College faculty worker and student mental health monitoring and intervention method and system

Abstract

The invention provides a college faculty worker and student psychological health monitoring and intervention method and system, and relates to a college ideological and political care system, and the method comprises the steps: employing a Monte Carlo simulation method to generate a large number of negative emotion propagation paths, carrying out the statistics of the coverage range and influence of negative emotion nodes, determining the influence range and key propagation nodes of negative emotion, and carrying out the calculation of the negative emotion propagation paths. Quantitative indexes reflecting the propagation penetration degree of the negative emotion among different communities are obtained, wherein the quantitative indexes comprise the negative emotion propagation speed, the influence range and the duration time; in a college ideological and political education system, positive emotion information is pushed to a target community and key nodes through a social network platform, the negative emotion propagation dynamic state is monitored, according to the real-time change of emotion distribution in an emotion propagation network, an emotion intervention strategy is dynamically adjusted, diffusion of negative emotions is controlled, and community emotion management measures are enhanced; and community emotion interaction mode analysis is perfected.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method and system for monitoring and intervening in the mental health of university faculty and students. Background Art

[0002] In the ideological and political education system of colleges and universities, the mental health of college faculty and students is directly related to the stability and harmony of the campus. However, in a complex campus environment, the spread of negative emotions is often hidden and sudden, and traditional mental health monitoring methods are difficult to detect and effectively intervene in time. Especially in today's prevalence of social networks, negative emotions may spread rapidly through virtual social circles, causing a wider range of impacts. The spread of negative emotions in college social networks involves multiple interrelated technical difficulties: First, the social network structure of college faculty and students is complex and changeable, and the emotional propagation characteristics of different communities are different, making it difficult to build a unified network topology model. Secondly, the emotional propagation process is affected by many factors, including individual emotional expression, community characteristics, network structure, etc., and it is unclear how to quantify the impact of these factors on emotional propagation. Furthermore, the propagation path and scope of influence of negative emotions are highly uncertain, and it is difficult to accurately predict their diffusion trend. Finally, different communities and individuals have different susceptibility and resistance to negative emotions, and how to formulate targeted intervention strategies faces challenges. These technical difficulties are interrelated and together constitute the core technical issues of predicting and intervening in the spread of negative emotions in colleges and universities. How to accurately predict and effectively intervene in the spread of negative emotions in the complex and ever-changing social networks of universities, and then monitor and intervene in the mental health of university faculty and students, is a key issue that needs to be urgently addressed. Summary of the invention

[0003] The present invention provides a method for monitoring and intervening in the mental health of university faculty and students, which mainly includes:

[0004] Obtain user relationship data and emotional expression data of university faculty and students on social networking platforms, and count the emotional density attributes of members of different communities, including the richness of emotional vocabulary and the proportion of different types of emotions;

[0005] Use sentiment analysis algorithms to identify negative emotions in sentiment expression data, build sentiment propagation network topology based on user relationship data, use community discovery algorithms to identify different communities, calculate sentiment density attributes within each community, and sentiment propagation direction, propagation intensity, and attenuation rate between communities;

[0006] A multi-factor correlation model of emotional communication is established in combination with the communication intensity attribute, which includes community size, number of communication paths, and average path length. The complexity of emotional interaction between communities is quantitatively evaluated to obtain quantitative indicators of the degree of communication penetration between different communities.

[0007] In the topological structure of the emotion propagation network, the key nodes on the propagation path are combined to evaluate the impact of community structure on emotion propagation. Different community structures are divided into different emotional states through clustering algorithms. The transition probability matrix of each emotional state is calculated based on the quantitative index of the degree of emotion propagation penetration among different communities. The emotion propagation efficiency under different community structures is calculated based on the transition probability matrix.

[0008] A large number of negative emotion propagation paths are generated using the Monte Carlo simulation method, the coverage and influence of negative emotion nodes are counted, the impact range and key propagation nodes of negative emotions are determined, and quantitative indicators reflecting the degree of propagation and penetration of negative emotions among different communities are obtained;

[0009] In view of the impact range and key transmission nodes of negative emotions, according to the quantitative analysis results of community emotional susceptibility and resistance, the target communities and key nodes for emotional intervention are determined, differentiated emotional intervention strategies are formulated, and community operation strategies are optimized. The community emotional susceptibility is the tendency of the community to accept and spread certain emotions, which is calculated by analyzing historical emotional transmission data. The community emotional resistance is the ability of the community to resist the spread of certain emotions, which is quantified by measuring the speed of emotional transmission and the degree of influence restriction;

[0010] Push positive emotional information to target communities and key nodes through social networking platforms, monitor the dynamics of negative emotional propagation, dynamically adjust emotional intervention strategies based on real-time changes in emotional distribution in the emotional propagation network, control the spread of negative emotions, strengthen community emotional management measures, and improve community emotional interaction pattern analysis.

[0011] The present invention provides a system for monitoring and intervening the mental health of college faculty and students, which mainly includes:

[0012] Data collection and emotion calculation module, used to obtain and analyze user emotion data;

[0013] Community emotion propagation network construction module, used to construct and analyze emotion propagation networks;

[0014] Emotional communication model construction and analysis module, used to build emotional communication models and analyze communication rules;

[0015] The negative emotion transmission simulation and intervention module is used to simulate and intervene in the transmission of negative emotions.

[0016] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0017] The invention discloses a method and system for monitoring and intervening the mental health of university faculty and students. The method constructs an emotion propagation network topology by acquiring user relationships and emotion expression data between university faculty and students on a social network platform. An emotion analysis algorithm is used to identify negative emotions, a community discovery algorithm is used to identify different communities, and the emotion density attributes within the community and the emotion propagation characteristics between communities are calculated. A multi-factor association model for emotion propagation is established in combination with the propagation intensity attribute to evaluate the impact of community structure on emotion propagation. The community is divided into different emotional states by an emotion clustering algorithm, and the emotional state transition probability matrix is ​​calculated. A Monte Carlo simulation method is used to generate a negative emotion propagation path, and its influence range and key propagation nodes are determined. According to the analysis results of community emotional susceptibility and resistance, a differentiated emotion intervention strategy is formulated, positive emotion information is pushed through a social platform, intervention measures are dynamically adjusted, and the spread of negative emotions is controlled. The invention realizes accurate analysis and effective intervention of university community emotions, and provides technical support for optimizing community operations and monitoring and intervening the mental health of university faculty and students in the ideological and political education system of universities. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 The present invention is a flow chart of a method for monitoring and intervening in the mental health of university faculty and students.

