AI-based college mental health comprehensive service system and method

Abstract

The invention relates to the technical field of psychological health management, and comprises an AI-based college psychological health comprehensive service system and method, and the system comprises a psychological behavior scoring module, an interactive behavior analysis module, a psychological health assessment module, a social intervention matching module, and a dynamic adaptation optimization module. According to the method, based on behavior data of learning, work and rest, social contact and the like, the time matching degree of behavior categories is calculated, the comprehensiveness of individual psychological health scoring is improved, social interaction data is combined, the social behavior resonance degree is analyzed, the psychological integrating degree between individuals is measured, semantic emotion polarity change is combined, and psychological state fluctuation is quantified; based on psychological health state change analysis, individuals with high social mode similarity are screened, pertinence of intervention matching is enhanced, fluctuation trends of individual social states are synchronously identified, psychological intervention strategies are adjusted according to changes of psychological health scores, personalized adaptability is improved, and fine adjustment of psychological health management of college students can be achieved.

Description

Technical Field

[0001] The present invention relates to the field of mental health management technology, and in particular to an AI-based comprehensive mental health service system and method for colleges and universities. Background Art

[0002] The field of mental health management technology includes methods and systems for the assessment, intervention and management of individual mental states. The core content includes mental health data collection, mental state analysis, psychological intervention strategy formulation and health management implementation. It is widely used in many industries such as medical care, education and social services. It mainly relies on psychological assessment scales, physiological signal monitoring, behavioral data analysis and other means to conduct a comprehensive assessment of individual psychological conditions, and combines artificial intelligence technology to provide personalized mental health intervention plans. In the university environment, students' mental health management involves emotional state monitoring, psychological problem identification, and consulting service optimization. The accuracy and accessibility of mental health services are improved through information technology.

[0003] Among them, the AI-based comprehensive service system for college mental health refers to a mental health management system built using artificial intelligence technology, which collects data, analyzes and optimizes services for the mental health status of college students. Specific methods include collecting students' self-assessment data through online psychological assessment tools, and combining campus smart terminal devices to obtain physiological indicator data, such as heart rate, skin electrical activity, etc.; using natural language processing technology to perform sentiment analysis on students' text or voice input to identify potential psychological risks; applying machine learning models to perform trend analysis on long-term mental health data, and establishing an individualized mental health assessment system based on classification models or neural network algorithms; combining intelligent recommendation technology based on knowledge graphs to provide personalized psychological counseling resource matching and intervention suggestions; optimizing virtual psychological counseling assistants through emotional computing technology so that they can provide human-computer interactive psychological counseling. The coordinated application of these means enables the system to provide multi-dimensional mental health management solutions.

[0004] The existing comprehensive mental health services in colleges and universities mainly rely on psychological assessment scales and physiological signal monitoring for mental health assessment, but lack in-depth analysis of social behavior patterns, resulting in the assessment of individual mental states being limited to static data and difficult to adapt to short-term fluctuations in mental states. Psychological assessment scales are subject to subjective self-assessment and are difficult to quantify dynamic changes in mental health. Physiological signal monitoring equipment is limited by the collection environment and individual differences, making it difficult to form long-term continuous data records, affecting the sustainability of mental health management. Behavioral data analysis mainly focuses on historical data and lacks real-time calculation of social interaction characteristics. It is unable to effectively measure the psychological fit between individuals, resulting in insufficient social matching capabilities for mental health management. Mental health interventions mostly rely on fixed plans and fail to optimize social matching in combination with the resonance of interactive behaviors, resulting in a lack of adaptability in intervention strategies. Since social recommendations fail to adjust dynamically, changes in mental states may lead to the failure of matching results, reduce the effectiveness of social interventions, and even lead to further accumulation of social pressure. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and propose an AI-based comprehensive college mental health service system and method.

[0006] In order to achieve the above objectives, the present invention adopts the following technical solutions: The AI-based comprehensive service system for college mental health includes: The psychological behavior scoring module obtains personal data of college students, calculates the time matching degree of behavior categories, sets behavior category weights and calculates mental health scores based on the weight mapping relationship to obtain individual mental health score data; The interactive behavior analysis module obtains social platform data based on the individual mental health score data, calculates the interaction time sequence matching degree and topic consistency, analyzes the behavior synchronization, and obtains the interactive behavior resonance data; The mental health assessment module obtains the emotional expression change data of the social platform based on the individual mental health score data and the interactive behavior resonance data, calculates the fluctuation amplitude of the mental state and the pressure fluctuation trend, and obtains the analysis results of the mental health state change; The social intervention matching module counts the student groups with the same similarity in social behavior patterns, sets social matching priorities, and screens interaction features, adjusts social recommendation relationships, and obtains personalized social intervention matching records; The dynamic adaptive optimization module obtains the target student's social interaction change data based on the personalized social intervention matching records, calculates the social behavior pattern deviation index, adjusts the psychological intervention situation, and obtains a mental health management optimization service plan.

[0007] As a further scheme of the present invention, the individual mental health score data includes behavior consistency score, psychological questionnaire parameter weight, mental health score weight mapping result, and individual psychological characteristic parameter matching value; the interactive behavior resonance data includes interaction timing matching degree, social topic consistency analysis results, and social behavior synchronization analysis results; the mental health state change analysis results include mental state fluctuation amplitude, mental health state change rate, psychological stress fluctuation trend analysis results, and negative emotion expression index; the personalized social intervention matching record includes social matching priority, target student interaction feature screening results, social recommendation adjustment plan, and social support matching relationship; the mental health management optimization service plan includes the degree of social behavior pattern deviation, individual social deviation index, mental health score fluctuation range adjustment value, and psychological intervention adaptation adjustment strategy.

[0008] As a further solution of the present invention, the psychological behavior scoring module includes: The behavior data acquisition submodule acquires data on study time, classroom concentration, homework submission time, exam preparation time, social interaction frequency, social network activity, work and rest stability, nighttime activity time, and exercise clock-in frequency. It summarizes the frequency and duration trends of each behavior category according to the time window to obtain the time distribution data of the behavior category. The behavior pattern calculation submodule sets the behavior category weight based on the behavior category time distribution data and the psychological questionnaire data, using the formula: ; Calculate the behavior matching deviation value , and combined with the time window to make matching adjustments, and obtain the matching degree of the behavior pattern, represents the weight of the i-th behavior category, represents the actual duration of the i-th behavior category, represents the standard duration of the ith behavior category, and n represents the total number of behavior categories; The mental health score calculation submodule calculates the mental health score based on the behavior pattern matching degree and the weight mapping relationship to obtain individual mental health score data.

