A method and system for intelligent assessment and early warning of mental health risks of adolescents

Abstract

This invention discloses an intelligent assessment and early warning method and system for adolescent mental health risks, specifically relating to the field of mental health early warning technology. It constructs a mental health risk propagation network by collecting individual adolescent psychological and behavioral data and social interaction data, analyzes the propagation direction and probability of mental health risks, and identifies risk-aggregated areas and high-influence nodes in social networks. By integrating psychological risk propagation path characteristic data and network-level psychological risk aggregation characteristic data, it ultimately determines the source nodes of mental health risks and clustered mental health risk events, achieving intelligent assessment and early warning of adolescent mental health risks. This intelligent assessment and early warning system can accurately identify the source and propagation path of mental health risks, improving the timeliness and accuracy of early warnings.

Description

Technical Field

[0001] This invention relates to the field of mental health early warning technology, and more specifically, to a method and system for intelligent assessment and early warning of mental health risks in adolescents. Background Technology

[0002] The widespread use of digital social platforms means that the mental state of teenagers is not only affected by their own factors, but also closely related to their social relationships and online interactions.

[0003] Currently, risk assessment and early warning for adolescent mental health mainly rely on independent analysis and feature identification of individual psychological and behavioral data, neglecting the interactive transmission effect of psychological risks in social relationship networks. This makes it difficult to identify and locate the initial source of psychological risk transmission and clustered risk events, resulting in a lag in risk warning and restricting the effectiveness and timeliness of early warning and management of adolescent mental health risks.

[0004] To address the aforementioned problems, a technical solution is provided. Summary of the Invention

[0005] In order to overcome the above-mentioned deficiencies of the prior art, embodiments of the present invention provide a method and system for intelligent assessment and early warning of adolescent mental health risks to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: A smart assessment and early warning method for adolescent mental health risks includes the following steps: S1: Collect psychological and behavioral data and corresponding social relationship interaction data of individual adolescents, and perform time alignment and structuring processing to generate individual psychological and behavioral characteristic data and social relationship correlation data; S2: Based on individual psychological and behavioral characteristics data and social relationship correlation data, construct psychological risk transmission network data including adolescent nodes and social connection relationships; S3: Based on psychological risk transmission network data, analyze the transmission direction and probability of psychological risks among adolescent nodes, and generate psychological risk transmission path characteristic data; S4: Based on psychological risk propagation network data, identify risk clustering areas and high-influence nodes in social networks, and generate network-level psychological risk clustering characteristic data; S5: Integrate psychological risk transmission path characteristic data with network-level psychological risk aggregation characteristic data to form comprehensive social network psychological risk assessment data; S6: Based on comprehensive assessment data of psychological risks from social networks, identify the source nodes of psychological risks and clustered psychological risk events, and output corresponding early warning results of mental health risks.

[0007] In a preferred embodiment, S1 specifically refers to: Collect psychological and behavioral data of adolescents in digital interaction scenarios over a continuous period of time, as well as social relationship interaction data corresponding to the psychological and behavioral data; A unified time reference is used to calibrate psychological and behavioral data and social relationship interaction data; Transform psychological and behavioral data into psychological and behavioral feature vectors to generate individual psychological and behavioral feature data; Social relationship interaction data is structured and encoded to generate social relationship association data.

[0008] In a preferred embodiment, S2 specifically refers to: Based on individual psychological and behavioral characteristics data and social relationship data, a set of nodes and a set of edges are constructed. Each individual adolescent is assigned a node, and the node attributes are set as the psychological and behavioral feature vector corresponding to the individual adolescent. Map the social interaction relationship between any two adolescents as an edge, and set the edge attributes as interaction frequency weight, interaction time sequence weight and interaction context similarity weight. By using node sets and edge sets to generate graph data structures, we can form psychological risk transmission network data that includes adolescent nodes and social connections.

[0009] In a preferred embodiment, S3 specifically refers to: Traverse all the connections between adolescent nodes in the psychological risk transmission network data, and calculate the similarity of the psychological and behavioral feature vectors between adolescent nodes corresponding to each connection edge; The intensity of psychological risk propagation for each connecting edge is determined based on the similarity of psychological and behavioral feature vectors. Based on the edge attributes in the psychological risk transmission network data, calculate the probability of psychological risk transmission for each edge. Based on the intensity and probability of psychological risk transmission, the direction of psychological risk transmission in psychological risk transmission network data is determined; Record the direction of psychological risk transmission and the corresponding probability of psychological risk transmission to generate characteristic data of psychological risk transmission paths.

[0010] In a preferred embodiment, S4 specifically refers to: Calculate the clustering coefficient of all adolescent nodes in the psychological risk transmission network data, identify nodes whose clustering coefficient exceeds a preset threshold, and determine the spatial range of the psychological risk clustering area; Calculate the node degree, node centrality, and node betweenness of each adolescent node in the psychological risk transmission network data to determine the candidate set of high-influence nodes; Calculate the cumulative value of the psychological risk propagation intensity of each adolescent node in the candidate set of high-influence nodes, and determine the adolescent nodes whose cumulative psychological risk propagation intensity exceeds the preset threshold as the final high-influence nodes. Record the spatial extent of psychological risk clustering areas and the final high-influence nodes to generate network-level psychological risk clustering characteristic data.

[0011] In a preferred embodiment, S5 specifically refers to: The direction of psychological risk propagation and the corresponding probability of psychological risk propagation are transformed into a psychological risk propagation path vector using a feature mapping method; The spatial extent of the psychological risk cluster area is encoded as a cluster area vector, and the final high-influence node is encoded as a high-influence node vector. Feature fusion is performed on the vectors of psychological risk propagation paths, aggregation areas, and high-influence nodes to obtain a comprehensive assessment vector of social network psychological risk, which serves as the comprehensive assessment data for social network psychological risk.

