A psychological abnormal behavior screening system for college students

Abstract

The application discloses a kind of psychological abnormal behavior screening system for college student group, it is related to mental health monitoring technical field, and its technical solution key points are: including by collecting the subjective state data and objective behavior data of student in course learning process, after data pre-processing, correlation analysis is carried out using recursive feature elimination, association rule mining and machine learning algorithm, constructs psychological abnormal risk prediction model, and real-time output psychological abnormal risk score;According to the evaluation result, individualized intervention suggestion is generated, and when risk score is lower than preset threshold, trigger graded early warning.The application realizes the real-time, accurate screening of student psychological abnormal behavior, overcomes the defects that traditional questionnaire survey feedback lags behind and is highly subjective, through graded early warning and individualized closed-loop intervention mechanism, effectively improves the learning self-efficacy and result of student, with significant educational application value.

Description

Technical Field

[0001] This invention relates to the field of mental health monitoring technology, and more specifically, to a screening system for abnormal psychological behaviors targeting university students. Background Technology

[0002] University students are at a critical stage of psychological development, facing multiple challenges such as academic pressure, job competition, and interpersonal relationships. The incidence of abnormal psychological behaviors (such as maladaptive learning, persistently low self-efficacy, and anxiety tendencies) is on the rise. Research shows that learning self-efficacy, i.e., students' subjective judgment of their ability to successfully complete learning tasks, is an important indicator reflecting the psychological state of university students, and abnormal changes in it often foreshadow broader psychological adaptation problems. Dynamic monitoring and data analysis of university students' course learning process to achieve early screening and timely intervention for abnormal psychological behaviors has become an important development direction in the fields of mental health education and teaching management in universities. Simultaneously, with the development of educational informatization and big data technology, the use of data mining, machine learning, and other technologies provides solid technical support for building efficient and accurate psychological screening and early warning systems.

[0003] Currently, the assessment of college students' mental state and screening for abnormal behavior mainly rely on the following methods: traditional questionnaire survey method: collecting students' self-reported data by distributing standardized questionnaires at the end of the teaching period or at specific time points; teacher observation and follow-up method: qualitatively assessing students' learning status and psychological performance through teacher-student interaction in classroom teaching, online courses, and after-class follow-ups; and performance-based evaluation method: indirectly inferring students' learning status and psychological adaptation level mainly based on students' usual homework grades, classroom quizzes, and final exam scores.

[0004] However, existing technologies still have the following problems:

[0005] Traditional questionnaire surveys have low frequency, significant feedback delays, and fail to reflect the dynamic changes in students' learning process. Furthermore, questionnaire results are influenced by students' subjective recollections and response biases, resulting in insufficient objectivity. Teacher observation and follow-up methods heavily rely on teachers' personal experience and level of involvement, are highly subjective, and are difficult to standardize and apply on a large scale. Teachers also struggle to simultaneously monitor subtle changes in every student, easily missing early warning signs. Performance-based evaluation methods are problematic because academic performance is the result of multiple factors, and its relationship with abnormal psychological behavior is not a simple linear one. More importantly, declining grades often only manifest after psychological problems have become quite severe, lacking early warning capabilities.

[0006] Meanwhile, among existing data analysis technologies, a few studies have attempted to analyze learning behavior data (such as login frequency and learning duration) using descriptive statistics or simple data visualization methods. However, these technologies mostly remain at the level of data aggregation and report display. On the one hand, they fail to effectively preprocess the raw, multi-source, heterogeneous data before data analysis, making it difficult to guarantee data quality. On the other hand, they fail to deeply explore the correlations between multi-source data and lack the ability to build iteratively optimized predictive models based on machine learning techniques such as recursive feature elimination and association rule mining. This makes it impossible to achieve accurate assessment and proactive early warning of students' psychological abnormality risks. Furthermore, in existing early warning and intervention mechanisms, even if a few systems have basic early warning functions, their early warning mechanisms are mostly triggered by a single threshold, lacking a hierarchical early warning design. The notification channels are singular and do not cover multiple roles. In addition, existing technologies generally lack a closed-loop follow-up mechanism after early warning is triggered, making it impossible to continuously track and provide feedback on the intervention effect. This leads to a disconnect between early warning and intervention, making it difficult to form an effective closed-loop management.

[0007] Therefore, the present invention aims to provide a psychological abnormality screening system for university students to solve the above problems. Summary of the Invention

[0008] The purpose of this invention is to provide a psychological abnormality screening system for university students. This invention overcomes the shortcomings of traditional questionnaire surveys, such as delayed feedback and strong subjectivity, by enabling real-time and accurate screening of students' psychological abnormalities. Through a tiered early warning and personalized closed-loop intervention mechanism, it effectively improves students' learning self-efficacy and academic performance, and has significant educational application value.

[0009] The above-mentioned technical objective of the present invention is achieved through the following technical solution: a psychological abnormality screening system for college students, including a data acquisition module, a data analysis module, a psychological abnormality assessment module, a personalized intervention suggestion module, and an early warning module;

[0010] The data acquisition module is used to collect multidimensional data of college students during the course learning process and transmit the collected multidimensional data to the data analysis module. The multidimensional data includes user subjective learning status data and user objective learning behavior data.

[0011] The data analysis module is used to receive multidimensional data transmitted by the data acquisition module, perform data preprocessing on the multidimensional data to output a standardized dataset, and then use data mining and machine learning algorithms to perform correlation analysis on the standardized dataset to identify key behavioral factors that affect the user's psychological state, and transmit the analysis results to the psychological abnormality assessment module.