[0009] As a further solution of the present invention, the interactive behavior analysis module includes: The social platform interaction feature extraction submodule collects data on social platform message response time, friend interaction frequency, group chat activity, topic switching mode, and emoticon usage ratio based on the individual mental health score data, analyzes the interaction behavior characteristics of different students, extracts the time series characteristics under each data dimension, calculates the mean and variance of the message response time between students, counts the periodicity of friend interaction frequency, analyzes the time distribution law of group chat activity, extracts the conversion probability matrix in the topic switching mode, obtains the distribution curve of the emoticon usage ratio, and obtains the interaction behavior time series characteristics; The social topic consistency calculation submodule calculates the frequency of common topics based on the temporal characteristics of the interactive behavior, counts the number of speaking turns and proportions of different students under the common topics, analyzes the coherence of the dialogue, calculates the information entropy difference between message sequences, and extracts the information complementarity features. The formula is: ; Calculating social topic consistency scores ,Get the social topic consistency index, where, represents the weight of the i-th round of dialogue, represents the semantic similarity of the i-th round of dialogue, N represents the total number of dialogue rounds, represents the information entropy of the i-th round, The baseline value of information entropy representing the topic; The social behavior synchronization analysis submodule calculates the matching degree of emoticon usage habits based on the social topic consistency index, counts the types and frequencies of emoticon usage by different students under the same topic, calculates the stability of the interaction rhythm, constructs an interaction timing distribution matrix by counting the mean and variance of the interaction time interval, evaluates the similarity of the interaction rhythm between individual students, and obtains the interaction behavior resonance data.

[0010] As a further solution of the present invention, the mental health assessment module includes: The emotion expression feature analysis submodule synchronously obtains the text data of college students on the social platform based on the individual mental health score data and the interactive behavior resonance data, extracts emotion-related words in the text, counts the proportion of negative words, the frequency of occurrence of emotion-related words and their changing trends over time, analyzes the changes in emotion expression characteristics under different social environments, calculates the rate of change of emotion expression in different time intervals, and obtains emotion expression feature data; The semantic emotion polarity calculation submodule analyzes the semantic emotion polarity of the text content based on the emotion expression feature data, using the formula: ; Calculate semantic sentiment polarity value , analyze the changing trend of individual emotional state and obtain the change record of emotional polarity, where represents the semantic emotional polarity value, represents the weight of sentiment words, represents the polarity score of the corresponding sentiment word, K represents the total number of sentiment words, represents the interaction context weight, represents the interaction frequency, and K represents the number of interaction data samples; The psychological state fluctuation judgment submodule calculates the psychological state fluctuation range of individual students in different time periods based on the emotion polarity change records, analyzes the abnormal fluctuations of individual students' psychological states in combination with the preset mental health warning threshold, judges the rate of change of mental health status, and obtains the analysis results of mental health status changes.

[0011] As a further solution of the present invention, the social intervention matching module includes: The social behavior pattern analysis submodule extracts the interaction frequency, number of interaction objects and interaction duration of individual students, calculates the interaction stability, interaction frequency change rate and group interaction density, and performs pattern classification to obtain the social behavior pattern classification results; The social matching priority calculation submodule builds a social matching weight matrix based on the social behavior pattern classification results and the similarity of the interaction stability, interaction frequency change rate and group interaction density among students, using the formula: ; Calculate the social matching index of student individual i , and sorted according to the matching index to obtain the social matching priority sorting results, among which, represents the matching weight between individual i and individual j, Represents the social mode classification number of individual j, represents the interaction frequency of individual j, T represents the social matching target frequency threshold, and Z represents the overall group size; The personalized social recommendation adjustment submodule excludes individual students with social pressure and isolation behavior based on the social matching priority sorting results and the results of the mental health status change analysis, adjusts the social recommendation relationship, and establishes a personalized social intervention matching record.

[0012] As a further solution of the present invention, the dynamic adaptive optimization module includes: The social interaction trend monitoring submodule obtains the social interaction change trend of the target student based on the personalized social intervention matching record, detects the frequency of friend contact and social activity participation rate, calculates the change value of social activity within the time period, analyzes the current social status change, screens the social trend fluctuation range within the target time window, and obtains social trend change data; The social deviation index calculation submodule adopts the formula based on the social trend change data: ; The social deviation index D is calculated, where Represents the social trend change data recorded for the i-th time, represents the mean of the social trend change parameter, 1 represents the number of data points of the social trend change parameter, represents the mental health score recorded at the jth time, represents the mean of the mental health score, and L represents the number of mental health score data points; The mental health management optimization submodule determines whether the individual student's social status exceeds the fluctuation range of the mental health score based on the social deviation index, adjusts the psychological intervention situation, updates the optimization plan, and obtains a mental health management optimization service plan.

[0013] The AI-based comprehensive mental health service method for colleges and universities includes the following steps: S1: Collect data on college students’ study time, classroom concentration, and homework submission time, calculate the time matching degree of behavior categories, set weights based on psychological questionnaire data, and obtain individual mental health score data; S2: Based on the individual mental health score data, collect the social platform message response time and friend interaction frequency data, calculate the interaction timing matching degree, and obtain the interaction behavior resonance data; S3: Based on the individual mental health score data and interactive behavior resonance data, analyze the emotional expression characteristics, semantic emotion polarity, and negative vocabulary ratio of the social platform, calculate the fluctuation amplitude of the mental state, and calculate the fluctuation trend of psychological pressure in combination with the mental health warning threshold to obtain the analysis results of the change of mental health status; S4: Based on the interactive behavior resonance data, count the groups with similar social patterns, set matching priorities, adjust social recommendations based on the trend of changes in mental health status, and obtain personalized social intervention matching records; S5: Based on personalized social intervention matching records, calculate the social deviation index, adjust the psychological intervention situation according to changes in social interactions, and obtain optimized service plans for mental health management.

[0014] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, based on the behavioral data of learning, work and rest, social interaction, etc., and calculating the time matching degree of behavior categories, the comprehensiveness of individual mental health scores is improved, and combined with social interaction data, the temporal matching degree, topic consistency and interaction synchronization of information exchange are analyzed, the resonance degree of social behavior is analyzed, the psychological fit between individuals is measured, and combined with the changes in semantic emotional polarity, the fluctuation of psychological state is quantified, and data support is provided for the rate of change of mental health status. Based on the analysis of changes in mental health status, individuals with higher social pattern similarity are screened, and social recommendations are adjusted according to the interaction resonance degree to enhance the pertinence of intervention matching, synchronously identify the fluctuation trend of individual social status, and adjust the psychological intervention strategy according to the changes in mental health scores to improve personalized adaptability. It can realize the refined adjustment of mental health management of college students, strengthen the accuracy of social intervention matching, and enhance the timeliness of mental state assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a system flow chart of the present invention; Figure 2 This is a flow chart of the psychological behavior scoring module of the present invention; Figure 3 This is a flow chart of the interactive behavior analysis module of the present invention; Figure 4 This is a flow chart of the mental health assessment module of the present invention; Figure 5 This is a flow chart of the social intervention matching module of the present invention; Figure 6 This is a flow chart of the dynamic adaptive optimization module of the present invention. DETAILED DESCRIPTION

