The multi-level intervention trigger module provides differentiated support to students according to the intervention level. When the system detects that a student has significant mental health risks, it automatically issues a high-priority warning message and recommends professional psychological counseling. The resource recommendation module recommends online psychology courses, offline counseling activities, and self-help training programs that match students’ stress types and emotional characteristics of social preferences; The notification scheduling module sends high-severity risk information to designated teachers and psychological counselors in a timely manner, facilitating follow-up and offline intervention at the first time.

7. The student psychological evaluation system based on AI reinforcement learning optimization according to claim 1 is characterized in that: The user interface layer includes: The student interface supports multi-platform login, provides task entry, personal data visualization, reward mechanism display, and psychological self-evaluation and help request channels; The teacher-side interface provides a class overview of student psychological assessment results, individual in-depth reports, intervention suggestions, and task management tools, so that teachers can adjust teaching and management strategies in a timely manner; The psychological counselor interface provides professional psychological assessment data visualization charts, multi-dimensional comparative analysis, intervention record management, and student case tracking functions; The management backend interface allows system administrators to assign permissions, adjust policies, back up and restore data, and perform privacy compliance configuration operations.

8. A student psychological evaluation method based on AI reinforcement learning optimization, characterized in that: The steps include: Step S1: Data collection, obtaining students' multimodal data from mobile applications, web pages and third-party platforms; Step S2; Data preprocessing and fusion: cleaning, normalizing and feature extraction of the collected data, and fusion to obtain the current comprehensive psychological state characteristics of the students; Step S3: Gamification task allocation, according to the comprehensive psychological state characteristics of the students, the reinforcement learning optimization layer adaptively selects and generates personalized gamification evaluation tasks; Step S4: Task execution and behavior collection: students complete the assigned tasks in the gamification evaluation module, and collect their behavior performance and physiological signals in real time; Step S5; Reinforcement learning strategy update, based on real-time feedback and collected data, iteratively updates the task allocation strategy through deep reinforcement learning algorithm; Step S6: Result feedback and intervention, generate student psychological assessment reports, automatically trigger intervention mechanisms for high-risk students detected, and push corresponding information to teachers and psychological counselors.

9. The student psychological evaluation method based on AI reinforcement learning optimization according to claim 8 is characterized in that: The method adopts a fusion formula based on weighted coefficients when fusing multimodal data; F fusion =ω1F behavior +ω2F physiology +ω3F social in; F behavior A feature vector representing students’ operational behaviors in gamification tasks; F physiology Feature vector representing students’ physiological data; F social A feature vector representing data related to students’ social interactions; ω1, ω2, ω3 are weighted coefficients used to balance the weights of each feature vector in the fusion result, satisfying ω1+ω2+ω3=1; By dynamically adjusting ω1, ω2, and ω3 at different stages, we can flexibly capture and evaluate students' psychological changes at different stages.

10. The student psychological evaluation method based on AI reinforcement learning optimization according to claim 8 is characterized in that: Also includes; Privacy protection: students’ personal identities are anonymized during data collection and transmission, and encryption algorithms are used to ensure the security of data transmission between the server and the client. Access control: set different permissions for users at all levels of the system to ensure that only authorized teachers, psychological counselors and administrators can access students' detailed psychological data; Data storage: Use distributed storage for data backup and redundancy, and perform secondary encryption storage on sensitive data to avoid data leakage; Compliance audit: Regularly audit and evaluate the system's data flow to ensure that the use and processing of students' psychological data complies with relevant laws and regulations as well as the compliance requirements of schools and educational institutions.

Images