[0012] The psychological abnormality assessment module is used to receive the analysis results transmitted by the data analysis module, establish and apply the psychological abnormality assessment model based on the analysis results, assess the user's psychological abnormality risk level in real time, and transmit the assessment results to the personalized intervention suggestion module and the early warning module. The psychological abnormality is at least manifested as an abnormal change in learning self-efficacy.

[0013] The personalized intervention suggestion module is used to receive the assessment results transmitted by the psychological abnormality assessment module, and generate and provide personalized learning suggestions and resource recommendations for the user based on the assessment results;

[0014] The early warning module is used to receive the assessment results transmitted by the psychological abnormality assessment module. When the user's psychological abnormality risk level in the assessment results is lower than a preset threshold, the early warning mechanism is automatically triggered to send an early warning notification to the user and the teaching terminal.

[0015] The present invention is further configured such that: the data acquisition module includes a subjective data acquisition unit and an objective data acquisition unit;

[0016] The subjective data collection unit is used to collect users' self-reported data through a dynamic questionnaire embedded in the online learning platform. The self-reported data includes learning interest, time management perception, self-assessment of course content mastery, and learning satisfaction.

[0017] The objective data acquisition unit is used to automatically record users' objective learning behavior data by monitoring the backend logs and frontend tracking points of the learning platform. The objective learning behavior data includes login frequency, learning duration, video viewing behavior data, forum discussion participation times, homework submission and late submission, online test scores, and answering time.

[0018] The present invention is further configured such that: the data analysis module includes a data preprocessing unit and a data analysis unit;

[0019] The data preprocessing unit is used to perform data cleaning, data formatting and data verification on the multidimensional data, and output a standardized dataset.

[0020] The data analysis unit is used to select the K key behavioral features that contribute the most to the prediction target from the initial feature set of the standardized dataset by combining recursive feature elimination with random forest algorithm, and to train the psychological abnormality risk prediction model based on the selected K key behavioral features using at least one machine learning algorithm selected from linear regression, random forest or neural network.

[0021] The present invention is further configured such that: the psychological abnormality assessment module loads the psychological abnormality risk prediction model, and calls the prediction model to perform forward prediction based on the latest characteristic data of the user in the current assessment period, and outputs a psychological abnormality risk score, wherein the psychological abnormality risk score ranges from 0 to 100, and the lower the score, the higher the psychological abnormality risk.

[0022] The present invention is further configured such that: the early warning module monitors the psychological abnormality risk score in real time, triggers a medium-level early warning when the score is below 40 points, and triggers a high-level early warning when the score is below 25 points, and the early warning notification is sent in parallel via email, SMS and platform in-site messages.

[0023] This invention also provides a method for screening abnormal psychological behaviors in college students, comprising the following steps:

[0024] S1. Data Collection: Collect multidimensional data of college students during the course learning process, including subjective learning status data and objective learning behavior data.

[0025] S2. Data Preprocessing: Cleaning, formatting, and validating the collected multidimensional data to output a standardized dataset.

[0026] S3. Data Analysis: Use data mining and machine learning algorithms to perform correlation analysis on the standardized dataset, identify key behavioral factors that affect users' psychological state, and output the analysis results of key behavioral factors.

[0027] S4. Psychological abnormality assessment: Based on the analysis results of the key behavioral factors, the psychological abnormality assessment model is used to assess the user's current psychological abnormality risk level in real time, and the psychological abnormality risk assessment results are output. The psychological abnormality is at least manifested as an abnormal change in learning self-efficacy.

[0028] S5. Generate intervention suggestions: Based on the output of the psychological abnormality risk assessment results, generate and provide personalized learning suggestions and resource recommendations for the user;

[0029] S6. Warning Trigger: When the output psychological abnormality risk assessment result shows that the user's psychological abnormality risk level is lower than the preset threshold, an warning is automatically triggered, and a notification is sent to the user and the teaching terminal.

[0030] The present invention is further configured such that: in step S1, the subjective learning status data is collected through a dynamic questionnaire embedded in the online learning platform, and the dynamic questionnaire is pushed out on a cycle of once every two weeks or after the completion of a teaching unit; the objective learning behavior data is collected by listening to the data interface of the learning management system or by periodically pulling log files, with a collection frequency of once every 5 minutes.

[0031] The present invention is further configured such that the data preprocessing in step S2 specifically includes: cleaning the original data, removing duplicate records, filtering outliers, and filling in missing values; formatting the cleaned data, including unifying the format of timestamps, encoding text data into numerical data, and normalizing the behavior counting features; and validating the formatted data according to preset logical rules, marking data that does not meet the rules and sending it to an exception queue.

[0032] The present invention is further configured as follows: In step S3, by combining recursive feature elimination with random forest algorithm, K key behavioral features that contribute the most to the prediction target are selected from the initial feature set of the standardized dataset, and the prediction target is learning self-efficacy score; and an association rule mining algorithm is used to explore the implicit association patterns between each behavioral feature; based on the selected K key behavioral features, at least one machine learning algorithm selected from linear regression, random forest or neural network is used to train the psychological abnormality risk prediction model.

[0033] The present invention also provides a psychological abnormality screening device for college students, including at least one processor; and a memory communicatively connected to at least one of the processors; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to implement a psychological abnormality screening method for college students.

[0034] In summary, the present invention has the following beneficial effects:

[0035] 1. This invention collects and analyzes students' subjective state data and objective behavioral data in real time during the course learning process. It uses recursive feature elimination, association rule mining and machine learning algorithms to build a psychological abnormality risk prediction model, realizing a dynamic and quantitative assessment of students' psychological state. Compared with traditional lagging assessment methods that rely on low-frequency questionnaires, teachers' subjective observation or final exam scores, this invention can advance the screening response time from the end of the semester to the daily learning process, reflect the micro-changes in students' learning self-efficacy in real time, improve the accuracy of students' psychological state assessment, and effectively overcome the defects of traditional methods such as serious feedback lag, strong subjectivity and insufficient objectivity.