[0019] Figure 2 It is a schematic diagram of a method for monitoring and intervening in the mental health of university faculty and students of the present invention.

[0020] Figure 3 This is another schematic diagram of a method for monitoring and intervening in the mental health of university faculty and students of the present invention. DETAILED DESCRIPTION

[0021] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] like Figure 1-3 In this embodiment, a method for monitoring and intervening in the mental health of college faculty and students may specifically include:

[0023] S101. Obtain user relationship data and emotional expression data of university faculty and student groups on social network platforms. Count the emotional density attributes of its members for different communities, including emotional vocabulary richness and the proportion of different types of emotions. The emotional vocabulary richness is calculated by dividing the number of preset emotional words used by each user by the total vocabulary of the user. The proportion of different types of emotions is calculated by counting the frequency of occurrence of each type of emotional vocabulary.

[0024] The social platform user identification is obtained for the university teacher group and the student group, and the target user text content is obtained from the social platform interface according to the user identification; the target user text content is segmented by a preset emotion dictionary, and six emotion labels of joy, anger, sadness, fear, disgust and surprise are added to the text content after the segmentation by a maximum entropy algorithm to obtain a user emotion vocabulary; the number of each type of emotion vocabulary is counted according to the user emotion vocabulary, and the user emotion vocabulary richness value is obtained by dividing the number of each type of emotion vocabulary by the total number of emotion vocabulary; the group emotion density distribution is calculated for the user emotion vocabulary richness value, and a group emotion distribution matrix is ​​generated according to the group emotion density distribution, and standardized group emotion distribution data is obtained through normalization processing.

[0025] Specifically, user identification data is collected on the social platform for the university teacher group and the student group, the text content corresponding to the target user identification is obtained from the social network platform interface, and user dictionary data is created according to the text content. The six types of emotion tags including joy, anger, sadness, fear, disgust, and surprise in the preset emotion dictionary are obtained from the database, the user text content is segmented in combination with the emotion dictionary data, and the maximum entropy algorithm is used to add emotion tags to the segmented data, and a user emotion vocabulary is established through the user emotion tag data. The number of different types of emotion words in the user emotion vocabulary is counted according to the user identification, the total number of emotion words is obtained by accumulating the number of each type of emotion words according to the counter, the user emotion vocabulary richness value is obtained by dividing the number of emotion words by the total number of words, and the community is divided according to the emotion vocabulary richness value of the user group through random forest. The group emotion density distribution is calculated for the emotion vocabulary richness value of the university teacher group and the student group, and the group emotion distribution matrix is ​​generated according to the proportion of each type of emotion words, and the standardized group emotion distribution data is obtained through normalization. The user identity information of the social platform is based on the corresponding relationship between the teacher's work number and the student's number. The user identification uses hash mapping to generate a unique code. For example, the teacher's work number 2023001 is mapped to the user identification hash_t_001, and the student's work number 20230101 is mapped to the user identification hash_s_001. The identification code is used to distinguish different groups of users. When obtaining text content, the original content and comment content posted by the user on the social platform are extracted, and the forwarded content and advertising content are removed to establish a user original text corpus. The user dictionary segments the original text based on the word segmentation tool, and records the word frequency and part of speech information. The emotion dictionary is constructed using psychology annotated corpus, which contains 6 basic emotion types. Each emotion type has multiple subcategories. For example, joy emotions include happy, happy, excited and other subcategories, and anger emotions include angry, hateful, indignant and other subcategories. Each entry in the emotion dictionary contains three attributes: word, emotion type, and emotion intensity. For example, the emotion type of the "happy" entry is joy, and the emotion intensity is 0.8. The maximum entropy algorithm combines the word context information in the annotation process to improve the accuracy of emotion annotation. The richness of user emotional vocabulary reflects the diversity of vocabulary used by users when expressing emotions. For example, if user A uses 20 different emotional vocabulary and the total vocabulary is 200, the richness of emotional vocabulary is 0.1. User B uses 40 different emotional vocabulary and the total vocabulary is 300, so the richness of emotional vocabulary is 0.13. Random forest divides user groups into communities based on the richness of emotional vocabulary characteristics, and identifies users with rich emotional expressions and users with conservative emotional expressions. The group emotional density distribution is obtained by counting the frequency of use of various emotional vocabulary in the community. For example, the vocabulary used by teachers to express joy accounts for 35%, the vocabulary used to express anger accounts for 15%, and the vocabulary used to express sadness accounts for 20%.The proportion of words used by students to express joy is 45%, the proportion of words used to express anger is 25%, and the proportion of words used to express sadness is 10%. The influence of the difference in the number of samples of different groups on the statistical results is eliminated through normalization processing, and the standardized group emotion distribution data is obtained. Normalization uses the maximum and minimum value method to map the proportion of various emotional words to the range of 0 to 1. In the teacher group, the normalized joy emotion density is 0.7, the anger emotion density is 0.3, and the sadness emotion density is 0.4. In the student group, the normalized joy emotion density is 0.9, the anger emotion density is 0.5, and the sadness emotion density is 0.2.

[0026] S102, using sentiment analysis algorithms to identify negative sentiments in sentiment expression data. Constructing sentiment propagation network topology based on user relationship data. Using community discovery algorithms to identify different communities, calculating sentiment density attributes within each community and sentiment propagation direction, propagation intensity, and attenuation rate between communities. The attenuation rate is the degree to which sentiment intensity decreases with propagation distance, and is calculated by comparing the changes in sentiment intensity at different propagation distances.

[0027] Negative emotion words in the text are obtained, and the emotion word feature vectors are extracted according to the word embedding method. The negative emotion training samples are annotated with the emotion dictionary to obtain the negative emotion score. For the negative emotion score, a social communication graph is constructed through the attention and interaction data between user identifiers, and the nodes with negative emotion intensity values ​​in the social communication graph are divided into communities using the label propagation algorithm. A propagation time series link is constructed according to the release time and forwarding time of the nodes in the social communication graph, and the Dijkstra algorithm is used to calculate the shortest path distance between the nodes, and the propagation intensity data is obtained through the negative emotion intensity value and the propagation path length. For the propagation time series link, the difference calculation of the negative emotion intensity value of the propagation path is performed, and the exponential function curve is used to fit the change of negative emotion intensity at different path distances, and the negative emotion attenuation rate is calculated through the propagation path length.