[0012] In a preferred embodiment, S6 specifically refers to: By calculating the risk intensity feature value corresponding to each adolescent node in the comprehensive assessment vector of psychological risk in social networks, the source nodes of psychological risk are identified. Based on the node identifiers of the psychological risk source nodes, the psychological risk propagation path vector corresponding to each psychological risk source node is traced back to determine the psychological risk diffusion path starting from the psychological risk source node. Based on the psychological risk diffusion path, the clustering structure of psychological risk source nodes that meet the preset clustering conditions is identified and determined as a clustered psychological risk event; Based on the source nodes of psychological risks and clustered psychological risk events, corresponding early warning results for mental health risks are generated.

[0013] On the other hand, the present invention provides an intelligent assessment and early warning system for adolescent mental health risks, comprising: Data acquisition module: Collects psychological and behavioral data and corresponding social relationship interaction data of individual adolescents, performs time alignment and structuring processing, and generates individual psychological and behavioral characteristic data and social relationship correlation data; Network construction module: Based on individual psychological and behavioral characteristics data and social relationship correlation data, construct psychological risk transmission network data including adolescent nodes and social connection relationships; The propagation analysis module analyzes the direction and probability of psychological risk propagation among adolescent nodes based on psychological risk propagation network data, and generates characteristic data of psychological risk propagation paths. Clustering Identification Module: Based on psychological risk propagation network data, it identifies risk clustering areas and high-influence nodes in social networks, and generates network-level psychological risk clustering characteristic data; Risk assessment module: Integrates data on the characteristics of psychological risk transmission paths with data on the aggregation characteristics of network-level psychological risks to form comprehensive assessment data on social network psychological risks; Early warning output module: Based on comprehensive assessment data of psychological risks from social networks, identify the source nodes of psychological risks and clustered psychological risk events, and output corresponding early warning results of mental health risks.

[0014] The technical effects and advantages of the intelligent assessment and early warning method and system for adolescent mental health risks of this invention are as follows: By collecting psychological and behavioral data of adolescents and corresponding social relationship interaction data, this study constructs data linking individual psychological and behavioral characteristics with social relationships, improving the completeness and authenticity of information related to mental health risks. It also builds a psychological risk transmission network data that includes adolescent nodes and social connections, introducing a network structure perspective to characterize the potential transmission paths of psychological risks within a group. By analyzing the transmission direction and probability of psychological risks within the network, it generates psychological risk transmission path characteristic data, reflecting the transmission trend of psychological risks among different adolescents from a dynamic relationship perspective. Furthermore, by identifying risk clustering areas and high-influence nodes, it forms network-level psychological risk clustering characteristic data, achieving a centralized representation of high-influence nodes and high-risk areas. Finally, by integrating psychological risk transmission path characteristic data and network-level psychological risk clustering characteristic data, it constructs comprehensive social network psychological risk assessment data, unifying the expression of individual risk information and group network structure information. Ultimately, based on the comprehensive social network psychological risk assessment data, it identifies psychological risk source nodes and clustered psychological risk events, achieving intelligent mental health risk early warning for group interaction scenarios. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of an intelligent assessment and early warning method for adolescent mental health risks according to the present invention; Figure 2 This is a schematic diagram of the structure of an intelligent assessment and early warning system for adolescent mental health risks according to the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1: Figure 1 This invention provides an intelligent assessment and early warning method for adolescent mental health risks, which includes the following steps: S1: Collect psychological and behavioral data and corresponding social relationship interaction data of individual adolescents, and perform time alignment and structuring processing to generate individual psychological and behavioral characteristic data and social relationship correlation data; S2: Based on individual psychological and behavioral characteristics data and social relationship correlation data, construct psychological risk transmission network data including adolescent nodes and social connection relationships; S3: Based on psychological risk transmission network data, analyze the transmission direction and probability of psychological risks among adolescent nodes, and generate psychological risk transmission path characteristic data; S4: Based on psychological risk propagation network data, identify risk clustering areas and high-influence nodes in social networks, and generate network-level psychological risk clustering characteristic data; S5: Integrate psychological risk transmission path characteristic data with network-level psychological risk aggregation characteristic data to form comprehensive social network psychological risk assessment data; S6: Based on comprehensive assessment data of psychological risks from social networks, identify the source nodes of psychological risks and clustered psychological risk events, and output corresponding early warning results of mental health risks.

[0018] S1: Collect psychological and behavioral data and corresponding social relationship interaction data of individual adolescents, and perform time alignment and structured processing to generate individual psychological and behavioral characteristic data and social relationship correlation data, including: Collect psychological and behavioral data of adolescents in digital interaction scenarios over a continuous period of time, as well as social relationship interaction data corresponding to the psychological and behavioral data; Psychological and behavioral data includes, but is not limited to, dialogue text data, voice behavior data, and behavioral interaction record data generated by adolescents in digital interaction scenarios. Dialogue text data refers to the text communication content generated by adolescents in instant messaging software or online counseling platforms; voice behavior data refers to the audio data generated by adolescents during voice chats and online psychological counseling; and behavioral interaction record data refers to the number of clicks, duration of screen dwell, and sequence of interactive option selections by adolescents in digital interaction scenarios. Social relationship interaction data includes interaction frequency data, interaction time sequence data, and interaction object association data between adolescents; interaction frequency data refers to the number of interactions between any two adolescents within a preset period; interaction time sequence data refers to the occurrence time and duration of interactions on a continuous timeline; and interaction object association data refers to the unique identifiers of all interaction objects corresponding to each adolescent.