[0036] 2. This invention implements a tiered early warning mechanism by pre-setting dual thresholds and intelligently matches differentiated notification recipients according to different early warning levels. Early warning notifications are sent in parallel via email, SMS, and platform in-site messages. At the same time, after an early warning is triggered, the system continuously monitors the changes in risk scores over the next 7 days, realizing automatic closed-loop management of early warning cancellation and escalation intervention. Compared with existing technologies that can only provide post-event summaries or single notifications, this invention realizes the transformation from passive response to proactive prevention and control, and constructs a complete screening-early warning-intervention-reassessment-closed-loop follow-up full-chain mechanism, which has significant effectiveness in early risk identification and effective intervention.

[0037] 3. This invention accurately identifies the risk characteristics of individual students and combines a hybrid strategy of rule engine and collaborative filtering to generate targeted intervention suggestions and resource recommendations for each student. Compared with the standardized general suggestions or random interventions that rely entirely on the teacher's personal experience in the prior art, this invention can match the most appropriate intervention measures according to the specific risk characteristics of each student, refer to the effective experience of similar successful cases, and accurately recommend learning resources according to their knowledge gaps, thereby improving the relevance of teaching and the learning effect of students.

[0038] 4. This invention ensures the quality of analyzed data through data preprocessing. On this basis, it uses a recursive feature elimination algorithm to automatically filter key behavioral features and combines association rule mining to explore implicit correlation patterns between features. Finally, it trains an iteratively optimized prediction model through machine learning algorithms, enabling the system to continuously update and optimize itself based on newly accumulated data. This solves the problem that most existing technologies remain at the level of descriptive statistics or simple report display, failing to deeply explore the correlation between multi-source data and lacking the ability to iteratively optimize the model. This leap from static statistics to dynamic learning significantly improves the intelligence level and adaptability of psychological abnormality screening at the technical level, making the screening results increasingly accurate as data accumulates.

[0039] 5. This invention is deployed on a general online learning platform, does not rely on specific hardware devices, and its data collection is compatible with mainstream learning behavior log formats and online questionnaire tools. The early warning and suggestion module can be seamlessly integrated with existing teaching management systems. Verified through practical application in multiple courses at different universities, this system can effectively adapt to student groups with different academic backgrounds. It demonstrates consistent and significant positive effects in enhancing learning self-efficacy, reducing the risk of psychological abnormalities, and improving teaching outcomes. It has the potential for large-scale application in higher education and provides strong technical support for university mental health education and teaching management. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the module structure of a psychological abnormal behavior screening system for college students in Embodiment 1 of the present invention;

[0041] Figure 2 This is a schematic diagram of the workflow of a psychological abnormal behavior screening system for college students in Embodiment 1 of the present invention;

[0042] Figure 3 This is a schematic diagram of the environmental boundary of a psychological abnormal behavior screening system for college students in Embodiment 1 of the present invention;

[0043] Figure 4 This is a schematic diagram of the interaction sequence of a psychological abnormality screening system for college students in Embodiment 1 of the present invention;

[0044] Figure 5 This is a schematic diagram of the service architecture of a psychological abnormal behavior screening system for college students in Embodiment 1 of the present invention;

[0045] Figure 6 This is a schematic diagram of the framework structure of a psychological abnormal behavior screening device for college students in Embodiment 5 of the present invention;

[0046] Figure 7 This is a flowchart illustrating the steps of a method for screening abnormal psychological behaviors in college students, as described in Embodiment 2 of the present invention. Detailed Implementation

[0047] The following is in conjunction with the appendix Figures 1-7 The present invention will be described in further detail below.

[0048] Example 1: A psychological abnormality screening system for university students includes a data acquisition module, a data analysis module, a psychological abnormality assessment module, a personalized intervention suggestion module, and an early warning module. The output of the data acquisition module is connected to the input of the data analysis module, the output of the data analysis module is connected to the input of the psychological abnormality assessment module, and the output of the psychological abnormality assessment module is connected to the inputs of the personalized intervention suggestion module and the early warning module, respectively. The modules interact with each other through synchronous call interfaces or asynchronous message queues.

[0049] In this embodiment, the data acquisition module is the system's data entry point, deployed on the front-end and back-end log servers of the online learning platform. It is responsible for collecting multi-dimensional data from university students during their course learning process. This module contains subjective data acquisition units and objective data acquisition units.

[0050] The subjective data collection unit is responsible for collecting users' self-reported data through dynamic questionnaires embedded in the online learning platform; the objective data collection unit is responsible for automatically recording users' objective learning behavior data by monitoring the backend logs and frontend event tracking of the learning platform. The data collection module aggregates all collected data, along with timestamps and unique user identifiers, into the system's shared data buffer, and then transmits all raw data in the buffer to the data analysis module via the internal data bus.

[0051] In this embodiment, the input end of the data analysis module is connected to the output end of the data acquisition module, and is responsible for receiving multidimensional data transmitted by the data acquisition module. The module contains a data preprocessing unit and a data analysis unit.

[0052] The data preprocessing unit is responsible for cleaning, formatting, and validating multidimensional data, and ultimately outputting a standardized dataset.

[0053] The data analysis unit is responsible for using data mining and machine learning algorithms to perform correlation analysis on standardized datasets and identify key behavioral factors that influence users' psychological states. The data analysis module then transmits the analysis results (including a list of key behavioral factors and a trained psychological abnormality risk prediction model) to the psychological abnormality assessment module.