[0016] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0017] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0018] See also Figure 1 , the AI-based comprehensive mental health service system for colleges and universities includes: The psychological behavior scoring module obtains data on college students' study time, classroom concentration, homework submission time, exam preparation time, social interaction frequency, and social network activity. It combines information on rest stability, nighttime activity time, and exercise clock-in frequency, summarizes the frequency and duration trends of each behavior category according to the time window, calculates the time matching degree of different behavior categories, and simultaneously combines psychological questionnaire data to set behavior category weights. It also calculates the psychological health score based on the weight mapping relationship to obtain individual psychological health score data. The interactive behavior analysis module collects data on social platform message response time, friend interaction frequency, group chat activity, topic switching mode, and emoticon usage ratio based on individual mental health score data, calculates the interaction timing matching degree between different students, calculates the consistency of social topics based on the frequency of common topics, dialogue coherence, and information complementarity, and analyzes the synchronization of college students' social behaviors in combination with emoticon usage habits and interaction rhythm stability to obtain interactive behavior resonance data; The mental health assessment module is based on individual mental health score data and interactive behavior resonance data. It counts the emotional expression characteristics, semantic emotion polarity changes, and negative vocabulary proportions of college students on social platforms, calculates the fluctuation amplitude of mental state in different time periods, determines the rate of change of mental health status, and calculates the fluctuation trend of psychological pressure in combination with the preset mental health warning threshold, and outputs the analysis results of mental health status changes; The social intervention matching module counts the student groups with the same similarity in social behavior patterns, sets the social matching priority according to the interactive behavior resonance data, and screens the interactive characteristics of the target students. It also adjusts the social recommendation relationship according to the analysis results of the changes in mental health status, and obtains personalized social intervention matching records. The dynamic adaptive optimization module is based on personalized social intervention matching records, collects the target students' social interaction trends, friend contact frequency, and social activity participation rate, calculates the degree of deviation of social behavior patterns, determines the individual social deviation index, and analyzes the fluctuation range in combination with individual mental health score data. It adjusts the psychological intervention situation according to changes in social interactions and obtains mental health management optimization service plans.

[0019] Individual mental health score data include behavioral consistency scores, psychological questionnaire parameter weights, mental health score weight mapping results, and individual psychological characteristic parameter matching values; interactive behavior resonance data include interactive timing matching, social topic consistency analysis results, and social behavior synchronization analysis results; mental health status change analysis results include mental state fluctuation amplitude, mental health status change rate, psychological stress fluctuation trend analysis results, and negative emotion expression index; personalized social intervention matching records include social matching priority, target student interaction feature screening results, social recommendation adjustment plan, and social support matching relationship; mental health management optimization service plan includes the degree of social behavior pattern deviation, individual social deviation index, mental health score fluctuation range adjustment value, and psychological intervention adaptation adjustment strategy.

[0020] See also Figure 2 , the psychological behavior scoring module includes: The behavior data acquisition submodule acquires data on study time, classroom concentration, homework submission time, exam preparation time, social interaction frequency, social network activity, work and rest stability, nighttime activity time, and exercise clock-in frequency. It summarizes the frequency and duration trends of each behavior category according to the time window to obtain the time distribution data of the behavior category. When obtaining data on study time, classroom concentration, homework submission time, exam preparation time, social interaction frequency, social network activity, work and rest stability, nighttime activity time, and exercise clock-in frequency, data collection is required from multiple sources. Study time can be recorded through the course login time, learning video playback time, and homework submission timestamp of the online learning platform. Classroom concentration can be obtained through the mouse click frequency, screen activity monitoring, and camera facial recognition of the online conference system. The homework submission time is recorded through the timestamp in the online homework system. The exam preparation time can be calculated by the proportion of study time in the days before the exam. For example, if a student's daily study time in the 7 days before the exam is 2.5h, 3.2h, 2.8h, 3.5h, 3.0h, 2.2h, and 3.8h, respectively, then the exam preparation time is calculated as follows: ; The frequency of social interaction can be quantified by the frequency of message sending on social applications and the number of interactions on social platforms. The activity of social networks is counted by the frequency of posting, the number of likes and comments, etc. The stability of daily routine is measured by whether the daily sleep period is within a specific range (such as 22:00-07:00). The nighttime activity time is counted by the online activity duration between 23:00-06:00. The frequency of exercise check-in can be obtained from exercise recording apps and smart wearable device data. After all data are collected, they need to be segmented and counted according to time windows (such as 1 hour, 1 day, 1 week), and then the frequency and duration trends of each behavior category are summarized. Among them, the setting of the time window is adjusted with reference to the work and rest habits and course arrangements of college students, and the basic unit is set to 1 hour. If a behavior does not occur in 3 consecutive time windows, it is classified as a low-frequency behavior. If a behavior occurs in 3 consecutive time windows, it is classified as a high-frequency behavior. The frequency calculation of this type of behavior category affects the subsequent matching weight setting, and finally the time distribution data of the behavior category is obtained.

[0021] The behavior pattern calculation submodule sets the behavior category weight based on the behavior category time distribution data and the psychological questionnaire data, using the formula: ; Calculate the behavior matching deviation value , and combined with the time window to make matching adjustments, and obtain the matching degree of the behavior pattern, represents the weight of the i-th behavior category, represents the actual duration of the i-th behavior category, represents the standard duration of the ith behavior category, and n represents the total number of behavior categories; Based on the time distribution data of behavior categories, we first need to calculate the time matching of different behavior categories. The time matching is defined as the degree of deviation between the actual behavior duration and the standard duration. The standard duration is calculated based on the behavior mean of the same group in the past three months. If the behavior is a high-frequency behavior, outliers that deviate by 1.5 times the standard deviation are excluded when calculating the mean to avoid extreme data affecting the matching calculation. For example, the historical mean of a student's study time is 3.2 hours, and the current data is 2.5 hours. The time matching is calculated as follows: ; Subsequently, the behavioral category weights are set in combination with the psychological questionnaire data. The psychological questionnaire will give an importance score of each behavioral category on mental health. For example, the importance scores of factors such as study time, social interaction frequency, and work and rest stability are 0.3, 0.25, and 0.45, respectively. The weight values ​​are set based on mental health-related research. A large amount of questionnaire survey data is used to analyze the impact of each behavioral category on the mental health score. If the explanatory power of a certain behavioral category in the mental health scoring model exceeds 20%, the weight is greater than 0.3. If the explanatory power of a certain behavioral category in the mental health scoring model is less than 10%, the weight is less than 0.15.

[0022] Taking study time and social interaction frequency as an example, assuming that a student's standard study time is 3.2 hours, the actual value is 2.5 hours, and the standard social interaction frequency is 20 times, the actual value is 15 times, then

[0023] The behavior matching deviation value is calculated and finally matched and adjusted based on the behavior category time distribution data of the time window. If the matching degree of a behavior category is lower than 0.2, the behavior is judged to be abnormal. When the abnormal behavior ratio is higher than 30%, the matching calculation method needs to be readjusted to make the calculation result more consistent with the individual's actual behavior pattern, and finally the behavior pattern matching degree is obtained.