[0019] The AI-powered mental health assessment agent automatically collects the above data at fixed intervals; for example, it can be set to automatically complete data collection every 24 hours.

[0020] A unified time reference is used to calibrate psychological and behavioral data and social relationship interaction data; The AI-powered mental health assessment agent automatically timestamps each dialogue text, each piece of voice behavior data, and each behavioral interaction record data based on the timestamps automatically attached when psychological and behavioral data and social relationship interaction data are generated. It also timestamps the interaction frequency data, interaction timing data, and interaction object association data in the social relationship interaction data. Using the smallest time granularity (e.g., 1 second) as the standard, the initial timestamps of psychological and behavioral data and social relationship interaction data are calibrated with a unified time benchmark.

[0021] Transform psychological and behavioral data into psychological and behavioral feature vectors to generate individual psychological and behavioral feature data; Dialogue text data needs to be transformed into fixed-dimensional semantic feature vectors using a text embedding model. The parameters of the text embedding model are determined in advance through parameter optimization and cross-validation on the training dataset. For example, the embedding dimension parameter of the text embedding model can be set to 256 dimensions. The speech behavior data is then denoised, and the denoised speech behavior data is converted into fixed-dimensional acoustic feature vectors using an acoustic feature extraction algorithm (such as the Mel frequency-cepply coefficient method) to obtain speech features. Behavioral interaction record data needs to be processed using statistical feature extraction methods. Interaction behavior data such as click counts, interface dwell time, and interactive option selection sequences are subjected to frequency statistics, mean calculation, and sequence pattern recognition to generate corresponding behavioral statistical feature vectors. The obtained dialogue text feature vectors, speech behavior feature vectors, and behavioral interaction record feature vectors are then fused into a unified psychological behavior feature vector through feature concatenation. During feature concatenation, the dimension and weight of the vectors can be determined through principal component analysis, thereby generating individual psychological behavior feature data.

[0022] Structured encoding of social relationship interaction data generates social relationship association data; The structured coding operation employs a relational graph coding method, specifically as follows: A node set is constructed using the unique identifiers of individual teenagers in the interaction object association data as node identifiers, ensuring that each teenager has a unique corresponding node in the node set; based on interaction frequency data and interaction time-series data, the interaction frequency weight and interaction time-series weight between any two nodes are calculated, for example, using a frequency normalization algorithm and a time-series decay weight algorithm. The frequency normalization algorithm normalizes the interaction frequency data to between 0 and 1, and the time-series decay weight algorithm determines the weight decay coefficient based on the time interval between the interaction occurrence and the current analysis time; the similarity of the interaction context between any two nodes is calculated, specifically by performing topic modeling (e.g., using LDA topic modeling) on ​​the dialogue text data corresponding to any two teenagers, and calculating the topic distribution similarity between any two nodes (e.g., using cosine similarity calculation method), thereby determining the interaction context similarity weight; the calculated interaction frequency weight, interaction time-series weight, and interaction context similarity weight are used as attributes of the edges between nodes to complete the structured coding and generate social relationship association data.

[0023] S2: Based on individual psychological and behavioral characteristics data and social relationship correlation data, construct psychological risk transmission network data including adolescent nodes and social connection relationships, including: Based on individual psychological and behavioral characteristics data and social relationship data, a set of nodes and a set of edges are constructed. For each individual adolescent, a single node is created in a node set, where each node in the node set uniquely corresponds to an individual adolescent; the attribute of each node is defined as the psychological and behavioral feature vector of the adolescent individual corresponding to that node.

[0024] Each individual adolescent is assigned a node, and the node attributes are set as the psychological and behavioral feature vector corresponding to the individual adolescent. Map the social interaction relationship between any two adolescents as an edge, and set the edge attributes as interaction frequency weight, interaction time sequence weight and interaction context similarity weight. By using node sets and edge sets to generate graph data structures, we can form psychological risk transmission network data that includes adolescent nodes and social connections.

[0025] Specifically, for each individual adolescent, a single node is created in a node set, where each node in the node set uniquely corresponds to an individual adolescent; the attribute of each node is defined as the psychological and behavioral feature vector of the adolescent individual corresponding to that node.

[0026] The social interaction relationship between any two adolescents is mapped to a single edge in the edge set. Specifically, based on the unique identifier of each adolescent, all possible node pairs are traversed to determine whether a valid social interaction relationship exists between the node pairs. The criteria for determining a valid social interaction relationship are: if the interaction frequency data value between any two adolescents is greater than zero in the social relationship association data, it indicates that there is a valid social interaction relationship between the node pairs, and thus an edge needs to be constructed for the corresponding node pair in the edge set; if the interaction frequency data value between any two adolescents is zero, then no corresponding edge needs to be constructed in the edge set.

[0027] For each pair of nodes that are determined to have a valid social interaction relationship, define the attributes of each edge, including interaction frequency weight, interaction time sequence weight, and interaction context similarity weight. The calculation method for interaction frequency weight is as follows: A frequency normalization algorithm is used to normalize the interaction frequency data between any two adolescent individuals to a preset range, such as between 0 and 1. The calculation method for interaction temporal weight is as follows: A temporal decay weight algorithm is used to determine the weight decay coefficient based on the time interval between the interaction occurrence time and the current analysis time for each edge corresponding to any two adolescent individuals. The shorter the interval between the interaction occurrence time and the current analysis time, the larger the weight decay coefficient. The calculation method for interaction context similarity weight is as follows: Topic modeling is performed on the dialogue text data corresponding to any two adolescent individuals. The topic modeling adopts the Latent Dirichlet Allocation Topic Model. The topic model parameters are determined empirically to determine the number of topics. For example, the number of topics is set to 10, and then the optimal parameter value for the number of topics is gradually determined through cross-validation, thereby obtaining the topic distribution probability vector of each adolescent individual's dialogue text data. Finally, the cosine similarity calculation method is used to obtain the interaction context similarity weight for the topic distribution probability vector between any two nodes. The closer the cosine similarity is to 1, the higher the similarity of the interaction context between the nodes.