[0054] In this embodiment, the input end of the psychological abnormality assessment module is connected to the output end of the data analysis module, and is responsible for receiving the analysis results transmitted by the data analysis module. After loading the psychological abnormality risk prediction model, the module calls the model to perform forward prediction based on the latest characteristic data of the user in the current assessment period, and outputs a psychological abnormality risk score (value 0-100, the lower the score, the higher the psychological abnormality risk). At the same time, it outputs several negative behavioral characteristics that contribute the most to the score and their respective contribution percentages. Psychological abnormality is at least manifested as an abnormal change in learning self-efficacy. The psychological abnormality assessment module transmits the assessment results (including psychological abnormality risk score, list of negative behavioral characteristics and contribution percentages, and abnormal decline markers) to the personalized intervention suggestion module and the early warning module, respectively.

[0055] In this embodiment, the input end of the personalized intervention suggestion module is connected to the output end of the psychological abnormality assessment module. It is responsible for receiving the assessment results transmitted by the psychological abnormality assessment module and generating personalized learning suggestions and resource recommendations for the user based on the assessment results. The module uses a hybrid strategy based on rule engine and collaborative filtering to generate intervention suggestions. At the same time, based on the learning units that the user has not completed or has low scores, it automatically retrieves and recommends corresponding learning resources from the course resource library. The generated intervention suggestions are output in the form of a structured report, displayed through the front-end interface, and pushed to the user's personal learning profile.

[0056] In this embodiment, the input end of the early warning module is connected to the output end of the psychological abnormality assessment module, and is responsible for receiving the assessment results transmitted by the psychological abnormality assessment module. This module monitors the psychological abnormality risk score in real time. When the score is lower than the preset threshold, a graded early warning mechanism is triggered, and early warning notifications are sent in parallel via email, SMS and platform in-site messages. After the early warning is triggered, the risk score change of the user is continuously monitored for the next 7 days. When the score rises back to the safe zone, an early warning cancellation notification is automatically sent. When the score does not improve after 3 consecutive early warnings, a manual intervention work order is generated and pushed to the teaching terminal.

[0057] Example 2: A method for screening abnormal psychological behaviors in university students, comprising the following steps:

[0058] S1, Data Acquisition.

[0059] This step is performed by the data acquisition module, which is responsible for collecting multidimensional data on university students during their course learning process. The multidimensional data includes subjective learning status data and objective learning behavior data.

[0060] Subjective learning status data is collected through a dynamic questionnaire embedded in the online learning platform. The questionnaire is automatically pushed to student users every two weeks according to a preset cycle, or triggered after a student completes a teaching unit (such as completing a chapter or passing a unit test). The questionnaire design includes questions in four dimensions: learning interest, time management perception, self-assessment of course content mastery, and learning satisfaction. Each question uses a five-point Likert scale (1 = completely disagree, 5 = completely agree). After students complete and submit the questionnaire, the system encapsulates the questionnaire data in JSON format, including a timestamp (accurate to the second) and a unique student identifier (student ID), and uploads it to the system database's raw data buffer via HTTPS security protocol.

[0061] Objective learning behavior data is collected through a combination of methods: monitoring the learning management system's Webhook interface (receiving user behavior events in real time, such as login, video playback, pause, drag and drop, assignment submission, forum posting, etc.) and periodically fetching log files (every 5 minutes). The collected data includes, but is not limited to: login frequency, single learning session duration and cumulative learning duration, video viewing behavior data (viewing duration, number of pauses / drags, completion rate), forum discussion participation (number of posts, number of replies, number of likes), assignment submission and late submission status (submission time, number of late submission days, number of revisions), online test scores, and time spent answering questions. The objective data collection unit initially aggregates the above behavior logs at an hourly granularity (merging multiple logins of the same student within one hour into a single session record) and stores them in the raw data buffer of the data warehouse in structured table format.

[0062] S2, Data Preprocessing.

[0063] This step is performed by the data preprocessing unit in the data analysis module, which is responsible for cleaning, formatting and validating the multidimensional raw data collected in step S1, and outputting a standardized dataset.

[0064] Data cleaning includes: removing duplicate records, deleting multiple identical records of the same student at the same timestamp due to network retransmissions or duplicate log entries; filtering outliers, identifying and removing obviously unreasonable data (such as study time exceeding 24 hours or being negative, test scores exceeding the 0-100 range, and video viewing time exceeding 120% of the total video duration); filling missing values, for continuous variables (such as study time, test scores) with a missing rate of less than 5%, using the average of the student's historical data for the same period, or using the class average if there is no historical data for the student; filling missing values ​​for categorical variables (such as whether homework was submitted) using the mode; and directly deleting records with missing or irreparable key fields such as student ID and timestamp.

[0065] Data formatting includes: converting timestamps from different data sources to the format "yyyy-MM-dd HH:mm:ss"; encoding text options in the questionnaire into numerical values ​​("strongly disagree" corresponds to 1, "disagree" to 2, "neutral" to 3, "agree" to 4, and "strongly agree" to 5); normalizing behavioral count features (number of posts, number of logins, and study time) by dividing the individual student's behavioral count by the highest count in the class and mapping it to the [0,1] interval; and constructing derived features: "learning regularity" is defined as the standard deviation of the login time interval (an inverse indicator, the smaller the value, the more regular the learning); "interaction activity" is defined as (number of posts + number of replies) divided by the highest interaction number in the class; and "learning engagement" is defined as (video viewing time + homework time) divided by the total course duration.

[0066] Data validation verifies data consistency through preset logical rules: the assignment submission time should not be earlier than the assignment release time; the assignment submission time should not be later than the deadline plus 7 days; the video viewing time should not exceed the total duration of the video; there should be a reasonable correlation between online test scores and answering time; any data that violates the rules is marked as "suspicious" and sent to an exception queue for manual review or automatic removal; the validated data forms a standardized dataset, stored in a two-dimensional table structure (each row corresponds to an aggregate record of a student within a day, and each column corresponds to a feature).