[0024] The mental health score calculation submodule calculates the mental health score based on the behavioral pattern matching degree and the weight mapping relationship to obtain the individual mental health score data; Based on the matching degree of the behavior pattern, the mental health score is further calculated. The score mapping relationship can be obtained by regressing the psychological questionnaire data and the historical mental health assessment data. For example, the mental health score range is set to 0-100 points, and the matching degree is mapped and calculated using the benchmark score curve. The benchmark score curve is constructed based on the mental health survey data of college students in the past five years to ensure that the score range matches the actual mental state. If the matching degree of a certain behavior category is lower than 0.4, the score of this category is additionally adjusted to avoid a single abnormal behavior leading to an overall score that is too low. The scoring formula is: =100×(1- ); The behavior matching deviation value calculated above ;

[0025] Finally, the individual mental health score data is obtained. If the individual mental health score data is lower than 60 points, it is necessary to further analyze the abnormal behavior categories in combination with the psychological questionnaire data, and set personalized adjustment strategies for the abnormal behavior categories, such as increasing social interaction, adjusting work and rest stability, etc., to provide adjustment suggestions that are more in line with the individual's psychological state. The calculation results show that the student's mental health score is 87.7 points, indicating that the mental state is relatively stable.

[0026] See also Figure 3 , the interactive behavior analysis module includes: The social platform interaction feature extraction submodule is based on individual mental health score data, collects social platform message response time, friend interaction frequency, group chat activity, topic switching mode, and emoticon usage ratio data, analyzes the interaction behavior characteristics of different students, extracts the time series characteristics under each data dimension, calculates the mean and variance of the message response time between students, counts the periodicity of friend interaction frequency, analyzes the time distribution law of group chat activity, extracts the conversion probability matrix in the topic switching mode, obtains the distribution curve of the emoticon usage ratio, and obtains the time series characteristics of the interaction behavior; Based on the individual mental health score data, it is necessary to monitor the social platform interaction behavior of different students and extract corresponding features. First, collect the students' message response time on the social platform. The data is calculated by analyzing the students' first response time to private chat and group chat messages. Record the students' message response intervals every day, and calculate their mean and variance to determine the stability of the response time. For example, the message response time of student A is 3 seconds, 5 seconds, 8 seconds, and 2 seconds, respectively. The mean is (3+5+8+2) / 4=4.5 seconds, and the variance is [(3-4.5)²+(5-4.5)²+(8-4.5)²+(2-4.5)²] / 4=5.25 seconds². The higher the value, the faster the student's response time. The fluctuation is large. Among them, the benchmark value of message response time is set as the average response time in the industry. This value is obtained based on the message response data of 5,000 students. The benchmark value is set to 3.5 seconds. When the student's mean deviates from the benchmark value by more than ±1 second, it is considered that there is an abnormal fluctuation in its response time; secondly, the frequency of friend interaction is collected, and the number of interactions between students and friends in a unit time (such as daily) is counted. The periodicity is calculated using time series data. For example, the number of times student B interacts with friends in a week is 10, 15, 14, 18, 9, 20, and 12 times, respectively. The standard deviation and periodic fluctuation value are calculated. If the standard deviation is small, it indicates that the student interaction frequency is relatively regular. If the change is large, it means that there is a sudden change. Changes, on this basis, the threshold of friend interaction frequency is set as the student group with more than 20 interactions or less than 5 interactions per day. This value is based on the behavioral data distribution of students of different age groups and different social habits, reflecting the normal range of interaction; at the same time, the group chat activity is analyzed, the number of students' speeches in different time periods is counted, and the activity distribution curve is constructed. For example, a student's group chat activity is 30 times in the morning, 45 times at noon, and 20 times in the evening. The activity distribution curve shows a trend of being active at noon and decreasing in the evening. The benchmark value of group chat activity is set to the daily average of 30 times. When the student's activity in a certain period of time is higher than 1.5 times or lower than 0.5 times the benchmark value, it indicates that the activity in that period is Then, the topic switching pattern is extracted and the conversation topic conversion probability matrix is ​​constructed. Assuming that the students' conversation involves three main topics A, B, and C, the probability of switching from topic A to topic B, C or staying on topic A is calculated, such as P(A→B)=0.4, P(A→C)=0.3, and P(A→A)=0.3. The matrix can reflect the students' conversation habits and stability. Among them, the conversion weight coefficient of the topic switching pattern is adjusted according to the conversation duration and topic switching frequency of different student groups. For example, if a student's conversation lasts for a long time (such as more than 20 minutes), the topic switching weight is reduced by 0.1. On the contrary, if the conversation duration is short (less than 5 minutes), the topic switching weight is increased by 0.2; Finally, analyze the proportion of expression usage, record the frequency of expression usage of students in each topic, and calculate the proportion. For example, in 100 conversations, student A used expressions 20 times, and the expression usage ratio is 20%. If the expression usage is mainly concentrated in a certain type of expression (such as positive emotions accounting for 70%), it can further reflect the social emotional tendencies of students and finally obtain the temporal characteristics of interactive behavior. .

[0027] The social topic consistency calculation submodule calculates the frequency of common topics based on the time series characteristics of interactive behaviors, counts the number of speaking turns and proportions of different students under common topics, analyzes the coherence of dialogues, calculates the information entropy differences between message sequences, and extracts information complementarity features using the formula: ; Calculating social topic consistency scores ,Get the social topic consistency index, where, represents the weight of the i-th round of dialogue, represents the semantic similarity of the i-th round of dialogue, N represents the total number of dialogue rounds, represents the information entropy of the i-th round, The baseline value of information entropy representing the topic; Based on the temporal characteristics of interactive behaviors, the consistency index of social topics needs to be calculated. First, the number of students’ speaking turns in the common topic is counted, and the number of times different students speak on the same topic is recorded and the proportion is calculated. For example, student C speaks 10 times under topic X, with a total of 50 speaking turns, and his topic proportion is 10 / 50=20%. Secondly, the coherence of the conversation is analyzed and the information entropy difference between message sequences is used for calculation. The information entropy calculation formula is: ; in is the probability of speaking under different topics. If the probability of student D speaking under topics A, B, and C is 0.5, 0.3, and 0.2 respectively, then the information entropy ; The baseline value of information entropy is set at 1.3 to 1.7. If it exceeds this range, it means that the student's topic coherence is low or too concentrated. Then the social topic consistency score is calculated, and the weight coefficient The setting is based on the proportion of students speaking on the topic, if the proportion exceeds 50%.

[0028] To calculate information complementarity, the word vectors of student message content are used to calculate the semantic relevance of different students on the same topic. The semantic relevance is calculated using the cosine similarity formula. If the message vectors of two students are: but:

[0029] Substituting the data, the calculated similarity is 0.8, indicating that the two students are highly similar in this topic. The threshold of semantic relevance is set to 0.7. If it is lower than this value, it is considered that the topic similarity between the two students is low. , and the number of dialogue turns N=5, then calculate: ; The calculated social topic consistency index is 0.812.