[0028] Specifically, for the four types of negative emotional words including anger, sadness, fear and disgust in user text data, the feature vector of each word is extracted through word embedding, the negative emotion training samples are annotated by the emotional dictionary, the negative emotion score of the user text is calculated from the neural network, and the text negative emotion intensity value is obtained according to the position relationship and word frequency of the emotional words. A social communication graph is constructed based on the attention and interaction data between user identifiers, and the negative emotion propagation relationship within a fixed time window is set on the graph structure. The nodes with negative emotion intensity values ​​are divided into communities by using label propagation, and the community negative emotion density value is obtained by the number of direct connections and edge weights of negative emotion nodes. The node release time and forwarding time are extracted from the negative emotion propagation graph to construct a propagation time series link, the shortest path distance between nodes is calculated according to the Dijkstra algorithm, the propagation intensity data is generated by the negative emotion intensity value and the propagation path length, and the negative emotion propagation direction is calculated for each propagation link. The difference calculation of the negative emotion intensity value of the propagation path is performed within the fixed propagation time window, and the exponential function curve is used to fit the change of negative emotion intensity at different path distances. The negative emotion attenuation rate is calculated by the propagation path length, and the cross-community propagation direction data is obtained by comparing the community negative emotion density values. Negative emotion words present different combination features in user texts. Each word is mapped to a 300-dimensional feature space through word embedding. In the feature space, similar emotion words are close to each other, while different emotion words are far away from each other. In the emotion dictionary, anger words are annotated with an intensity value of 0.8, sadness words are annotated with an intensity value of 0.6, fear words are annotated with an intensity value of 0.7, and disgust words are annotated with an intensity value of 0.5. During the training process, the neural network learns the correspondence between word features and emotion intensity, and calculates the comprehensive emotion intensity value for sentences containing multiple negative emotion words. The social communication graph is constructed through the attention relationship and interactive behavior between users. The nodes represent users and the edges represent the communication channels between users. Within a fixed time window of 24 hours, if user A posts negative emotion content and his fan user B reposts or comments on the content, a directed communication edge is established between user A and user B. Label propagation starts from the seed node with higher negative emotion intensity and realizes community division by iteratively updating the labels of adjacent nodes. The density value of negative emotions in a community reflects the degree of aggregation of negative emotions among members within the community. A density value of 0.8 indicates that negative emotions are actively propagated within the community. The propagation time series link records the diffusion process of negative emotional content among users, including information such as publishing time, forwarding time, and propagation path. For example, user A publishes negative emotional content at 10:00, user B forwards it at 10:30, and user C forwards it at 11:00, forming a propagation link of A→B→C. The Dijkstra algorithm calculates the shortest path distance between user nodes. A distance of 2 means that negative emotions need to pass through 2 user nodes to propagate. The propagation intensity weakens as the path distance increases. The intensity from the source user to the first-degree propagation user is 0.9, and the intensity to the second-degree propagation user drops to 0.6.Negative emotions show attenuation characteristics during the propagation process, and the attenuation data is obtained by calculating the difference in emotion intensity between adjacent propagation nodes. Within the 3-hop propagation distance, the intensity of negative emotions decays from the initial value of 1.0 to 0.7, 0.4, and 0.2, showing an exponential decay trend. There are differences in the propagation characteristics of different communities. The decay rate of high-density communities is slower, and the spread of negative emotions is wider; the decay rate of low-density communities is faster, and the spread of negative emotions is limited. The cross-community propagation direction is related to the density difference between communities. Negative emotions tend to spread from high-density communities to low-density communities.

[0029] S103. Establish a multi-factor association model of emotion transmission in combination with the transmission intensity attribute, wherein the transmission intensity attribute includes community size, number of transmission paths, and average path length, quantitatively evaluate the complexity of emotional interaction between communities, and obtain a quantitative index of the degree of transmission penetration between different communities. The multi-factor association model combines community size, number of transmission paths, and average path length to generate a complexity score.

[0030] The connection data between nodes in the community interaction network is obtained, and the number of nodes inside the community is counted according to the connection data to obtain the community scale value; a community propagation adjacency matrix is ​​generated for the node connection data, and the product of the node propagation ratio and the propagation path set is calculated through the community propagation adjacency matrix to obtain the propagation intensity parameter; a multidimensional feature tensor is constructed according to the propagation intensity parameter, and a neural network is used to extract the product of the community scale feature vector and the path feature vector from the multidimensional feature tensor to obtain the community interaction complexity parameter; a community penetration matrix is ​​constructed according to the community interaction complexity parameter, and if the ratio of the number of source community nodes to the number of target community nodes multiplied by the propagation path length exceeds a preset threshold, a corresponding community penetration level indicator is generated.

[0031] Specifically, node connection data is extracted from the community interaction network, the community size value is determined by counting the number of nodes inside each community, the number of all shortest propagation paths between communities within a fixed time window is counted according to the Bellman-Ford algorithm, and the average propagation path length is obtained by calculating the mean of the path length. The community propagation adjacency matrix is ​​generated for the node identifier, and the set of all propagation paths between nodes is obtained by breadth-first traversal. The node propagation ratio is calculated by dividing the out-degree value by the in-degree value, and the propagation intensity parameter is obtained by multiplying the propagation ratio by the path set size. A multidimensional feature tensor is constructed for the propagation intensity parameter, and a neural network is used to extract the community size feature vector, the path number feature vector, and the average path length feature vector from the feature tensor. The community interaction complexity parameter is calculated by multiplying the community size feature vector and the path feature vector. The community penetration matrix is ​​constructed based on the community interaction complexity parameter, and the penetration benchmark value is calculated by the ratio of the number of source community nodes to the number of target community nodes. The penetration degree value between communities is obtained by multiplying the penetration benchmark value by the propagation path length, and the community penetration level index is generated according to the penetration degree value. The node connections in the social interaction network show hierarchical propagation characteristics, and the nodes establish connections through behaviors such as following, commenting, and forwarding. For community A containing 500 nodes and community B containing 300 nodes, 45 shortest propagation paths were found in a 24-hour time window through the Bellman-Ford algorithm, with an average path length of 3.2, indicating that information usually needs to pass through 3 to 4 nodes to propagate between the two communities. The community propagation adjacency matrix records the direct propagation relationship between nodes. The value 1 in the matrix indicates that there is a propagation link, and the value 0 indicates that there is no propagation link. Breadth-first traversal starts from the source node and discovers all possible propagation paths hierarchically. In community A, node X propagates outward to 10 nodes and receives propagation from 5 nodes. The propagation ratio is 2.0, which is multiplied by the 25 propagation paths in which the node participates to obtain the propagation strength parameter of 50.0. The propagation strength parameter reflects the propagation influence of the node in the network. The multidimensional feature tensor integrates the propagation characteristics of three dimensions: community size, number of paths, and path length. The neural network extracts the community scale feature vector [500,300], the path number feature vector