[0067] S3, Data Analysis.

[0068] This step is performed by the data analysis unit in the data analysis module. It is responsible for using data mining and machine learning algorithms to perform correlation analysis on the standardized dataset, identify key behavioral factors that affect users' psychological state, and output the key behavioral factor analysis results.

[0069] Key behavioral feature selection: By combining recursive feature elimination with the random forest algorithm, the K key behavioral features (K = 10-15) that contribute the most to the prediction target are selected from the initial feature set of the standardized dataset (usually containing 50-80 original and derived features). The prediction target is the learning self-efficacy score, which is derived from the self-efficacy scale scores collected periodically in step S1. The specific process of recursive feature elimination is as follows: First, train the random forest model using all features, sort the features according to their importance, and delete the feature with the lowest importance; then retrain the model on the remaining features and sort them again, repeating the above process until the number of remaining features reaches the preset value of K. As verified in this embodiment, the most significant key behavioral features usually include: the rate of decrease in learning time over two consecutive weeks, the number of late assignments, the video completion rate, the frequency of forum interaction, and the regularity of learning periods.

[0070] Association rule mining: Association rule mining algorithms (Apriori or FP-Growth) are used to explore implicit association patterns between various behavioral features. Students' various behavioral events are transformed into Boolean transaction data, with a minimum support threshold of 0.1 and a minimum confidence threshold of 0.6, to mine frequent itemsets and strong association rules. For example, this analysis might reveal a confidence level of 0.75 between "high video dragging frequency" and "low online test scores," and significant support and confidence levels between "no login for 3 consecutive days" and "late homework submission."

[0071] Training of the Psychological Abnormality Risk Prediction Model: Based on K selected key behavioral features, the psychological abnormality risk prediction model is trained using at least one machine learning algorithm selected from linear regression, random forest, or neural networks. 80% of the samples in the standardized dataset are used as the training set, and 20% as the validation set. k-fold cross-validation (k=5 or 10) is used to adjust the model hyperparameters (for random forest algorithms, this includes the number of trees, maximum depth, minimum number of leaf node samples, etc.; for neural network algorithms, this includes the number of hidden layers, number of neurons, learning rate, Dropout ratio, etc.). Mean squared error or mean absolute error is used as the loss function for the regression task. After training, the model is stored as a serialized file in a model repository. The model is retrained weekly on an offline cluster based on newly accumulated data to achieve incremental learning and iterative model optimization.

[0072] S4. Psychological abnormality assessment.

[0073] This step is performed by the psychological abnormality assessment module, which is responsible for assessing the user's current psychological abnormality risk level in real time based on the key behavioral factor analysis results output in step S3, and outputting the psychological abnormality risk assessment results; psychological abnormality is at least manifested as abnormal changes in learning self-efficacy.

[0074] Real-time risk scoring: Load the psychological abnormality risk prediction model trained in step S3, call the model to make forward predictions based on the latest feature data of users within the current assessment period (daily or every 6 hours), and output a psychological abnormality risk score S, with a value of 0-100. The lower the score, the higher the psychological abnormality risk (i.e., the lower the learning self-efficacy). Taking the random forest model as an example, S is the mean of the prediction results of all decision trees; taking the neural network model as an example, S is the result of the activation value of the output layer neurons after inverse normalization.

[0075] Output of Negative Behavioral Feature Contribution: Using interpretability tools of the model (such as SHAP value analysis or the feature importance built into random forests), the output shows the top few negative behavioral features that contribute the most to the score, along with their respective percentage contributions. SHAP value analysis, based on Shapley values ​​in game theory, quantifies the marginal contribution of each feature to the prediction result. For example, the output might be: "Learning time decreased by 35% week-on-week, contribution 40%; Homework submitted late twice, contribution 35%; Forum interaction was 0, contribution 25%."

[0076] Personalized baseline maintenance and abnormal decline judgment: A personalized baseline B (the user's average psychological abnormality risk score over the past four weeks) is maintained for each user; when the user's current score S is lower than B×(1-20%) and also lower than (class mean - 1.5×class standard deviation), the current status is marked as "abnormal decline"; the design of dual reference standards (combining individual longitudinal comparison and group horizontal comparison) avoids the interference of individual differences and overall class deviation on the evaluation results.

[0077] The assessment results (psychological abnormality risk score, list of negative behavioral characteristics and contribution percentage, abnormal decline marker, timestamp) are simultaneously written into a time-series database to support the visualization analysis of risk trends for individual students and groups.

[0078] S5. Generate intervention recommendations.

[0079] This step is performed by the personalized intervention suggestion module, which generates and provides personalized learning suggestions and resource recommendations for the user based on the psychological abnormality risk assessment results output in step S4.

[0080] Rule-based intervention suggestion generation: The personalized intervention suggestion module has a pre-stored mapping table between risk characteristics and intervention measures. When a negative behavioral characteristic in the psychological abnormality assessment result matches a risk characteristic in the mapping table, the rule engine automatically extracts the corresponding intervention measure. The mapping table is shown in Table 1.

[0081] Table 1. Mapping Table of Risk Characteristics and Intervention Measures

[0082]

[0083] Supplementary suggestions based on collaborative filtering: The collaborative filtering strategy retrieves the top 10 "successful improvement cases" from the database that are similar to the user's risk feature vector (cosine similarity > 0.7) and showed the greatest improvement in psychological abnormality risk score after one week. The high-frequency effective measures adopted in these cases (the top 3 with the highest frequency) are extracted as a supplement to the rule engine's suggestions.