[0028] The unified integration process of the graph data structure is as follows: all nodes in the node set are sequentially inserted into the graph data structure, with each node identified by its unique adolescent individual identifier and its psychological and behavioral feature vector as its node attribute; all edges in the edge set are sequentially inserted into the graph data structure, with the identifiers of any two nodes serving as the start and end points of the edge, and the interaction frequency weight, interaction timing weight, and interaction context similarity weight are sequentially assigned to each corresponding edge to form the connection relationship between nodes and edges; through the above unified integration process, the graph data structure reflects the network topology characteristics of all adolescent individual nodes and their social interaction relationships.

[0029] S3: Based on psychological risk transmission network data, analyze the transmission direction and probability of psychological risks among adolescent nodes, and generate characteristic data of psychological risk transmission paths, including: Traverse all the connections between adolescent nodes in the psychological risk transmission network data, and calculate the similarity of the psychological and behavioral feature vectors between adolescent nodes corresponding to each connection edge; Based on the psychological risk transmission network data, this study traverses all adolescent nodes and their connections within the network. Specifically, using a graph traversal algorithm, all adolescent nodes in the network are visited sequentially according to their node identifiers. For each visited node, each edge is traversed, meaning all connecting edges originating from that node are analyzed. Each edge corresponds to two adolescent nodes: the starting node and the ending node, representing two individual adolescents. For each pair of adolescent nodes corresponding to an edge, psychological and behavioral feature vectors are extracted from their node attributes.

[0030] For the psychological and behavioral feature vectors extracted from the two adolescent nodes corresponding to each connecting edge, the cosine similarity calculation method is used to calculate the degree of similarity. Specifically, the cosine similarity is equal to the inner product of the two psychological and behavioral feature vectors divided by the product of the magnitudes of the two psychological and behavioral feature vectors. The closer the cosine similarity is to 1, the higher the degree of similarity between the psychological and behavioral feature vectors of the two adolescent nodes.

[0031] The intensity of psychological risk propagation for each connecting edge is determined based on the similarity of psychological and behavioral feature vectors. The intensity of psychological risk transmission is defined as a monotonically increasing function of the similarity of psychological and behavioral feature vectors. For example, the intensity of psychological risk transmission is set as an exponential function expression of the similarity of psychological and behavioral feature vectors: the intensity of psychological risk transmission is equal to an exponential function with the natural constant as the base and the cosine similarity as the exponent, ensuring that the intensity of psychological risk transmission varies between 1 and the natural constant.

[0032] Based on the edge attributes in the psychological risk transmission network data, calculate the probability of psychological risk transmission for each edge. The interaction frequency weight, interaction timing weight, and interaction context similarity weight are fused. Specifically, the psychological risk propagation probability is equal to the sum of the interaction frequency weight multiplied by the first fusion coefficient, the interaction timing weight multiplied by the second fusion coefficient, and the interaction context similarity weight multiplied by the third fusion coefficient. The first, second, and third fusion coefficients are all non-negative real numbers, and their sum is 1. The fusion coefficients are determined by using a pre-established psychological risk propagation probability labeled dataset and employing cross-validation and grid search methods to determine the optimal combination of fusion coefficients. For example, through multiple rounds of cross-validation and grid search, the first fusion coefficient can be determined to be 0.4, the second fusion coefficient to be 0.3, and the third fusion coefficient to be 0.3. Thus, the psychological risk propagation probability corresponding to each connection edge is obtained.

[0033] Based on the intensity and probability of psychological risk transmission, the direction of psychological risk transmission in psychological risk transmission network data is determined; The rule for determining the direction of psychological risk propagation is as follows: between two adolescent nodes on each connecting edge, psychological risk always propagates from the adolescent node with the higher product of psychological risk propagation intensity and psychological risk propagation probability to the adolescent node with the lower product of psychological risk propagation intensity and psychological risk propagation probability. For example, for a connecting edge between two adolescent nodes A and B, if the product of psychological risk propagation intensity and psychological risk propagation probability between nodes A and B is 0.6, and the product of psychological risk propagation intensity and psychological risk propagation probability between nodes B and A is 0.4, then the direction of psychological risk propagation is from node A to node B.

[0034] Record the direction of psychological risk transmission and the corresponding probability of psychological risk transmission, and generate characteristic data of psychological risk transmission path; For each connection edge in the psychological risk propagation network data, a corresponding psychological risk propagation path record is generated. Each psychological risk propagation path record includes a propagation start node identifier, a propagation end node identifier, and a psychological risk propagation probability. All psychological risk propagation path records are integrated to form psychological risk propagation path feature data.