[45] , and the average path length feature vector [3.2] from the feature tensor. The interaction complexity parameter of community A and community B is 0.85. The larger the value, the more complex the communication relationship between communities. The interaction complexity parameter comprehensively reflects the differences in community scale, the diversity of communication paths, and the characteristics of communication distance. The community penetration matrix describes the phenomenon of information transmission and penetration between different communities. The penetration baseline value of community A to community B is 1.67, which means that the number of nodes in the source community is 1.67 times that of the target community. Combined with the average communication path length of 3.2, the penetration degree value is calculated to be 5.34. A penetration degree value greater than 5.0 is classified as high penetration, between 3.0 and 5.0 is classified as moderate penetration, and less than 3.0 is classified as low penetration.The penetration level index reflects the scope and intensity of information dissemination between communities. High penetration indicates that the source community has significant information penetration on the target community. In practical applications, community A has a high penetration feature on community B, indicating that the larger community A occupies a dominant position in the dissemination process, and information is more likely to penetrate from community A to community B.

[0032] S104. In the topological structure of the emotion propagation network, the influence of the community structure on the emotion propagation is evaluated in combination with the key nodes on the propagation path. Different community structures are divided into different emotional states through clustering algorithms. The transition probability matrix of each emotional state is calculated based on the quantitative index of the degree of penetration of emotions among different communities, and the emotion propagation efficiency under different community structures is calculated based on the transition probability matrix. The transition probability matrix represents the probability of changing from one emotional state to another.

[0033] The number of connections between the upper and lower four layers of neighboring nodes of the network node is obtained, the node betweenness value is calculated according to the number of connections, and the key propagation node is identified through the node betweenness value; for the path connectivity between the key propagation nodes, a hierarchical clustering method is used to construct a community structure feature matrix, and the community structure feature matrix includes the internal connectivity rate and the number of external edges of the community; according to the emotional expression data of the community structure feature matrix, a probability state machine is used to record the emotional state conversion data, and the cross-community emotional propagation amount is obtained through the community penetration degree value; a state transition probability matrix is ​​generated for the cross-community emotional propagation amount and the emotional state conversion data, and the state balance coefficient is obtained by dividing the number of positive emotional state nodes by the number of negative emotional state nodes, and the propagation efficiency value is calculated according to the state balance coefficient and the state transition probability matrix.

[0034] Specifically, the node in-degree and out-degree values ​​are extracted from the topological structure of the emotion propagation network, the node betweenness is calculated by the number of connections between the upper and lower four-layer neighbor nodes, the key propagation nodes are identified according to the node betweenness values, and the community aggregation values ​​are calculated based on the path connectivity between the key nodes in each community. According to the community aggregation values ​​and emotion propagation data, hierarchical clustering is used to divide the three structures of isolated communities, intermediate communities, and clustered communities. The community structure feature matrix is ​​constructed by the internal connectivity rate and the number of external edges of the community. The structure feature matrix is ​​divided into positive emotional state, negative emotional state, and neutral emotional state according to the emotional expression data of each community in a fixed time window. The community structure feature matrix is ​​segmented and counted according to the 24-hour time window, and the probability state machine is used to record the emotional state conversion data in two adjacent time windows. The amount of cross-community emotional propagation is extracted from the community penetration value, and the emotional state transition probability matrix is ​​generated according to the propagation amount and state conversion data. The transition ratio between adjacent states is calculated for the emotional state transition probability matrix. The state balance coefficient is obtained by dividing the number of nodes in the positive emotional state by the number of nodes in the negative emotional state. The transmission efficiency value is obtained by multiplying the state balance coefficient by the transition probability matrix. The structural impact of the transmission efficiency value under different community structures is compared. In the emotional communication network, the in-degree and out-degree of the node reflect the information dissemination ability. For example, the in-degree of node A is 15 and the out-degree is 5, which means that it receives information input from 15 users and disseminates information to 5 users. The information flow capacity is evaluated by counting the number of connections between the four layers of neighbors of the node. Node B has 20 first-layer neighbors, 35 second-layer neighbors, 50 third-layer neighbors, and 80 fourth-layer neighbors. The calculated betweenness value is 0.8, which exceeds the preset threshold of 0.7 and is determined to be a key communication node. The community aggregation value is calculated by the number of shortest paths between key nodes. The more paths there are, the stronger the internal connectivity of the community. The community structure is divided into three types according to the aggregation characteristics. The internal connectivity rate of the isolated community is lower than 0.3, and there is a lack of communication paths between key nodes; the internal connectivity rate of the intermediate community is between 0.3 and 0.7, and it has the ability to spread across communities; the internal connectivity rate of the clustered community is higher than 0.7, and the paths between key nodes are rich. The community structure feature matrix includes two dimensions: internal connectivity rate and external connection number. The internal connectivity rate of the clustered community C is 0.8 and the number of external connections is 25, reflecting a strong information aggregation ability. The emotional state division is based on the emotional tendency of the text. The proportion of positive emotional texts exceeding 60% is divided into positive state, the proportion of negative emotional texts exceeding 60% is divided into negative state, and the rest is divided into neutral state. The emotional state transition shows regular changes within the 24-hour time window, and the probabilistic state machine records the state transition process from 00:00 to 24:00.In the intermediate community D, the probability of a positive emotional state turning into a neutral state is 0.4, and the probability of turning into a negative state is 0.2; the probability of a neutral state turning into a positive state is 0.3, and the probability of turning into a negative state is 0.3; the probability of a negative state turning into a positive state is 0.2, and the probability of turning into a neutral state is 0.5. The degree of community penetration reflects the intensity of emotional transmission across communities. A penetration degree of 0.6 means that 60% of emotional information is transmitted across communities. The transmission efficiency value comprehensively considers the characteristics of state transfer and emotional balance. The state balance coefficient is calculated by dividing the number of positive nodes by the number of negative nodes. In the clustered community E, there are 300 positive emotional nodes and 100 negative emotional nodes. The state balance coefficient is 3.0. Combined with the state transfer probability matrix, the transmission efficiency value is 0.85. By comparison, it is found that the transmission efficiency value of the clustered community is significantly higher than that of the isolated community and the intermediate community, indicating that a tight community structure is conducive to emotional transmission and diffusion. The impact of community structure on emotional transmission is reflected in many aspects such as transmission path, transmission speed, and transmission scope.