[0084] Learning resource recommendation: Based on the learning units that the user has not completed or scored low (e.g., an online test score below 70 points for a certain chapter), the system automatically retrieves and recommends corresponding learning resources from the course resource library, including video tutorials, accompanying exercises, and extended reading materials. The resources are then sorted by two weighted indicators: matching degree and historical performance score, and a recommendation list (usually the top 5 items) is output. The weighted sorting formula is: Overall score = 0.6 × matching degree + 0.4 × historical performance score.

[0085] Structured report output: The generated intervention recommendations are output in the form of a structured report, which includes the current psychological risk level (normal / mildly abnormal / moderately abnormal / severely abnormal), description of the main risk behaviors (listing the two negative behavioral characteristics with the highest contribution and their percentages), three specific action recommendations (generated jointly by the rule engine and collaborative filtering), and a list of recommended learning resources (including resource names, access links, and estimated learning duration). The report is generated in HTML format, displayed through the front-end interface, and simultaneously pushed to the user's personal learning profile database in JSON format.

[0086] S6, Warning triggered.

[0087] This step is executed by the early warning module, which automatically triggers an early warning and sends a notification to the user and the teaching terminal when the psychological abnormality risk assessment result output by step S4 shows that the user's psychological abnormality risk level is lower than a preset threshold.

[0088] Dual-threshold grading judgment: The early warning module monitors the psychological abnormality risk score S in real time and adopts a dual-threshold grading judgment mechanism; when S < 40 points, a medium-level early warning is triggered, indicating that the student's psychological state has shown obvious abnormalities and needs to be paid attention to in time; when S < 25 points, a high-level early warning is triggered, indicating that the student's psychological state has reached a more serious level of abnormality and needs to be intervened immediately.

[0089] Multi-role notification recipients: Intermediate alerts include the user and their instructor; advanced alerts include the user, instructor, counselor, and school counseling center.

[0090] Multi-channel parallel notification: Warning notifications are sent in parallel through three channels: email, SMS, and platform in-app messages. Emails are sent via SMTP protocol to a formatted mail server with the subject "[Psychological and Behavioral Warning] Please pay attention to your learning status," and the body includes the current score, a description of the main negative behavioral characteristics, an analysis of potential causes, and a link to personalized intervention suggestions. SMS messages are sent via HTTP API to a third-party SMS gateway with the content "[School Psychological Screening] Your recent learning status has been abnormal; please log in to the learning platform to view personalized suggestions." Platform in-app messages are displayed as a red dot in the online learning platform's notification center; clicking it redirects to the intervention suggestion report page.

[0091] Closed-loop follow-up mechanism: After an alert is triggered, the alert module continuously monitors the user's psychological risk score changes over the next 7 days, recording the score value daily and generating trend data. When the score rises back to the safe zone (≥45 points), the system automatically sends an alert cancellation notification via email and in-system messages, informing the user and teacher that the risk has been mitigated. If the score does not improve or continues to decline after 3 consecutive alerts (i.e., 3 consecutive weeks), the system automatically generates a manual intervention work order and pushes it to the college's psychological counseling station or counselor's workbench, where psychological counseling professionals will intervene to provide face-to-face counseling.

[0092] Example 3: Application of a Psychological Abnormal Behavior Screening System in University "Entrepreneurship and Employment Guidance" Course

[0093] In this embodiment, in order to improve the teaching effectiveness of the "Entrepreneurship and Employment Guidance" course, universities introduce a psychological abnormal behavior screening system for university students, as shown in Embodiment 1. This system can analyze students' learning data in real time, identify changes in their learning status, and provide timely warnings and personalized learning suggestions.

[0094] System Deployment and Data Collection: The system is deployed on the school's online learning platform and seamlessly integrated with the "Entrepreneurship and Employment Guidance" course; the system begins collecting students' learning behavior data, including login frequency, study time, homework completion status, online test scores, etc.

[0095] Real-time data analysis: The system uses machine learning algorithms to analyze the collected data in real time and calculate students' self-efficacy levels. Self-efficacy, as a psychological concept, refers to an individual's subjective judgment of whether they can successfully complete a certain behavior. In this embodiment, the system comprehensively evaluates students' self-efficacy through multiple dimensions such as homework completion, test scores, and study time.

[0096] Warning mechanism triggered: The system detects that a student's self-efficacy level has been declining for several consecutive weeks, falling below the class average and showing a clear downward trend; the system immediately triggers the warning mechanism and sends a warning notification to the student and the teacher.

[0097] Intervention measures: After receiving the warning, the teacher immediately communicated with the student to understand their recent learning situation and difficulties encountered; it was found that the student had recently encountered a bottleneck in writing a business plan and felt confused and frustrated; based on the personalized learning suggestions provided by the system, the teacher developed a targeted tutoring plan for the student, including providing additional learning resources and arranging one-on-one tutoring.

[0098] Evaluation of Results: After a period of counseling and intervention, the student's self-efficacy level gradually recovered and even exceeded the class average; the student's academic performance also improved, especially in the writing of business plans.

[0099] Parameter results: Before the warning: The student's self-efficacy level was 60 (out of 100), while the class average was 75; After the warning: After one month of intervention, the student's self-efficacy level increased to 85, exceeding the class average (still 75); Improved academic performance: The student's final exam score improved from 70 to 85, showing significant progress.

[0100] Table 2. Changes in Students' Self-Efficacy Levels

[0101]

[0102] Table 2 shows the changes in the student's self-efficacy level. Before the warning, it showed a downward trend, and after the intervention, it gradually increased and exceeded the class average.

[0103] Table 3 Comparison of Student Academic Performance

[0104]

[0105] Table 3 shows the student's progress in academic performance. The final exam score after the warning period improved by 15 points compared to before the warning, and exceeded the class average.