[0035] S4: Based on psychological risk propagation network data, identify risk clustering areas and high-influence nodes in social networks, and generate network-level psychological risk clustering characteristic data, including: Calculate the clustering coefficient of all adolescent nodes in the psychological risk transmission network data, identify nodes whose clustering coefficient exceeds a preset threshold, and determine the spatial range of the psychological risk clustering area; The clustering coefficient is calculated by dividing the actual number of edges connecting each adolescent node to its directly connected nodes in the psychological risk transmission network data by the maximum possible number of edges connecting directly connected nodes. It quantifies the density of the local network where a node resides; a clustering coefficient closer to 1 indicates a denser network of connections between the nodes surrounding the corresponding adolescent node. The calculation method is as follows: For any adolescent node in the psychological risk transmission network data, extract all other adolescent nodes directly connected to that node, defining them as the set of directly connected nodes; count the actual number of edges connecting any two nodes in the set; calculate the maximum possible number of edges connecting the nodes within the set, which is equal to half the number of nodes multiplied by the number of nodes minus 1; finally, divide the actual number of edges by the maximum number of edges to obtain the clustering coefficient of the adolescent node.

[0036] Adolescent nodes are identified based on a preset clustering coefficient threshold. Nodes with clustering coefficients exceeding this threshold are designated as member nodes of psychological risk clustering areas. The preset clustering coefficient threshold is determined as follows: based on historical psychological risk propagation network data, the distribution characteristics of clustering coefficients among adolescent nodes in areas where actual psychological risk events have occurred are analyzed to establish a correspondence between clustering coefficients and the probability of actual psychological risk events. Cross-validation is used to determine the optimal threshold for the clustering coefficient, which is the boundary point distinguishing between psychological risk clustering areas and non-clustering areas. For example, through historical data analysis and cross-validation, the preset clustering coefficient threshold can be set to 0.75. Adolescent nodes with clustering coefficients greater than or equal to 0.75 are considered member nodes of psychological risk clustering areas. The spatial extent of the psychological risk clustering area is determined based on the direct connections between member nodes; that is, the spatial extent of the psychological risk clustering area includes all adjacent member nodes or nodes interconnected by direct edges.

[0037] Calculate the node degree, node centrality, and node betweenness of each adolescent node in the psychological risk transmission network data to determine the candidate set of high-influence nodes; Node degree is defined as the number of edges directly connected to each adolescent node, i.e., the number of other adolescent nodes directly connected to that adolescent node, used to measure the local connectivity of adolescent nodes in the psychological risk propagation network data; node centrality is defined as the reciprocal of the sum of the shortest path lengths between each adolescent node and all other adolescent nodes in the psychological risk propagation network data, used to quantify the importance of adolescent nodes in the psychological risk propagation network data; node betweenness is defined as the ratio of the number of times the shortest path between all other pairs of adolescent nodes in the psychological risk propagation network data passes through a specific adolescent node to the number of shortest paths between all pairs of nodes, used to reflect the bridging role of adolescent nodes in the psychological risk propagation path.

[0038] Node degree is obtained by counting the number of edges directly connected to each adolescent node; node centrality is obtained by calculating the shortest path length between every two nodes in the psychological risk propagation network data using the Floyd shortest path algorithm, and then summing the shortest path lengths from each adolescent node to other nodes and taking the reciprocal; node betweenness is obtained by counting the number of times a certain node appears in the shortest path between all node pairs in the psychological risk propagation network data and comparing it with the total number of paths.

[0039] The candidate set of high-influence nodes is determined based on three calculated indicators: node degree, node centrality, and node betweenness. Specifically, the node degree, node centrality, and node betweenness are normalized. The minimum value of each indicator is subtracted from the maximum value, and then divided by the difference between the minimum and maximum values ​​to obtain the three normalized indicators. For each adolescent node, the sum of the three normalized indicators is calculated. Adolescent nodes whose sum of the three normalized indicators exceeds a preset high-influence threshold are identified as members of the candidate set of high-influence nodes. The preset high-influence threshold is determined by labeling historical psychological risk event data and using cross-validation and statistical analysis based on historical actual psychological risk transmission network data to obtain the optimal threshold. For example, the preset high-influence threshold can be set to 1.5, meaning nodes whose sum of the three normalized indicators exceeds 1.5 are included in the candidate set of high-influence nodes.

[0040] Calculate the cumulative value of the psychological risk propagation intensity of each adolescent node in the candidate set of high-influence nodes, and determine the adolescent nodes whose cumulative psychological risk propagation intensity exceeds the preset threshold as the final high-influence nodes. Traverse all direct edges of adolescent nodes in each candidate set and extract the psychological risk propagation intensity; sum up the psychological risk propagation intensity corresponding to all edges of adolescent nodes in each candidate set to obtain the cumulative value of psychological risk propagation intensity corresponding to adolescent nodes.

[0041] The method for determining the preset threshold of the cumulative value of psychological risk transmission intensity is as follows: based on historical psychological risk transmission network data, the optimal threshold is determined by analyzing the correlation between the actual cumulative value of psychological risk transmission intensity and historical psychological risk events, and using statistical analysis and cross-validation methods. For example, the preset threshold of the cumulative value of psychological risk transmission intensity can be set to 20, that is, adolescent nodes with a cumulative value of psychological risk transmission intensity exceeding 20 are determined as the final high-influence nodes.

[0042] Record the spatial extent of psychological risk clustering areas and the final high-influence nodes to generate network-level psychological risk clustering characteristic data; The spatial extent of the psychological risk cluster area is recorded as a set of member node identifiers, that is, the set of all node identifiers that are interconnected and whose clustering coefficient exceeds the preset clustering coefficient threshold; the final high-influence node is recorded as a list of identifiers of the final high-influence node.