[0035] S105. Generate a large number of negative emotion propagation paths using the Monte Carlo simulation method, and count the coverage and influence of negative emotion nodes. Determine the scope of influence and key propagation nodes of negative emotions, and obtain quantitative indicators reflecting the degree of propagation and penetration of negative emotions among different communities. The quantitative indicators include the propagation speed, scope of influence, and duration of negative emotions.

[0036] A set of negative nodes with polarity values ​​lower than a preset threshold value in the social network nodes is obtained, the propagation path is generated by the Monte Carlo method based on the negative node set, and the propagation path is time-series sorted by the node release timestamp; a hierarchical traversal is performed on the propagation path to count the number of covered nodes, the propagation influence range is calculated based on the number of covered nodes, and the node degree centrality is obtained by multiplying the in-degree value and the out-degree value of the covered node; the node propagation time series is extracted from the propagation path, and a state transfer chain of the time series is established by using a Bayesian network, and the propagation speed is obtained if the number of nodes in the propagation path is divided by the duration; a propagation matrix is ​​constructed based on the propagation speed and the number of covered nodes, and the ratio of the number of nodes in the source community to the number of nodes in the target community is calculated from the propagation matrix to obtain the propagation penetration value, and a quantitative indicator is generated through the propagation speed, the scope of influence, and the duration.

[0037] Specifically, a set of negative nodes with emotion polarity values ​​lower than a preset threshold is obtained from social network nodes. For negative nodes, the Monte Carlo method is used to set the walk depth and walk times to generate a propagation path. The paths are sorted in time according to the node release timestamp and forwarding timestamp, and the negative emotion propagation paths are screened by the path length and node polarity value. For each negative emotion propagation path, the number of node coverage is counted by hierarchical traversal, and the propagation influence range is calculated by accumulating the proportion of covered nodes to the total nodes of the community. The node degree centrality is generated by the product of the in-degree value and the out-degree value of the covered node, and the key propagation nodes are obtained from the degree centrality value sorting. The node propagation time series is extracted from the negative emotion propagation path, and a time series state transfer chain is established for the propagation sequence using a Bayesian network. The duration is calculated based on the first and last appearance times of the negative emotion node on the propagation path, and the propagation speed is obtained by dividing the number of nodes on the propagation path by the duration. The community propagation matrix is ​​constructed based on the propagation speed value and node coverage rate. The propagation penetration value is obtained by calculating the ratio of the number of source community nodes to the number of target community nodes from the propagation matrix. The quantitative indicators are generated through the three-dimensional values ​​of propagation speed, influence range, and duration. The characteristics of negative emotion propagation between different communities are compared based on the quantitative indicators. Negative emotion nodes are identified in social networks through the emotional polarity value of text content. The content published by the node is obtained through sentiment analysis. The polarity value between -1 and 1 is obtained. When the polarity value is lower than -0.6, it is determined to be a negative node. The Monte Carlo random walk starts from the negative node, sets the walk depth to 4 layers, and the number of walks to 1000 times, generating a large number of possible propagation paths. Node A publishes negative content at 10:00, node B forwards it at 10:30, and node C forwards it at 11:00, forming a time-ordered propagation path ABC. Nodes with a path length exceeding 4 or containing a polarity value greater than -0.6 will be filtered out. The negative emotion propagation path calculates the influence range through hierarchical traversal. In community X, the total number of nodes is 1000, the number of nodes covered by negative emotions is 300, and the propagation influence range is 0.3. The node degree centrality reflects the propagation ability. The in-degree of node D is 15, the out-degree is 10, and the degree centrality value is 150; the in-degree of node E is 8, the out-degree is 4, and the degree centrality value is 32. The nodes with the top 10% of degree centrality values ​​are identified as key propagation nodes, which play a key hub role in the process of negative emotion propagation. The Bayesian network considers the temporal dependency between nodes when modeling the propagation sequence, and records the state transition probability of each propagation step. In the propagation path FGHI, the negative emotion first appears at node F at 14:00 and finally appears at node I at 16:00, lasting for 2 hours. The propagation path contains 4 nodes, and the propagation speed is 2 nodes / hour, reflecting the diffusion rate of negative emotions in the network. The community propagation matrix describes the propagation characteristics of negative emotions among different communities. The source community Y contains 500 nodes. After propagating to the target community Z, it affects 300 nodes, and the propagation penetration value is 0.6.The quantitative index vector [2, 0.3, 2] is constructed by using the propagation speed of 2 nodes / hour, the impact range of 0.3, and the duration of 2 hours. Comparison shows that the quantitative indicators of communities P and Q of similar size are [1.5, 0.25, 3] and [2.5, 0.35, 1.5] respectively, indicating that negative emotions in community Q spread faster and have a wider impact, but the duration is shorter, reflecting the impact of different community structures on the characteristics of negative emotion propagation.

[0038] S106. Based on the quantitative analysis results of the community's emotional susceptibility and resistance, determine the target community and key nodes for emotional intervention, formulate differentiated emotional intervention strategies, and optimize the community operation strategy, with respect to the scope of influence and key transmission nodes of negative emotions. The community's emotional susceptibility is the tendency of the community to accept and spread certain emotions, which is calculated by analyzing historical emotional transmission data. The community's emotional resistance is the ability of the community to resist the spread of certain emotions, which is quantified by measuring the speed of emotional transmission and the degree of limitation of the impact.

[0039] The propagation speed value is obtained by dividing the number of negative emotion propagation nodes in a fixed time window by the total number of community nodes, and the propagation speed value is multiplied by the coverage node ratio to obtain the community emotion susceptibility value; the node risk score is calculated by multiplying the node in-degree value by the negative emotion forwarding rate, wherein the negative emotion forwarding rate is obtained by dividing the number of historical negative emotion forwarding times of the node by the total number of forwarding times of the node; a feature vector is constructed by the community emotion susceptibility value and the resistance value, and spectral clustering is performed on the node risk score data to obtain the community danger level; the connectivity coefficient is calculated for the propagation topology structure corresponding to the community danger level, and the connectivity coefficient is divided by the negative emotion propagation speed to obtain the propagation resistance coefficient, and the propagation resistance coefficient is multiplied by the node average risk score to generate the community operation priority sequence.