[0030] The social behavior synchronization analysis submodule calculates the matching degree of emoticon usage habits based on the social topic consistency index, counts the types and frequencies of emoticon usage of different students under the same topic, calculates the stability of interaction rhythm, constructs the interaction time series distribution matrix by counting the mean and variance of the interaction time interval, evaluates the similarity of interaction rhythm between individual students, and obtains the interaction behavior resonance data; Based on the social topic consistency index, it is necessary to analyze the synchronization of students' social behaviors. First, calculate the matching degree of emoticon usage habits, and count the types and frequencies of emoticon usage of different students under the same topic. For example, student E sent a total of 100 messages under topic Y, including 30 emoticons, and the emoticon usage rate was calculated to be 30%. At the same time, count the proportion of different students using the same emoticon under the topic. If the overlap of expressions of two students reaches 70%, it indicates that the emoticon usage habits are relatively consistent; secondly, calculate the stability of the interaction rhythm, and count the mean and variance of the interaction time interval. For example, the interaction time intervals of student F are 2 seconds, 2 seconds, and 3 seconds, respectively. 4 seconds, 5 seconds, 3 seconds, and 6 seconds, then the mean is 4 seconds and the variance is 2.5 seconds². The threshold of rhythm stability is set to 3 seconds. If the variance is higher than this value, it is considered that the student's interaction rhythm is unstable. If the variance is small, it indicates that the interaction rhythm is relatively stable. Next, an interaction time series distribution matrix is ​​constructed to record the interaction time distribution of different students. For example, in one day, the interaction time of student G is 10 times in the morning, 15 times at noon, and 5 times in the evening, then the matrix is ​​[10, 15, 5]. It is compared with the interaction matrices of other students to calculate the similarity score. Finally, the interaction behavior resonance data is calculated to measure the synchronization of students' social behaviors and obtain the interaction behavior resonance data.

[0031] Table 1 Interaction behavior parameters student Mean response time (seconds) Interaction frequency (times / day) Group chat activity (times) Ratio of emoticon usage (%) A 4.5 12 32 20 B 3.2 15 40 25 As shown in Table 1, the main interaction characteristics of students on social platforms are displayed, and this data can be used for subsequent social behavior analysis.

[0032] See also Figure 4 , the mental health assessment modules include: The emotion expression feature analysis submodule is based on individual mental health score data and interactive behavior resonance data, synchronously obtains text data of college students on social platforms, extracts emotion-related words in the text, counts the proportion of negative words, the frequency of emotion-related words and their changing trends over time, analyzes the changes in emotion expression characteristics in different social environments, calculates the rate of change of emotion expression in different time intervals, and obtains emotion expression feature data; Obtain text data of college students on social platforms, organize the student texts on the social platforms in chronological order, and extract the vocabulary involved in emotional expression, including negative emotional words, positive emotional words and neutral emotional words. At the same time, use the timestamp of the text as the basis for analysis. For example, on a certain social platform, a student posted the content "I feel very bad today" at 8 am. At this time, "bad" can be extracted as a negative emotional word, and the publishing time of the text can be recorded. When counting the frequency of emotion-related words, calculate the proportion of different types of emotional words in a certain time period. For example, within an hour, the student published a total of 10 texts containing emotional expression, of which negative emotional words appeared 6 times, positive emotional words appeared 3 times, and neutral words appeared 1 time, then the proportion of negative words is 60%. Combined with time series analysis, observe the changing trend of emotional vocabulary, further call the interaction data of the social platform, including likes, comments, forwarding and other behaviors, and analyze the changes in emotional expression characteristics in different social environments. For example, the frequency of the student's negative emotional words increases in the time period with less interaction (such as 12:00 a.m. to 5:00 a.m.), and decreases in the time period with more interaction (such as 2:00 p.m. to 5:00 a.m.), then it can be judged that social interaction affects emotional expression. By calculating the rate of change of emotional expression per unit time, setting a fixed time window (such as 30 minutes), and calculating the fluctuation value of emotional words in adjacent windows, the characteristic parameters of emotional expression are finally obtained. In this process, it is necessary to set an emotional fluctuation benchmark value to measure the emotional changes of different students in the same time range. The benchmark value can be calculated based on the student's historical emotional fluctuation mean. For example, the standard deviation of all emotional polarity values ​​in the past 7 days is taken as the benchmark value. If the emotional fluctuation on a certain day exceeds 1.5 times of the benchmark value, the emotional fluctuation is judged to be abnormal. For example, the standard deviation of the emotional polarity of a student for 7 days is 0.4, and the emotional fluctuation benchmark value is set to 0.6. If the emotional polarity value fluctuation in a certain time period of the day reaches 0.7, it is judged that the emotional fluctuation in that time period is large.

[0033] The semantic sentiment polarity calculation submodule analyzes the semantic sentiment polarity of the text content based on the emotion expression feature data, using the formula: ; Calculate semantic sentiment polarity value Analyze the changing trend of individual emotional states and obtain the change records of emotional polarity, where represents the semantic emotional polarity value, represents the weight of sentiment words, represents the polarity score of the corresponding sentiment word, K represents the total number of sentiment words, represents the interaction context weight, represents the interaction frequency, and K represents the number of interaction data samples; Based on the emotion expression feature data, the semantic emotion polarity of the text content is further analyzed, and each emotion-related word is converted into a numerical emotion tendency score. The weight parameters are defined to distinguish the degree of influence of different emotion words on the overall emotion. For example, "happy" is given a positive weight of 0.8, while "angry" is given a negative weight of -0.9. In order to obtain a more accurate emotion polarity measurement, the polarity score of each emotion word is set and calculated in combination with its frequency of occurrence. For example, in a certain period of time, "disappointed" appears 5 times (score -0.7) and "happy" appears 3 times (score 0.8) in the student's text. The calculation formula is as follows: ; Bring in data calculation:

[0034] The calculated semantic sentiment polarity value is 0.2292, and the changing trend of students' individual emotional states is analyzed simultaneously. For example, if a student's emotional polarity value fluctuates between -0.5 and 0.5 within a day, it can be judged that his or her emotions are relatively stable. If the emotional polarity value drops to -0.8 or rises to 0.9 in a certain period of time, it means that the emotions fluctuate violently during that period of time, thereby obtaining a record of emotional polarity changes.

[0035] During the calculation process, the weight parameters are set based on the influence strength of the emotional words, which can be adjusted by the co-occurrence frequency of the emotional words in the historical data and the contexts in which they are used by students. For example, when the proportion of negative emotions exceeds 50%, the weight of the negative emotional words is appropriately increased by 10% to improve the sensitivity of the calculation results. If the default weight of a negative emotional word is -0.8, and the proportion of negative emotions in this period reaches 60%, the adjusted weight is -0.88, so that extreme emotions have a greater impact in the calculation.