[0106] Conclusion: This embodiment demonstrates that the application of the intelligent learning support system in the "Entrepreneurship and Employment Guidance" course has achieved significant results. The system can analyze students' learning data in real time, promptly identify students' learning problems, and help students overcome difficulties and improve learning outcomes through early warning mechanisms and personalized learning suggestions. This proves the effectiveness of the system in improving students' learning efficiency and self-efficacy.

[0107] Example 4: Application of a Psychological Abnormal Behavior Screening System in the "Current Affairs and Policy" Course in Universities

[0108] In this embodiment, in order to enhance the teaching effectiveness of the "Current Affairs and Policy" course and improve students' ideological and political literacy and learning efficiency, universities decided to introduce a psychological abnormal behavior screening system for university students, as shown in Embodiment 1. This system analyzes students' learning behavior, interests, and knowledge mastery to tailor learning suggestions and resource recommendations for each student. After a semester of practical application, the system has significantly improved students' learning efficiency and satisfaction, and has also improved the overall teaching effectiveness of ideological and political courses.

[0109] System Deployment and Data Collection: The intelligent personalized learning system was integrated into the school's online teaching platform and closely combined with the "Current Affairs and Policy" course; the system began to comprehensively collect students' learning data, including learning time, browsing content, interaction, test scores, etc.

[0110] Personalized learning suggestion generation: The system uses advanced data analysis algorithms to deeply mine students' learning data, identify students' learning needs and weaknesses; based on these analysis results, the system generates personalized learning suggestions for each student, including recommended learning resources, suggested learning paths, and targeted practice questions.

[0111] Resource Recommendation and Learning Support: The system automatically recommends relevant learning resources, such as video lectures, article readings, and online discussions, based on students' learning progress and interests. At the same time, the system also provides real-time learning support, including online Q&A, learning progress tracking, and feedback on learning outcomes.

[0112] Effectiveness evaluation and feedback: After one semester of implementation, the school organized a survey on students' learning efficiency and satisfaction to evaluate the application effect of the system; at the same time, teachers also conducted a comprehensive evaluation of the system's teaching effectiveness based on students' learning performance and classroom interaction.

[0113] Parameter results: Improved learning efficiency: Average study time increased by 20%, homework completion speed increased by 15%, and average test scores increased by 10 points (out of 100); Improved satisfaction: Student satisfaction with the course increased from 75% to 90%, and satisfaction with personalized learning suggestions reached 95%; Improved teaching effectiveness: The overall pass rate of ideological and political courses increased by 10 percentage points, and the excellent rate also increased by 5 percentage points.

[0114] Table 4. Graph of Changes in Student Learning Efficiency

[0115]

[0116] Table 4 shows the changes in students' learning efficiency before and after implementing the intelligent personalized learning system. It can be seen that the average study time increased, the speed of completing assignments improved, and test scores also significantly improved.

[0117] Table 5 Comparison of Student Satisfaction and Teaching Effectiveness

[0118]

[0119] Table 5 shows a comparison of student satisfaction and teaching effectiveness before and after the implementation of the intelligent personalized learning system. It can be seen that students' satisfaction with the courses and personalized suggestions has increased significantly, while the pass rate and the rate of excellence in the courses have also increased significantly.

[0120] Conclusion: This embodiment demonstrates that the application of the intelligent personalized learning system in the "Current Affairs and Policy" course has achieved significant results. By providing students with personalized learning suggestions and resource recommendations, the system effectively improves students' learning efficiency and satisfaction, while also enhancing the overall teaching effect of ideological and political education. This fully proves the effectiveness and feasibility of the system in improving the quality of ideological and political education.

[0121] Example 5: A psychological abnormality screening device for university students, comprising at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the processor, the instructions being executed by the processor to implement a psychological abnormality screening method for university students as described in Example 2, the method comprising: data acquisition: collecting multidimensional data of university students during course learning, wherein the multidimensional data includes subjective learning status data and objective learning behavior data of users; data preprocessing: performing data cleaning, data formatting, and data verification on the collected multidimensional data, outputting a standardized dataset; data analysis: employing data mining and machine learning algorithms to analyze the data. The system performs correlation analysis on standardized datasets to identify key behavioral factors influencing users' psychological state and outputs the key behavioral factor analysis results. Psychological anomaly assessment: Based on the output key behavioral factor analysis results, a psychological anomaly assessment model is used to assess the user's current psychological anomaly risk level in real time and outputs the psychological anomaly risk assessment results. Psychological anomalies are at least manifested as abnormal changes in learning self-efficacy. Intervention suggestions are generated: Based on the output psychological anomaly risk assessment results, personalized learning suggestions and resource recommendations are generated and provided for the user. Early warning trigger: When the output psychological anomaly risk assessment results show that the user's psychological anomaly risk level is below a preset threshold, an early warning is automatically triggered, sending a notification to the user and the teaching platform.

[0122] This specific embodiment is merely an explanation of the present invention and is not intended to limit the invention. After reading this specification, those skilled in the art can make modifications to this embodiment without contributing any inventive step, but such modifications are protected by patent law as long as they are within the scope of the claims of the present invention.