[0040] Specifically, the community propagation path within a fixed time window is extracted from the historical negative emotion propagation data. The propagation speed is calculated by dividing the number of negative emotion propagation nodes per unit time by the total number of community nodes. The community emotion susceptibility value is obtained by multiplying the propagation speed by the percentage of covered nodes. The probability of propagation obstruction is obtained by dividing the number of negative emotion propagation interruption points by the total number of propagation paths, thereby calculating the community emotion resistance value. For the nodes in the community with the highest community emotion susceptibility value, logistic regression is used to calculate the risk score of the node. The negative emotion forwarding rate is obtained by dividing the number of historical forwarding negative emotions of the node by the total number of forwarding times of the node. The node risk score is calculated by multiplying the node in-degree value by the negative emotion forwarding rate. The target node for emotional intervention is selected based on the risk score. A two-dimensional feature vector is constructed from the community emotion susceptibility value and resistance value. Spectral clustering is used to divide the risk score data of all nodes into high-risk communities, medium-risk communities, and low-risk communities. The community intervention level is determined by the community's danger level and the average risk score of the node. Two types of operational measures, positive guidance and negative blocking, are generated according to the intervention level. The communication topology structure is extracted for communities with different intervention levels, and the community connectivity coefficient is calculated from the node connection matrix. The communication resistance coefficient is obtained by dividing the connectivity coefficient by the negative emotion transmission speed. The community operation priority sequence is generated by multiplying the resistance coefficient by the average risk score of the node, and positive guidance content is given priority to high-priority communities. In the analysis of community emotion transmission, the fixed time window is usually set to 24 hours to record the transmission trajectory of negative emotion nodes. The total number of nodes in community A is 1000. Within 1 hour, the negative emotion transmission covers 150 nodes, the transmission speed is 0.15, the coverage node ratio is 0.15, and the calculated emotional susceptibility value is 0.0225. There are 25 transmission interruption points in 100 transmission paths, and the probability of transmission obstruction is 0.25, indicating that community A has a certain emotional resistance. The node risk score reflects the degree of danger of the node in the transmission of negative emotions. The total number of forwarding times of node B is 200 times, of which 50 times are negative emotion forwarding, and the negative emotion forwarding rate is 0.25. The in-degree value of node B is 40, indicating that 40 other nodes have established connections with it, and the node risk score is 10. Logistic regression takes nodes with higher risks in historical communication data as positive samples during training to construct a node risk predictor. In community C, nodes with a node risk score of more than 8.0 are included in the set of intervention target nodes. Spectral clustering divides the danger level based on the community feature vector. The feature vector of community D is [0.0225, 0.25], indicating high susceptibility and medium resistance. According to the clustering results, community D is classified as a medium-risk community with an average node risk score of 6.5. The medium-risk community adopts a mild intervention level. Operational measures include pushing positive content to key nodes and restricting the spread of negative content on high-risk nodes. The community communication topology reflects the connection relationship between nodes. The average path length between any two nodes in community E is 3, and the connectivity coefficient is 0.33.The propagation speed of negative emotions in community E is 0.2, and the calculated propagation resistance coefficient is 1.65. Combined with the average node risk score of 5.0, the operational priority of community E is 8.25. In the case of limited operational resources, priority is given to placing positive content in community E to guide group emotions to develop in a positive direction. The implementation of operational measures is based on the principle of differentiation of community characteristics, and intervention measures of corresponding intensity are taken for communities of different risk levels. High-risk communities adopt strict content review and communication control, medium-risk communities focus on the balance between positive guidance and negative constraints, and low-risk communities mainly carry out routine positive operation and maintenance.

[0041] S107. Push positive emotion information to target communities and key nodes through social networking platforms and monitor the dynamics of negative emotion transmission. According to the real-time changes in the distribution of emotions in the emotion transmission network, dynamically adjust the emotion intervention strategy, control the spread of negative emotions, strengthen community emotion management measures, and improve the analysis of community emotion interaction patterns.

[0042] Obtain community keywords and text keywords, calculate cosine similarity based on the community keywords and the text keywords to obtain a content relevance score; receive text information of the target content, use a naive Bayes classifier to identify the emotional polarity of the text information, and if the negative emotional intensity value of the text information exceeds a preset threshold, determine that the target content is negative communication content; calculate a ratio based on the number of texts of the negative communication content and the number of texts of the positive communication content to obtain an intervention effect value, and use the intervention effect value to update parameters for the content theme; calculate the propagation speed through the propagation time of the negative communication content between nodes, obtain a diffusion index based on the propagation speed and the number of node coverage, and use the diffusion index and the positive content absorption rate to generate a management and control effect value.

[0043] Specifically, extract the feature vector of positive emotional content from the content library, calculate the cosine similarity of community keywords through text keywords, obtain the content relevance score according to the word frequency value, divide the day and night active periods according to the distribution of historical posting time of the community, generate the forwarding tendency index from the historical forwarding data of the target node, and determine the priority of positive content delivery according to the relevance score. Extract the negative emotional communication content of the social platform regularly through text keywords, use naive Bayes to identify the emotional polarity of the text, filter the target content according to the threshold of the negative emotional intensity value, count the node forwarding time interval and forwarding level from the communication link, and trigger the warning signal according to the proportion of nodes covered by negative emotions exceeding the preset threshold. Calculate the responsiveness of the positive content delivery results and the negative emotional communication data, use the decision tree to perform time series modeling on the trend of community emotional changes, calculate the intervention effect value through the ratio of the number of positive and negative emotional texts, dynamically update the delivery content theme and delivery frequency according to the intervention effect value, and adjust the positive content push rules based on the warning signal data. An emotional interaction evaluation matrix is ​​constructed for the community emotional control indicators. The diffusion index is calculated based on the speed and coverage of negative emotional transmission. The control effect value is generated based on the absorption rate of positive content and the attenuation rate of negative emotions. The emotional interaction modes of different communities are quantitatively compared, and the content delivery mechanism is continuously optimized from the control effect value. The relevance of positive content keywords and community keywords determines the delivery priority. For example, the high-frequency keywords of community A include "graduate entrance examination", "preparation for examination", and "learning". The positive content "persistence in learning will eventually succeed" is calculated with the community keywords to obtain a similarity score of 0.85. The community active period is determined by statistically analyzing the distribution of posting time. The data shows that the community has the highest number of posts between 20:00 and 23:00, accounting for 35% of the total number of posts for the whole day. The number of forwardings of learning topics by target node B in the past 30 days accounts for 60% of the total number of forwardings, and the forwarding tendency index is 0.6. The identification of negative emotional transmission content is based on the judgment of text emotional polarity. Naive Bayes learns the characteristics of emotional dictionary through training corpus and labels the text into three categories: negative, neutral, and positive. Texts with a sentiment polarity value lower than -0.7 are marked as strongly negative content and need to be monitored. In the communication link, if node C is forwarded by 100 nodes within 15 minutes after publishing negative content, and the covered nodes account for 20% of the total community, an early warning signal is triggered. The node forwarding level reflects the depth of communication. Deep communication above 4 layers indicates that negative emotions have a tendency to spread. The intervention effect evaluation is achieved by comparing the trend of positive and negative emotions. Within 24 hours after the positive content was released in community D, the ratio of the number of positive emotion texts to the number of negative emotion texts increased from 0.8 to 1.2, and the intervention effect value increased by 0.4. The decision tree dynamically adjusts the delivery strategy according to the intervention effect value. When the effect value is lower than 0.5, the delivery frequency is increased, and when the effect value is higher than 1.5, the existing frequency is maintained. The frequency of early warning signals also affects the delivery adjustment. Three consecutive warnings will trigger the update of the delivery theme. The emotional interaction evaluation matrix comprehensively considers the characteristics of communication and control.The speed of negative emotion propagation in community E is 50 new nodes per hour, and the impact range reaches 30% of the total community, with a diffusion index of 15. Positive content is forwarded 300 times within 24 hours, with an absorption rate of 0.3; the speed of negative emotion propagation is reduced by 40%, with a decay rate of 0.4, and a comprehensive calculation results in a control effect value of 0.35. By comparing the control effect values ​​of different communities, the best emotional interaction mode is selected. Community F adopts a high-frequency, small-dose positive content delivery method, and the control effect value reaches 0.6, proving that this interaction mode is more suitable for the characteristics of this type of community.