[0036] The psychological state fluctuation judgment submodule calculates the fluctuation range of individual students' psychological states in different time periods based on the records of emotional polarity changes, analyzes the abnormal fluctuation of individual students' psychological states in combination with the preset psychological health warning threshold, judges the change rate of psychological health states, and obtains the analysis results of psychological health state changes; Based on the records of emotion polarity changes, the fluctuation range of students' individual psychological states in different time periods is calculated. By setting the mental health warning threshold, the abnormal fluctuations of students' individual psychological states are analyzed. The warning threshold is set based on the individual's historical emotional fluctuation range, and twice the standard deviation of the daily average emotional fluctuation is calculated as the warning line. For example, if the standard deviation of the emotional polarity change of a student in the past 30 days is 0.35, the warning threshold is set to 0.7. If the emotional change amplitude exceeds 0.7 on a certain day, it is marked as high-risk emotional fluctuation. If the emotional polarity change amplitude exceeds this threshold, there is an abnormal psychological state.

[0037] For example, in the monitoring data of a certain day, the student's sentiment polarity value is 0.2 at 9 am, and drops sharply to -0.8 at 11 am. The change range is calculated as follows:

[0038] Since 1.0 is greater than the warning threshold of 0.7, the mood fluctuations during this time period are relatively large and require further attention. The trend of psychological pressure fluctuations can be obtained and compared with historical data to determine the rate of change of mental health status. For example, by comparing the changes in emotional polarity over the past week, if the fluctuation value of emotional polarity change on a certain day is found to be much higher than other times, it can be inferred that the student has greater psychological pressure on that day, and finally the analysis results of mental health status changes can be obtained.

[0039] See also Figure 5 , the social intervention matching module includes: The social behavior pattern analysis submodule extracts the interaction frequency, number of interaction objects and interaction duration of individual students, calculates the interaction stability, interaction frequency change rate and group interaction density, and performs pattern classification to obtain the social behavior pattern classification results; In the social behavior pattern analysis submodule, the social interaction data of the student group is first obtained, including information such as the interaction frequency, number of interaction objects, and interaction duration of each student. For example, a student interacted with classmates 15 times in a week, involving 5 different classmates, and each interaction lasted an average of 10 minutes. Next, the interaction stability of each student is calculated, that is, the degree of fluctuation of the interaction frequency over a period of time. Assuming that the number of interactions of the student in four weeks is 12 times, 15 times, 14 times, and 16 times respectively, the interaction stability can be measured by calculating the standard deviation, and the result is about 1.58, indicating that the interaction frequency of the student is relatively stable. Then, the interaction frequency change rate is calculated to reflect the growth or decrease trend of the interaction frequency. For example, if the student interacted 15 times this week and 12 times last week, the change rate is (15-12) / 12≈0.25, indicating an increase of 25%. In addition, the group interaction density is calculated, that is, the ratio of the actual number of interactions in the entire student group to the maximum number of interactions that may occur. Assuming there are 100 students and the actual number of interactions is 450, the maximum possible number of interactions is 100. (100-1) / 2=4950 times, and the group interaction density is 450 / 4950≈0.091. In this calculation process, the maximum interaction number benchmark value of 4950 used in the density calculation is calculated based on the total number of people in the group. This value shows a quadratic growth trend with the increase of the number of students G, that is, G(G-1) / 2. The density threshold is set to 0.1. According to past data observations, when this value is lower than 0.1, it means less interaction, and higher than 0.3 means active interaction. Finally, students are divided into different social behavior pattern categories, such as high-frequency interaction group and low-frequency interaction group, to obtain the social behavior pattern classification results.

[0040] The social matching priority calculation submodule builds a social matching weight matrix based on the social behavior pattern classification results and the similarity of the interaction stability, interaction frequency change rate and group interaction density among students, using the formula: ; Calculate the social matching index of student individual i , and sorted according to the matching index to obtain the social matching priority sorting results, among which, represents the matching weight between individual i and individual j, Represents the social mode classification number of individual j, represents the interaction frequency of individual j, T represents the social matching target frequency threshold, and Z represents the overall group size; In the social matching priority calculation submodule, the social matching index of each pattern group is calculated based on the social behavior pattern classification results. First, a social matching weight matrix is ​​constructed to measure the matching degree between students. For example, the matching weight between student 1 and student 2 is 0.8, and the matching weight between student 1 and student 3 is 0.5. Then, the matching index of each student is calculated using the formula.

[0041] The target frequency threshold T is set to 12, which is calculated based on the average interaction frequency of the student group. If the average interaction frequency of the student group exceeds 15, the threshold is adjusted to 15. If it is lower than 8, it is adjusted to 10 to accommodate fluctuations in the social activity levels of different groups.

[0042] Assume that there are three students, whose social mode classification numbers are 1, 2, and 3, respectively, and whose interaction frequencies are 10, 15, and 20, respectively. The target frequency threshold is 12, and the matching weight matrix is ​​as follows: Table 2 Student 1 Student 2 Student 3 Student 1 1 0.8 0.5 Student 2 0.8 1 0.6 Student 3 0.5 0.6 1 According to Table 2, for student 1, the matching index is calculated as follows:

[0043] After calculating the matching index of each student, they are sorted to obtain the social matching priority ranking results.

[0044] The personalized social recommendation adjustment submodule excludes individual students with social pressure and isolation behavior based on the social matching priority sorting results and the mental health status change analysis results, adjusts the social recommendation relationship, and establishes a personalized social intervention matching record; In the personalized social recommendation adjustment submodule, based on the social matching priority sorting results, the student combination with a higher matching priority is screened. For example, the top 10% of students in the matching index are selected for further analysis. Combined with the results of the analysis of changes in mental health status, students with recent social pressure or isolation behavior are eliminated. Assuming that among the students screened above, two students have shown social pressure in recent mental health assessments, they will be removed from the recommendation list. The benchmark value for changes in mental health status is set based on the psychological assessment data of the past three months. If the assessment score is lower than 50 (out of 100) and the decline is more than 10 points, it is considered that there is a risk of social pressure. At this time, the recommendation priority of the student is reduced to less than 50%. If the score is lower than 40 and the decline is more than 15 points, it is directly removed from the recommendation list. Finally, adjust the social recommendation relationship, establish a personalized social intervention matching record, and provide a customized social intervention plan for each student.

[0045] See also Figure 6 , the dynamic adaptive optimization module includes: The social interaction trend monitoring submodule obtains the social interaction change trend of the target students based on the personalized social intervention matching records, detects the frequency of contact with friends and the participation rate in social activities, calculates the change value of social activity within the time period, analyzes the current social status changes, screens the social trend fluctuation range within the target time window, and obtains the social trend change data; Based on the personalized social intervention matching records, the social interaction trend monitoring of the target students is obtained by collecting data such as their social activity participation rate, frequency of contact with friends, and overall social interaction change trends. The specific implementation steps are as follows. First, detect the social behavior data of the target students in different time periods, including the frequency of friend interactions on social software, group chat participation, and participation records in face-to-face social activities. For example, a student has 10 private chats with friends, 30 group chat interactions, and 2 offline social activities in a week. These data constitute the student’s initial social behavior data set. Secondly, calculate the change value of its social activity. This value calculates the change in daily social interaction by setting a time window. When the time window is set to 7 days, assuming that the number of interactions on the first day is 20 times and the number of interactions on the seventh day is 10 times, the social activity change value can be expressed as , indicating that social activity has dropped 10 times. Subsequently, the data is compared and analyzed with historical social data to obtain the current social status change rate. For example, historical data shows that the student's average social activity change rate in the past three months is 5%, and the current calculated change rate is -10%. The student's social activity has dropped beyond the historical average. Further screen the social trend fluctuation range within a specific time window, such as calculating the social trend change parameters of all 7-day rolling windows in the past month and determining the range. For example, the minimum change rate is -15% and the maximum change rate is 8%. Then the current change rate of -10% is at a low level. Finally, the social trend change data is obtained for the subsequent calculation of the social deviation index.