[0043] S5: Integrating data on the characteristics of psychological risk transmission paths with data on the clustering characteristics of network-level psychological risks, a comprehensive assessment of social network psychological risks is formed, including: The direction of psychological risk propagation and the corresponding probability of psychological risk propagation are transformed into a psychological risk propagation path vector using a feature mapping method; Each psychological risk propagation path record is extracted from the psychological risk propagation path feature data. Each record includes a propagation start node identifier, a propagation end node identifier, and a corresponding psychological risk propagation probability. An independent psychological risk propagation probability sub-vector is constructed for each adolescent node in the psychological risk propagation network data. The dimension of each sub-vector is equal to the total number of adolescent nodes in the network data, with each dimension corresponding to a specific adolescent node. The value of a specific dimension is determined as follows: if the current adolescent node has a psychological risk propagating to the adolescent node corresponding to that dimension, then the psychological risk propagation probability of the current adolescent node to that node is taken; otherwise, it is set to 0. Using this method, a corresponding psychological risk propagation probability sub-vector is constructed for each adolescent node. All psychological risk propagation probability sub-vectors are then concatenated sequentially based on the adolescent node identifier order to obtain the final psychological risk propagation path vector.

[0044] The spatial extent of the psychological risk cluster area is encoded as a cluster area vector, and the final high-influence node is encoded as a high-influence node vector. The method for generating the psychological risk clustering region vector is as follows: Based on the set of member node identifiers recorded in the spatial range of the psychological risk clustering region of the network-level psychological risk clustering feature data center, all adolescent nodes in the psychological risk propagation network data are sequentially labeled, and a binary vector with the same dimension as the number of adolescent nodes is constructed. Each dimension in the vector corresponds to one adolescent node. For each dimension, if the identifier of the corresponding adolescent node is included in the set of member node identifiers recorded in the spatial range of the psychological risk clustering region, the value corresponding to that dimension is set to 1; otherwise, it is set to 0. This forms a clustering region vector that can identify the spatial distribution characteristics of the psychological risk clustering region.

[0045] The method for generating the final high-influence node vector is as follows: Based on the list of final high-influence node identifiers recorded in the network-level psychological risk cluster feature data, all adolescent node identifiers in the psychological risk transmission network data are also sequentially labeled, and a binary vector with a dimension equal to the number of adolescent nodes is constructed, where each dimension represents the corresponding adolescent node identifier; for each dimension, if the identifier of the corresponding adolescent node is in the list of final high-influence node identifiers, the dimension value is set to 1, otherwise it is set to 0, thereby obtaining a high-influence node vector that can express the spatial location characteristics of the final high-influence nodes.

[0046] Feature fusion is performed on the vectors of psychological risk propagation paths, aggregation areas, and high-influence nodes to obtain a comprehensive assessment vector of social network psychological risk, which serves as comprehensive assessment data for social network psychological risk. Vector dimension normalization is performed on the vectors of psychological risk transmission path, cluster area, and high-influence node. For each feature vector, the minimum and maximum values ​​of all dimensions in the feature vector are calculated. The minimum value is subtracted from the value of each dimension and then divided by the difference between the maximum and minimum values, thereby uniformly mapping the vector dimension values ​​to between 0 and 1.

[0047] After dimensionality normalization, corresponding feature fusion weights are assigned to the psychological risk propagation path vector, clustering region vector, and high-influence node vector, and then weighted summation is performed. The method for determining the feature fusion weights is as follows: based on historical social network psychological risk assessment data, the feature data of psychological risk propagation paths and network-level psychological risk clustering features of multiple labeled historical psychological risk events are analyzed, and the optimal combination of feature fusion weights is determined using multi-round cross-validation and grid search methods. For example, the feature fusion weight of the psychological risk propagation path vector is set to 0.5, the feature fusion weight of the clustering region vector is set to 0.25, and the feature fusion weight of the high-influence node vector is set to 0.25.

[0048] The values ​​of each dimension in the psychological risk transmission path vector are multiplied by the corresponding feature fusion weight of 0.5, the values ​​of each dimension in the clustering region vector are multiplied by the corresponding feature fusion weight of 0.25, and the values ​​of each dimension in the high-influence node vector are multiplied by the corresponding feature fusion weight of 0.25. Finally, each dimension is summed one by one to form the fused social network psychological risk comprehensive assessment vector.

[0049] The number of dimensions in the merged social network psychological risk comprehensive assessment vector is equal to the number of dimensions in the psychological risk propagation path vector, clustering area vector, and high-influence node vector. The merged social network psychological risk comprehensive assessment vector is the social network psychological risk comprehensive assessment data.

[0050] S6: Based on comprehensive psychological risk assessment data from social networks, identify the source nodes of psychological risks and clustered psychological risk events, and output corresponding mental health risk warning results, including: By calculating the risk intensity feature value corresponding to each adolescent node in the comprehensive assessment vector of psychological risk in social networks, the source nodes of psychological risk are identified. Each dimension of the comprehensive assessment vector for psychological risks in social networks corresponds to a specific adolescent node. The value of each dimension represents the comprehensive impact of psychological risks on that specific adolescent node, defined as a risk intensity feature value. A higher risk intensity feature value indicates a higher potential psychological risk for the corresponding adolescent node. Based on historical social network psychological risk event data, a preset risk threshold for the risk intensity feature value is set. For example, the preset risk threshold is determined by using ROC curve analysis on the risk intensity feature values ​​in the historical dataset; the preset risk threshold for the risk intensity feature value can be set to 0.8. Adolescent nodes with a risk intensity feature value greater than or equal to 0.8 are identified as source nodes of psychological risks. The number of source nodes of psychological risks can be one or more.