[0044] It should be noted that the above examples are only some specific embodiments of the present invention. Obviously, the present invention is not limited to the above embodiments, and there are many variations. All variations that can be directly derived or associated with the content disclosed by a person skilled in the art should be considered as the protection scope of the present invention.

Claims

1. A method for monitoring and intervening in the mental health of college faculty and students, characterized in that: The method comprises: Obtain user relationship data and emotional expression data of university faculty and students on social networking platforms, and count the emotional density attributes of members of different communities, including the richness of emotional vocabulary and the proportion of different types of emotions; Use sentiment analysis algorithms to identify negative emotions in sentiment expression data, build sentiment propagation network topology based on user relationship data, use community discovery algorithms to identify different communities, calculate sentiment density attributes within each community, and sentiment propagation direction, propagation intensity, and attenuation rate between communities; A multi-factor correlation model of emotional communication is established in combination with the communication intensity attribute, which includes community size, number of communication paths, and average path length. The complexity of emotional interaction between communities is quantitatively evaluated to obtain quantitative indicators of the degree of communication penetration between different communities. In the topological structure of the emotion propagation network, the key nodes on the propagation path are combined to evaluate the impact of community structure on emotion propagation. Different community structures are divided into different emotional states through clustering algorithms. The transition probability matrix of each emotional state is calculated based on the quantitative index of the degree of emotion propagation penetration among different communities. The emotion propagation efficiency under different community structures is calculated based on the transition probability matrix. A large number of negative emotion propagation paths are generated using the Monte Carlo simulation method, the coverage and influence of negative emotion nodes are counted, the impact range and key propagation nodes of negative emotions are determined, and quantitative indicators reflecting the degree of propagation and penetration of negative emotions among different communities are obtained; In view of the impact range and key transmission nodes of negative emotions, according to the quantitative analysis results of community emotional susceptibility and resistance, the target communities and key nodes for emotional intervention are determined, differentiated emotional intervention strategies are formulated, and community operation strategies are optimized. The community emotional susceptibility is the tendency of the community to accept and spread certain emotions, which is calculated by analyzing historical emotional transmission data. The community emotional resistance is the ability of the community to resist the spread of certain emotions, which is quantified by measuring the speed of emotional transmission and the degree of influence restriction; Push positive emotional information to target communities and key nodes through social networking platforms, monitor the dynamics of negative emotional propagation, dynamically adjust emotional intervention strategies based on real-time changes in emotional distribution in the emotional propagation network, control the spread of negative emotions, strengthen community emotional management measures, and improve community emotional interaction pattern analysis.

2. The method according to claim 1, characterized in that The user relationship data and emotional expression data of university faculty and students on the social network platform are obtained, and the emotional density attributes of members of different communities are counted, including the richness of emotional vocabulary and the proportion of different types of emotions, including: Obtaining social platform user identifiers for university teacher groups and student groups, and obtaining target user text content from the social platform interface according to the user identifiers; Using a preset emotion dictionary to perform word segmentation processing on the target user's text content, and adding six types of emotion labels, namely, joy, anger, sadness, fear, disgust and surprise, to the text content after word segmentation processing by using a maximum entropy algorithm to obtain a user emotion vocabulary; According to the user emotion vocabulary table, the number of each type of emotion vocabulary is counted, and the number of each type of emotion vocabulary is divided by the total number of emotion vocabulary to obtain the user emotion vocabulary richness value; The group emotion density distribution is calculated according to the richness value of the user emotion vocabulary, a group emotion distribution matrix is ​​generated according to the group emotion density distribution, and standardized group emotion distribution data is obtained through normalization processing.

3. The method according to claim 1, characterized in that The method uses a sentiment analysis algorithm to identify negative emotions in the sentiment expression data, constructs a sentiment propagation network topology based on user relationship data, uses a community discovery algorithm to identify different communities, and calculates the sentiment density attributes within each community and the sentiment propagation direction, propagation intensity, and attenuation rate between communities, including: Obtain negative emotion words in the text, extract emotion word feature vectors based on word embedding method, and use emotion dictionary to annotate negative emotion training samples to obtain negative emotion scores; For the negative emotion score, a social communication graph is constructed through attention and interaction data between user identifiers, and a label propagation algorithm is used to divide the nodes with negative emotion intensity values ​​in the social communication graph into communities; Constructing a propagation time sequence link according to the node publishing time and forwarding time in the social propagation graph, using the Dijkstra algorithm to calculate the shortest path distance between nodes, and obtaining propagation intensity data through the negative emotion intensity value and the propagation path length; For the transmission timing link, the difference calculation is performed on the negative emotion intensity value of the transmission path, an exponential function curve is used to fit the change of negative emotion intensity at different path distances, and the negative emotion attenuation rate is calculated through the transmission path length.