Claims

1. A psychological abnormality screening system for university students, characterized in that: It includes a data acquisition module, a data analysis module, a psychological abnormality assessment module, a personalized intervention suggestion module, and an early warning module; The data acquisition module is used to collect multidimensional data of college students during the course learning process and transmit the collected multidimensional data to the data analysis module. The multidimensional data includes user subjective learning status data and user objective learning behavior data. The data analysis module is used to receive multidimensional data transmitted by the data acquisition module, perform data preprocessing on the multidimensional data to output a standardized dataset, and then use data mining and machine learning algorithms to perform correlation analysis on the standardized dataset to identify key behavioral factors that affect the user's psychological state, and transmit the analysis results to the psychological abnormality assessment module. The psychological abnormality assessment module is used to receive the analysis results transmitted by the data analysis module, establish and apply the psychological abnormality assessment model based on the analysis results, assess the user's psychological abnormality risk level in real time, and transmit the assessment results to the personalized intervention suggestion module and the early warning module. The psychological abnormality is at least manifested as an abnormal change in learning self-efficacy. The personalized intervention suggestion module is used to receive the assessment results transmitted by the psychological abnormality assessment module, and generate and provide personalized learning suggestions and resource recommendations for the user based on the assessment results; The early warning module is used to receive the assessment results transmitted by the psychological abnormality assessment module. When the user's psychological abnormality risk level in the assessment results is lower than a preset threshold, the early warning mechanism is automatically triggered to send an early warning notification to the user and the teaching terminal.

2. The psychological abnormality screening system for university students according to claim 1, characterized in that: The data acquisition module includes a subjective data acquisition unit and an objective data acquisition unit; The subjective data collection unit is used to collect users' self-reported data through a dynamic questionnaire embedded in the online learning platform. The self-reported data includes learning interest, time management perception, self-assessment of course content mastery, and learning satisfaction. The objective data acquisition unit is used to automatically record users' objective learning behavior data by monitoring the backend logs and frontend tracking points of the learning platform. The objective learning behavior data includes login frequency, learning duration, video viewing behavior data, forum discussion participation times, homework submission and late submission, online test scores, and answering time.

3. The psychological abnormality screening system for university students according to claim 1, characterized in that: The data analysis module includes a data preprocessing unit and a data analysis unit; The data preprocessing unit is used to perform data cleaning, data formatting and data verification on the multidimensional data, and output a standardized dataset. The data analysis unit is used to select the K key behavioral features that contribute the most to the prediction target from the initial feature set of the standardized dataset by combining recursive feature elimination with random forest algorithm, and to train the psychological abnormality risk prediction model based on the selected K key behavioral features using at least one machine learning algorithm selected from linear regression, random forest or neural network.

4. A psychological abnormality screening system for university students according to claim 1, characterized in that: The psychological abnormality assessment module loads the psychological abnormality risk prediction model and calls the prediction model to perform forward prediction based on the latest characteristic data of the user in the current assessment period, and outputs a psychological abnormality risk score. The psychological abnormality risk score ranges from 0 to 100, and the lower the score, the higher the psychological abnormality risk.

5. A psychological abnormality screening system for university students according to claim 4, characterized in that: The early warning module monitors the psychological abnormality risk score in real time. When the score is below 40, a medium-level early warning is triggered. When the score is below 25, a high-level early warning is triggered. The early warning notification is sent in parallel via email, SMS and platform in-site messages.

6. A method for screening abnormal psychological behaviors in university students, applied to the screening system described in any one of claims 1-5, characterized in that: Includes the following steps: S1. Data Collection: Collect multidimensional data of college students during the course learning process, including subjective learning status data and objective learning behavior data. S2. Data Preprocessing: Cleaning, formatting, and validating the collected multidimensional data to output a standardized dataset. S3. Data Analysis: Use data mining and machine learning algorithms to perform correlation analysis on the standardized dataset, identify key behavioral factors that affect users' psychological state, and output the analysis results of key behavioral factors. S4. Psychological abnormality assessment: Based on the analysis results of the key behavioral factors, the psychological abnormality assessment model is used to assess the user's current psychological abnormality risk level in real time, and the psychological abnormality risk assessment results are output. The psychological abnormality is at least manifested as an abnormal change in learning self-efficacy. S5. Generate intervention suggestions: Based on the output of the psychological abnormality risk assessment results, generate and provide personalized learning suggestions and resource recommendations for the user; S6. Warning Trigger: When the output psychological abnormality risk assessment result shows that the user's psychological abnormality risk level is lower than the preset threshold, an warning is automatically triggered, and a notification is sent to the user and the teaching terminal.

7. A method for screening abnormal psychological behaviors in university students according to claim 6, characterized in that: In step S1, the subjective learning status data is collected through a dynamic questionnaire embedded in the online learning platform. The dynamic questionnaire is pushed out every two weeks or after the completion of a teaching unit. The objective learning behavior data is collected by monitoring the data interface of the learning management system or by periodically pulling log files. The collection frequency is once every 5 minutes.

8. The method for screening abnormal psychological behaviors in college students according to claim 6, characterized in that: The data preprocessing in step S2 specifically includes: cleaning the original data, removing duplicate records, filtering outliers, and filling in missing values; formatting the cleaned data, including unifying the format of timestamps, encoding text data into numerical data, and normalizing the behavior counting features; and validating the formatted data according to preset logical rules, marking data that does not meet the rules and sending it to the exception queue.

9. A method for screening abnormal psychological behaviors in university students according to claim 6, characterized in that: In step S3, the K key behavioral features that contribute the most to the prediction target are selected from the initial feature set of the standardized dataset by recursive feature elimination combined with the random forest algorithm. The prediction target is to learn self-efficacy scores. The association rule mining algorithm is used to explore the implicit association patterns between the behavioral features. Based on the selected K key behavioral features, at least one machine learning algorithm selected from linear regression, random forest or neural network is used to train the psychological abnormality risk prediction model.

10. A psychological abnormality screening device for university students, characterized in that: The method includes at least one processor; and a memory communicatively connected to at least one of the processors; wherein the memory stores instructions executable by the processor to implement a method for screening abnormal psychological behaviors of university students according to any one of claims 6-9.

Images