[0046] The social deviation index calculation submodule is based on social trend change data and uses the formula: ; The social deviation index D is calculated, where Represents the social trend change data recorded for the i-th time, represents the mean of the social trend change parameter, 1 represents the number of data points of the social trend change parameter, represents the mental health score recorded at the jth time, represents the mean of the mental health score, and L represents the number of mental health score data points; Based on the social trend change data, to calculate the degree of deviation of social behavior patterns, we first need to calculate the baseline value of the individual's historical social status, which is determined by the average of their long-term social behavior. For example, if a student's daily social activity averaged 15 times in the past three months, then his social baseline value is 15. Then, calculate the deviation of the current social status from the baseline value. For example, if a student's social interaction times in the last 7 days are 10, 8, 12, 9, 15, 7, 11, then ; Calculate the deviation of each value from the mean and sum them: ; 1=7, we get: ; Then calculate the standard deviation of the mental health score. The student's mental health scores for the past 7 days are 80, 85, 78, 82, 79, 77, and 81. The mean is: ; Calculate the variance: ; Finally, the social deviation index is calculated: ; The corresponding student’s social deviation index is then obtained as 5.08.

[0047] The mental health management optimization submodule determines whether the individual student's social status exceeds the fluctuation range of the mental health score based on the social deviation index, adjusts the psychological intervention situation, updates the optimization plan, and obtains the mental health management optimization service plan; According to the social deviation index, it is judged whether the individual's social status exceeds the fluctuation range of the mental health score. The fluctuation range of the mental health score is set based on the group mean deviation setting benchmark, that is, a large number of historical mental health score data of students are selected and their mean is calculated. and standard deviation , and then determine the normal fluctuation range. Suppose a statistical analysis is conducted on 200 students in a school to obtain the average mental health score =80, standard deviation =5, the health score fluctuation range is set to [ — , + 】=【75,85】, that is, individuals within this range are considered to have a normal mental health score. If it is lower than 75 or higher than 85, it means that the mental state is abnormal and the intervention plan needs to be adjusted.

[0048] The threshold for judging the social deviation index was set based on the standard deviation multiple of social behavior to define the abnormal range. The social deviation index data of 1,000 students were collected and their mean was calculated. =3 and standard deviation =2.4, and + Thresholds are set in the form of three intervention levels: Mild deviation (0≤D<2): normal social fluctuations, no adjustment of intervention plan; Moderate deviation (2≤D<5.4): Social trends need to be monitored and existing intervention measures maintained; Severe deviation (D≥5.4): Social behavior deviates significantly and requires increased psychological intervention, such as regular social training tasks or psychological counseling.

[0049] Combined with the above calculation results, a student's social deviation index is 5.08, which is in the moderate deviation range, matching his mental health score of 80 points, which is in line with the fluctuation range of mental health scores. Therefore, it is recommended to keep the current psychological intervention measures unchanged and continue to monitor social behavior. Assuming that another student's social deviation index is 6.2, and his mental health score drops to 74, the system determines that his social status is beyond the normal range and requires social intervention adjustments, such as increasing regular social training tasks or the frequency of psychological counseling, and then obtaining a mental health management optimization service plan.

[0050] The AI-based comprehensive mental health service method for colleges and universities includes the following steps: S1: Collect data on college students’ study time, classroom concentration, and homework submission time, calculate the time matching degree of behavior categories, set weights based on psychological questionnaire data, and obtain individual mental health score data; S2: Based on the individual mental health score data, collect the social platform message response time and friend interaction frequency data, calculate the interaction timing matching degree, and obtain the interaction behavior resonance data; S3: Based on the individual mental health score data and interactive behavior resonance data, analyze the emotional expression characteristics, semantic emotion polarity, and negative vocabulary ratio of the social platform, calculate the fluctuation amplitude of the mental state, and calculate the fluctuation trend of psychological pressure in combination with the mental health warning threshold to obtain the analysis results of the change of mental health status; S4: Based on the interactive behavior resonance data, count the groups with similar social patterns, set matching priorities, adjust social recommendations based on the trend of changes in mental health status, and obtain personalized social intervention matching records; S5: Based on personalized social intervention matching records, calculate the social deviation index, adjust the psychological intervention situation according to changes in social interactions, and obtain optimized service plans for mental health management.

[0051] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. AI-based comprehensive mental health service system for colleges and universities, characterized by: The system comprises: The psychological behavior scoring module obtains personal data of college students, calculates the time matching degree of behavior categories, sets behavior category weights and calculates mental health scores based on the weight mapping relationship to obtain individual mental health score data; The interactive behavior analysis module obtains social platform data based on the individual mental health score data, calculates the interaction time sequence matching degree and topic consistency, analyzes the behavior synchronization, and obtains the interactive behavior resonance data; The mental health assessment module obtains the emotional expression change data of the social platform based on the individual mental health score data and the interactive behavior resonance data, calculates the fluctuation amplitude of the mental state and the pressure fluctuation trend, and obtains the analysis results of the mental health state change; The social intervention matching module counts the student groups with the same similarity in social behavior patterns, sets social matching priorities, and screens interaction features, adjusts social recommendation relationships, and obtains personalized social intervention matching records; The dynamic adaptive optimization module obtains the target student's social interaction change data based on the personalized social intervention matching records, calculates the social behavior pattern deviation index, adjusts the psychological intervention situation, and obtains a mental health management optimization service plan.

2. The AI-based comprehensive college mental health service system according to claim 1 is characterized in that: The individual mental health score data includes behavior consistency score, psychological questionnaire parameter weight, mental health score weight mapping result, and individual psychological characteristic parameter matching value; the interactive behavior resonance data includes interaction timing matching degree, social topic consistency analysis result, and social behavior synchronization analysis result; the mental health state change analysis result includes mental state fluctuation amplitude, mental health state change rate, psychological stress fluctuation trend analysis result, and negative emotion expression index; the personalized social intervention matching record includes social matching priority, target student interaction feature screening result, social recommendation adjustment plan, and social support matching relationship; The mental health management optimization service plan includes the degree of social behavior pattern deviation, individual social deviation index, mental health score fluctuation range adjustment value, and psychological intervention adaptation adjustment strategy.