4. The method according to claim 1, characterized in that: The multi-factor association model of emotion propagation is established by combining the propagation intensity attribute, wherein the propagation intensity attribute includes the community size, the number of propagation paths, and the average path length, and quantitatively evaluates the complexity of emotional interaction between communities to obtain quantitative indicators of the degree of propagation penetration between different communities, including: Obtaining the connection data between nodes in the community interaction network, and counting the number of nodes inside the community according to the connection data to obtain the community size value; Generate a community propagation adjacency matrix for the node connection data, and calculate the node propagation ratio and the propagation path set product through the community propagation adjacency matrix to obtain the propagation intensity parameter; Constructing a multidimensional feature tensor according to the propagation intensity parameter, and using a neural network to extract the product of the community scale feature vector and the path feature vector from the multidimensional feature tensor to obtain a community interaction complexity parameter; A community penetration matrix is ​​constructed according to the community interaction complexity parameter. If the ratio of the number of source community nodes to the number of target community nodes multiplied by the propagation path length exceeds a preset threshold, a corresponding community penetration level indicator is generated.

5. The method according to claim 1, characterized in that In the topological structure of the emotion propagation network, the key nodes on the propagation path are combined to evaluate the impact of the community structure on the emotion propagation. Different community structures are divided into different emotion states through a clustering algorithm. The transition probability matrix of each emotion state is calculated based on the quantitative index of the degree of emotion propagation penetration among different communities. The emotion propagation efficiency under different community structures is calculated based on the transition probability matrix, including: Obtain the number of connections of the upper and lower four layers of neighbor nodes of the network node, calculate the node betweenness value according to the number of connections, and identify the key propagation node through the node betweenness value; According to the path connectivity between the key communication nodes, a hierarchical clustering method is used to construct a community structure feature matrix, which includes the internal connectivity rate and the number of external edges of the community; According to the emotional expression data of the community structure feature matrix, a probabilistic state machine is used to record emotional state conversion data, and the amount of emotional transmission across communities is obtained through the community penetration degree value; A state transition probability matrix is ​​generated for the cross-community emotion propagation amount and the emotion state conversion data, a state balance coefficient is obtained by dividing the number of positive emotion state nodes by the number of negative emotion state nodes, and a propagation efficiency value is calculated based on the state balance coefficient and the state transition probability matrix.

6. The method according to claim 1, characterized in that The Monte Carlo simulation method is used to generate a large number of negative emotion propagation paths, count the coverage and influence of negative emotion nodes, determine the impact range and key propagation nodes of negative emotions, and obtain quantitative indicators reflecting the degree of propagation and penetration of negative emotions among different communities, including: Acquire a set of negative nodes whose polarity values ​​are lower than a preset threshold value in the social network nodes, generate the propagation path according to the set of negative nodes by using the Monte Carlo method, and perform time sequence sorting on the propagation path according to the node release timestamps; Performing hierarchical traversal on the propagation path to count the number of covered nodes, calculating the propagation influence range according to the number of covered nodes, and obtaining the node degree centrality by multiplying the in-degree value and the out-degree value of the covered node; Extracting a node propagation time series from the propagation path, using a Bayesian network to establish a state transfer chain of the time series, and obtaining a propagation speed by dividing the number of nodes in the propagation path by the duration; A propagation matrix is ​​constructed according to the propagation speed and the number of covered nodes, and the propagation penetration value is obtained by calculating the ratio of the number of source community nodes to the number of target community nodes from the propagation matrix. A quantitative index is generated through the propagation speed, impact range and duration.

7. The method according to claim 1, characterized in that The scope of influence and key transmission nodes of negative emotions are determined according to the quantitative analysis results of community emotional susceptibility and resistance, and the target community and key nodes of emotional intervention are determined, and differentiated emotional intervention strategies are formulated to optimize community operation strategies. The community emotional susceptibility is the tendency of the community to accept and spread certain emotions, which is calculated by analyzing historical emotional transmission data. The community emotional resistance is the ability of the community to resist the spread of certain emotions, which is quantified by measuring the speed of emotional transmission and the degree of influence restriction, including: The propagation speed value is obtained by dividing the number of negative emotion propagation nodes in a fixed time window by the total number of community nodes, and the propagation speed value is multiplied by the coverage node ratio to obtain the community emotion susceptibility value; The node risk score is calculated by multiplying the node in-degree value by the negative emotion forwarding rate, where the negative emotion forwarding rate is obtained by dividing the number of negative emotion forwarding times of the node by the total number of forwarding times of the node; A feature vector is constructed by using the community emotional susceptibility value and resistance value, and spectral clustering is performed on the node risk score data to obtain the community danger level; The connectivity coefficient is calculated for the communication topology structure corresponding to the community danger level, the connectivity coefficient is divided by the negative emotion propagation speed to obtain the communication resistance coefficient, and the communication resistance coefficient is multiplied by the average risk score of the node to generate the community operation priority sequence.

8. The method according to claim 1, characterized in that The method of pushing positive emotion information to target communities and key nodes through social network platforms, monitoring the dynamics of negative emotion propagation, dynamically adjusting emotion intervention strategies according to real-time changes in emotion distribution in the emotion propagation network, controlling the spread of negative emotions, strengthening community emotion management measures, and improving community emotion interaction pattern analysis includes: Obtaining community keywords and text subject words, and calculating cosine similarity based on the community keywords and the text subject words to obtain a content relevance score; Receive text information of target content, use a naive Bayes classifier to identify the emotional polarity of the text information, and if the negative emotional intensity value of the text information exceeds a preset threshold, determine that the target content is negative communication content; Calculating a ratio according to the number of texts of the negatively propagated content and the number of texts of the positively propagated content to obtain an intervention effect value, and using the intervention effect value to update parameters of the content theme; The propagation speed is calculated by the propagation time of the negative content between nodes, the diffusion index is obtained according to the propagation speed and the number of nodes covered, and the control effect value is generated by using the diffusion index and the positive content absorption rate.

9. A system for monitoring and intervening mental health of college faculty and students, characterized in that: The system comprises: Data collection and emotion calculation module, used to obtain and analyze user emotion data; Community emotion propagation network construction module, used to construct and analyze emotion propagation networks; Emotional communication model construction and analysis module, used to build emotional communication models and analyze communication rules; The negative emotion transmission simulation and intervention module is used to simulate and intervene in the transmission of negative emotions.

Images