[0051] Based on the node identifiers of the psychological risk source nodes, the psychological risk propagation path vector corresponding to each psychological risk source node is traced back to determine the psychological risk diffusion path starting from the psychological risk source node. Each adolescent node has a corresponding psychological risk propagation probability sub-vector. Each dimension in the psychological risk propagation probability sub-vector represents the probability of the current node propagating psychological risk to the node corresponding to that dimension. When tracing the psychological risk diffusion path, the psychological risk propagation probability sub-vector of each psychological risk source node is determined. The adolescent nodes corresponding to the dimensions with values ​​greater than 0 in the psychological risk propagation probability sub-vector are identified, which are the direct propagation nodes of the psychological risk source node. Starting from each direct propagation node, the process is repeated backwards to determine the propagation nodes for the next stage, until all values ​​of the psychological risk propagation probability sub-vector are 0, indicating that there are no more propagation nodes. After completing this process, the psychological risk diffusion path starting from each psychological risk source node is obtained. Each psychological risk diffusion path begins with the identifier of the psychological risk source node and sequentially records the identifiers of each adolescent node along the way until the end of the path.

[0052] Based on the psychological risk diffusion path, the clustering structure of psychological risk source nodes that meet the preset clustering conditions is identified and determined as a clustered psychological risk event; A clustering structure analysis matrix for psychological risk source nodes is constructed. The rows and columns of the matrix correspond to the identifiers of all psychological risk source nodes. Each element in the matrix represents the topological similarity between two corresponding psychological risk source nodes in the psychological risk diffusion path. The topological similarity is calculated by counting the number of nodes in the psychological risk diffusion path corresponding to any two psychological risk source nodes, and then dividing the number of nodes in the intersection by the sum of the number of nodes in the two diffusion paths to obtain the topological similarity. The closer the topological similarity is to 1, the more similar the spatial topological structure of the two psychological risk source nodes in the psychological risk diffusion path.

[0053] Based on the clustering structure analysis matrix of psychological risk source nodes, a hierarchical clustering algorithm is used for clustering analysis. Specifically, each psychological risk source node is initialized as a separate cluster. According to the topological similarity in the clustering structure analysis matrix, the two clusters with the highest topological similarity exceeding a preset clustering threshold are merged to generate new clusters. This process is repeated until all topological similarities are less than the preset clustering threshold. The clustering threshold is set by analyzing the correspondence between historical psychological risk event datasets and actual clustered psychological risk events to determine the optimal range of values ​​for the clustering threshold. For example, the preset clustering threshold can be set to 0.6. After the clustering process is complete, clusters with a number of members greater than or equal to a preset clustering size threshold are identified as clustered psychological risk events. The clustering size threshold is determined based on the statistical distribution characteristics of the number of nodes in historical clustered psychological risk events. For example, the preset clustering size threshold can be set to 3, meaning that a cluster with 3 or more member nodes is identified as a clustered psychological risk event.

[0054] Based on the source nodes of psychological risks and clustered psychological risk events, corresponding early warning results for mental health risks are generated. The early warning results for psychological risk source nodes record the unique identifier, risk intensity characteristic value, and corresponding psychological risk diffusion path of each psychological risk source node; the early warning results for clustered psychological risk events record the set of unique identifiers of all psychological risk source nodes included in each clustered psychological risk event and the corresponding clustering structure analysis matrix of psychological risk source nodes; the above early warning results for psychological risk source nodes and early warning results for clustered psychological risk events serve as the final early warning results for mental health risks.

[0055] Intervention actions or risk management plans can be developed based on the results of mental health risk warnings. For example, different intervention measures can be set for each source of mental health risk, including personalized psychological counseling or continuous monitoring; and for each cluster of mental health risk events, a group psychological intervention plan or collective mental health activities can be developed.

[0056] Example 2: The difference between Example 2 and Example 1 is that this example introduces an intelligent assessment and early warning system for adolescent mental health risks.

[0057] Figure 2 A schematic diagram of the structure of an intelligent assessment and early warning system for adolescent mental health risks is provided. The intelligent assessment and early warning system for adolescent mental health risks includes: Data acquisition module: Collects psychological and behavioral data and corresponding social relationship interaction data of individual adolescents, performs time alignment and structuring processing, and generates individual psychological and behavioral characteristic data and social relationship correlation data; Network construction module: Based on individual psychological and behavioral characteristics data and social relationship correlation data, construct psychological risk transmission network data including adolescent nodes and social connection relationships; The propagation analysis module analyzes the direction and probability of psychological risk propagation among adolescent nodes based on psychological risk propagation network data, and generates characteristic data of psychological risk propagation paths. Clustering Identification Module: Based on psychological risk propagation network data, it identifies risk clustering areas and high-influence nodes in social networks, and generates network-level psychological risk clustering characteristic data; Risk assessment module: Integrates data on the characteristics of psychological risk transmission paths with data on the aggregation characteristics of network-level psychological risks to form comprehensive assessment data on social network psychological risks; Early warning output module: Based on comprehensive assessment data of psychological risks from social networks, identify the source nodes of psychological risks and clustered psychological risk events, and output corresponding early warning results of mental health risks.

[0058] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0059] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0060] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0061] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0062] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0063] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0064] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0065] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0066] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intelligent assessment and early warning of adolescent mental health risks, characterized in that, Includes the following steps: S1: Collect psychological and behavioral data and corresponding social relationship interaction data of individual adolescents, and perform time alignment and structuring processing to generate individual psychological and behavioral characteristic data and social relationship correlation data; S2: Based on individual psychological and behavioral characteristics data and social relationship correlation data, construct psychological risk transmission network data including adolescent nodes and social connection relationships; S3: Based on psychological risk transmission network data, analyze the transmission direction and probability of psychological risks among adolescent nodes, and generate psychological risk transmission path characteristic data; S4: Based on psychological risk propagation network data, identify risk clustering areas and high-influence nodes in social networks, and generate network-level psychological risk clustering characteristic data; S5: Integrate psychological risk transmission path characteristic data with network-level psychological risk aggregation characteristic data to form comprehensive social network psychological risk assessment data; S6: Based on comprehensive assessment data of psychological risks from social networks, identify the source nodes of psychological risks and clustered psychological risk events, and output corresponding early warning results of mental health risks.