3. The AI-based comprehensive college mental health service system according to claim 1 is characterized in that: The psychological behavior scoring module includes: The behavior data acquisition submodule acquires data on study time, classroom concentration, homework submission time, exam preparation time, social interaction frequency, social network activity, work and rest stability, nighttime activity time, and exercise clock-in frequency. It summarizes the frequency and duration trends of each behavior category according to the time window to obtain the time distribution data of the behavior category. The behavior pattern calculation submodule sets the behavior category weight based on the behavior category time distribution data and the psychological questionnaire data, using the formula: ; Calculate the behavior matching deviation value , and combined with the time window to make matching adjustments, and obtain the matching degree of the behavior pattern, represents the weight of the i-th behavior category, represents the actual duration of the i-th behavior category, represents the standard duration of the ith behavior category, and n represents the total number of behavior categories; The mental health score calculation submodule calculates the mental health score based on the behavior pattern matching degree and the weight mapping relationship to obtain individual mental health score data.

4. The AI-based comprehensive college mental health service system according to claim 1 is characterized in that: The interactive behavior analysis module includes: The social platform interaction feature extraction submodule collects data on social platform message response time, friend interaction frequency, group chat activity, topic switching mode, and emoticon usage ratio based on the individual mental health score data, analyzes the interaction behavior characteristics of different students, extracts the time series characteristics under each data dimension, calculates the mean and variance of the message response time between students, counts the periodicity of friend interaction frequency, analyzes the time distribution law of group chat activity, extracts the conversion probability matrix in the topic switching mode, obtains the distribution curve of the emoticon usage ratio, and obtains the interaction behavior time series characteristics; The social topic consistency calculation submodule calculates the frequency of common topics based on the temporal characteristics of the interactive behavior, counts the number of speaking turns and proportions of different students under the common topics, analyzes the coherence of the dialogue, calculates the information entropy difference between message sequences, and extracts the information complementarity features. The formula is: ; Calculating social topic consistency scores ,Get the social topic consistency index, where, represents the weight of the i-th round of dialogue, represents the semantic similarity of the i-th round of dialogue, N represents the total number of dialogue rounds, represents the information entropy of the i-th round, The baseline value of information entropy representing the topic; The social behavior synchronization analysis submodule calculates the matching degree of emoticon usage habits based on the social topic consistency index, counts the types and frequencies of emoticon usage by different students under the same topic, calculates the stability of the interaction rhythm, constructs an interaction timing distribution matrix by counting the mean and variance of the interaction time interval, evaluates the similarity of the interaction rhythm between individual students, and obtains the interaction behavior resonance data.

5. The AI-based comprehensive college mental health service system according to claim 1 is characterized in that: The mental health assessment module includes: The emotion expression feature analysis submodule synchronously obtains the text data of college students on the social platform based on the individual mental health score data and the interactive behavior resonance data, extracts emotion-related words in the text, counts the proportion of negative words, the frequency of occurrence of emotion-related words and their changing trends over time, analyzes the changes in emotion expression characteristics under different social environments, calculates the rate of change of emotion expression in different time intervals, and obtains emotion expression feature data; The semantic emotion polarity calculation submodule analyzes the semantic emotion polarity of the text content based on the emotion expression feature data, using the formula: ; Calculate semantic sentiment polarity value , analyze the changing trend of individual emotional state and obtain the change record of emotional polarity, where represents the semantic emotional polarity value, represents the weight of sentiment words, represents the polarity score of the corresponding sentiment word, K represents the total number of sentiment words, represents the interaction context weight, represents the interaction frequency, and K represents the number of interaction data samples; The psychological state fluctuation judgment submodule calculates the psychological state fluctuation range of individual students in different time periods based on the emotion polarity change records, analyzes the abnormal fluctuations of individual students' psychological states in combination with the preset mental health warning threshold, judges the rate of change of mental health status, and obtains the analysis results of mental health status changes.

6. The AI-based comprehensive college mental health service system according to claim 1 is characterized in that: The social intervention matching module includes: The social behavior pattern analysis submodule extracts the interaction frequency, number of interaction objects and interaction duration of individual students, calculates the interaction stability, interaction frequency change rate and group interaction density, and performs pattern classification to obtain the social behavior pattern classification results; The social matching priority calculation submodule builds a social matching weight matrix based on the social behavior pattern classification results and the similarity of the interaction stability, interaction frequency change rate and group interaction density among students, using the formula: ; Calculate the social matching index of student individual i , and sorted according to the matching index to obtain the social matching priority sorting results, among which, represents the matching weight between individual i and individual j, Represents the social mode classification number of individual j, represents the interaction frequency of individual j, T represents the social matching target frequency threshold, and Z represents the overall group size; The personalized social recommendation adjustment submodule excludes individual students with social pressure and isolation behavior based on the social matching priority sorting results and the results of the mental health status change analysis, adjusts the social recommendation relationship, and establishes a personalized social intervention matching record.

7. The AI-based comprehensive college mental health service system according to claim 1 is characterized in that: The dynamic adaptive optimization module includes: The social interaction trend monitoring submodule obtains the social interaction change trend of the target student based on the personalized social intervention matching record, detects the frequency of friend contact and social activity participation rate, calculates the change value of social activity within the time period, analyzes the current social status change, screens the social trend fluctuation range within the target time window, and obtains social trend change data; The social deviation index calculation submodule adopts the formula based on the social trend change data: ; The social deviation index D is calculated, where Represents the social trend change data recorded for the i-th time, represents the mean of the social trend change parameter, 1 represents the number of data points of the social trend change parameter, represents the mental health score recorded at the jth time, represents the mean of the mental health score, and L represents the number of mental health score data points; The mental health management optimization submodule determines whether the individual student's social status exceeds the fluctuation range of the mental health score based on the social deviation index, adjusts the psychological intervention situation, updates the optimization plan, and obtains a mental health management optimization service plan.

8. The AI-based comprehensive mental health service method for colleges and universities is characterized by: The system according to any one of claims 1 to 7 is implemented, comprising the following steps: S1: Collect data on college students’ study time, classroom concentration, and homework submission time, calculate the time matching degree of behavior categories, set weights based on psychological questionnaire data, and obtain individual mental health score data; S2: Based on the individual mental health score data, collect the social platform message response time and friend interaction frequency data, calculate the interaction timing matching degree, and obtain the interaction behavior resonance data; S3: Analyze the emotional expression characteristics, semantic emotion polarity, and negative vocabulary ratio of the social platform according to the individual mental health score data and the interactive behavior resonance data, calculate the fluctuation amplitude of the mental state, and calculate the fluctuation trend of psychological pressure in combination with the mental health warning threshold to obtain the analysis results of the change of mental health status; S4: Based on the interactive behavior resonance data, count the groups with similar social patterns, set matching priorities, adjust social recommendations based on the trend of changes in mental health status, and obtain personalized social intervention matching records; S5: Based on the personalized social intervention matching records, the social deviation index is calculated, the psychological intervention situation is adjusted according to the changes in social interaction, and the mental health management optimization service plan is obtained.

Images