2. The intelligent assessment and early warning method for adolescent mental health risks according to claim 1, characterized in that, S1, specifically: Collect psychological and behavioral data of adolescents in digital interaction scenarios over a continuous period of time, as well as social relationship interaction data corresponding to the psychological and behavioral data; A unified time reference is used to calibrate psychological and behavioral data and social relationship interaction data; Transform psychological and behavioral data into psychological and behavioral feature vectors to generate individual psychological and behavioral feature data; Social relationship interaction data is structured and encoded to generate social relationship association data.

3. The intelligent assessment and early warning method for adolescent mental health risks according to claim 2, characterized in that, S2, specifically: Based on individual psychological and behavioral characteristics data and social relationship data, a set of nodes and a set of edges are constructed. Each individual adolescent is assigned a node, and the node attributes are set as the psychological and behavioral feature vector corresponding to the individual adolescent. Map the social interaction relationship between any two adolescents as an edge, and set the edge attributes as interaction frequency weight, interaction time sequence weight and interaction context similarity weight. By using node sets and edge sets to generate graph data structures, we can form psychological risk transmission network data that includes adolescent nodes and social connections.

4. The intelligent assessment and early warning method for adolescent mental health risks according to claim 3, characterized in that, S3, specifically: Traverse all the connections between adolescent nodes in the psychological risk transmission network data, and calculate the similarity of the psychological and behavioral feature vectors between adolescent nodes corresponding to each connection edge; The intensity of psychological risk propagation for each connecting edge is determined based on the similarity of psychological and behavioral feature vectors. Based on the edge attributes in the psychological risk transmission network data, calculate the probability of psychological risk transmission for each edge. Based on the intensity and probability of psychological risk transmission, the direction of psychological risk transmission in psychological risk transmission network data is determined; Record the direction of psychological risk transmission and the corresponding probability of psychological risk transmission to generate characteristic data of psychological risk transmission paths.

5. The intelligent assessment and early warning method for adolescent mental health risks according to claim 4, characterized in that, S4, specifically: Calculate the clustering coefficient of all adolescent nodes in the psychological risk transmission network data, identify nodes whose clustering coefficient exceeds a preset threshold, and determine the spatial range of the psychological risk clustering area; Calculate the node degree, node centrality, and node betweenness of each adolescent node in the psychological risk transmission network data to determine the candidate set of high-influence nodes; Calculate the cumulative value of the psychological risk propagation intensity of each adolescent node in the candidate set of high-influence nodes, and determine the adolescent nodes whose cumulative psychological risk propagation intensity exceeds the preset threshold as the final high-influence nodes. Record the spatial extent of psychological risk clustering areas and the final high-influence nodes to generate network-level psychological risk clustering characteristic data.

6. The intelligent assessment and early warning method for adolescent mental health risks according to claim 5, characterized in that, S5, specifically: The direction of psychological risk propagation and the corresponding probability of psychological risk propagation are transformed into a psychological risk propagation path vector using a feature mapping method; The spatial extent of the psychological risk cluster area is encoded as a cluster area vector, and the final high-influence node is encoded as a high-influence node vector. Feature fusion is performed on the vectors of psychological risk propagation paths, aggregation areas, and high-influence nodes to obtain a comprehensive assessment vector of social network psychological risk, which serves as the comprehensive assessment data for social network psychological risk.

7. The intelligent assessment and early warning method for adolescent mental health risks according to claim 6, characterized in that, S6, specifically: By calculating the risk intensity feature value corresponding to each adolescent node in the comprehensive assessment vector of psychological risk in social networks, the source nodes of psychological risk are identified. Based on the node identifiers of the psychological risk source nodes, the psychological risk propagation path vector corresponding to each psychological risk source node is traced back to determine the psychological risk diffusion path starting from the psychological risk source node. Based on the psychological risk diffusion path, the clustering structure of psychological risk source nodes that meet the preset clustering conditions is identified and determined as a clustered psychological risk event; Based on the source nodes of psychological risks and clustered psychological risk events, corresponding early warning results for mental health risks are generated.

8. A smart assessment and early warning system for adolescent mental health risks, used to implement the smart assessment and early warning method for adolescent mental health risks as described in any one of claims 1-7, characterized in that, include: Data acquisition module: Collects psychological and behavioral data and corresponding social relationship interaction data of individual adolescents, performs time alignment and structuring processing, and generates individual psychological and behavioral characteristic data and social relationship correlation data; Network construction module: Based on individual psychological and behavioral characteristics data and social relationship correlation data, construct psychological risk transmission network data including adolescent nodes and social connection relationships; The propagation analysis module analyzes the direction and probability of psychological risk propagation among adolescent nodes based on psychological risk propagation network data, and generates characteristic data of psychological risk propagation paths. Clustering Identification Module: Based on psychological risk propagation network data, it identifies risk clustering areas and high-influence nodes in social networks, and generates network-level psychological risk clustering characteristic data; Risk assessment module: Integrates data on the characteristics of psychological risk transmission paths with data on the aggregation characteristics of network-level psychological risks to form comprehensive assessment data on social network psychological risks; Early warning output module: Based on comprehensive assessment data of psychological risks from social networks, identify the source nodes of psychological risks and clustered psychological risk events, and output corresponding early warning results of mental health risks.

Images