Family stress tolerance evaluation system based on knowledge graph

Abstract

The invention discloses a family stress tolerance evaluation system based on a knowledge graph. The family stress tolerance evaluation system comprises a data acquisition layer for collecting basic information, medical history, family membership and social resources of a family of a community cerebral apoplexy patient; the knowledge graph construction layer is used for constructing a family stress resistance knowledge graph based on the collected data; the evaluation model layer is used for constructing a family resilience evaluation model, carrying out quantitative evaluation based on data in the knowledge graph, processing questionnaire survey data of an individual by utilizing logistic regression, quantitatively evaluating family resilience of an individual level, and combining family relationship data in the knowledge graph to obtain a family resilience evaluation model; the method comprises the following steps: carrying out resilience evaluation on a family level by utilizing a graph neural network GCN, quantifying resilience contribution of a social relation level by utilizing the graph neural network GCN and combining social relation data in a knowledge graph, determining weights of an individual, a family and a social relation in an evaluation model, carrying out weighted summation according to the weights, and calculating the resilience of the individual, the family and the social relation in the evaluation model. And a comprehensive evaluation result of the family stress tolerance is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of family resilience assessment, and particularly to a family resilience assessment system based on a knowledge graph. Background Art

[0002] As a family risk event, stroke has a profound impact on patients and their families. It not only disrupts the balance of the patient's family system but also has a negative impact on family recovery and the quality of family life. Assessing the family resilience of community stroke patients is particularly important. The significance of family resilience assessment lies in that it can help us systematically understand and analyze the coping abilities and recovery mechanisms of stroke patients' families in the face of difficulties. Through assessment, we can reveal the internal mechanism of family resilience, provide targeted support and intervention for families, thereby promoting family recovery and post-discharge rehabilitation of patients. In addition, the assessment results can also provide a scientific basis for policymakers, medical workers, and social workers to formulate more effective family support and rehabilitation strategies.

[0003] Traditional family resilience assessment methods may lack systematicness, comprehensiveness, and accuracy, and it is difficult to accurately reflect the comprehensive impact of the three levels of individuals, families, and social relationships. Therefore, a family resilience assessment system based on a knowledge graph is proposed. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of existing family resilience assessment methods in the prior art, aiming at the defects of traditional methods in terms of systematicness, comprehensiveness, and accuracy, especially the problem of being difficult to accurately assess the comprehensive impact of the three levels of individuals, families, and social relationships, and to propose a family resilience assessment system based on a knowledge graph.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: A family resilience assessment system based on a knowledge graph, comprising: Data acquisition layer: collecting the basic information, medical history, family member relationships, and social resources of families of community stroke patients; Knowledge graph construction layer: based on the collected data, constructing a knowledge graph of family resilience, the knowledge graph including related concepts, attributes, and relationships of family resilience, such as the roles of family members, the intensity of family relationships, social resource support, etc., introducing semantic rules to describe the relationships between family members, the corresponding relationships between resilience strategies and challenges during the construction of the knowledge graph, and using natural language processing to extract key information from text data to enrich the knowledge graph; Evaluation Model Layer: Based on three levels of individual, family, and social relationships, construct an evaluation model for family resilience. Conduct quantitative evaluation based on the data in the knowledge graph. Use logistic regression to process the individual's questionnaire survey data. Through feature selection and weight assignment, quantitatively evaluate the family resilience at the individual level. Combine the family relationship data in the knowledge graph and use the graph neural network GCN to evaluate the family-level resilience. The graph neural network GCN captures the relationships between family members and evaluates the overall family resilience level through the feature representation of nodes and edges. Use the graph neural network GCN, combine the social relationship data in the knowledge graph, and evaluate the impact of social relationships on family resilience. Quantify the contribution of resilience at the social relationship level through the connection strength and importance between nodes. Combine the relevant theories and empirical studies of family resilience, determine the weights of the three levels of individual, family, and social relationships in the evaluation model through regression analysis, and perform weighted summation according to the weights to obtain the comprehensive evaluation result of family resilience; Result Display and Application Layer: Display the evaluation results to users. According to the evaluation results, provide personalized family intervention suggestions, such as family communication skills training, social resource connection, etc.

[0006] The above technical solution further includes: Furthermore, the data collection methods in the data collection layer include questionnaire surveys, clinical interviews, and family records. Design a questionnaire covering multiple dimensions and distribute it to target families through online or offline methods. The questionnaire content covers basic family information (such as the age, gender, occupation of family members, etc.), medical history (including the diagnosis, treatment, and rehabilitation of stroke patients, etc.), family member relationships (such as parent-child relationships, marital relationships, etc.), and social resources (such as social support networks, community participation, etc.). Encode the open-ended questions in the questionnaire and convert the text data into numerical data. At the same time, standardize each indicator to eliminate the dimensional differences between different indicators. The clinical interview is conducted by professional medical staff or psychologists to interview stroke patients and their family members face-to-face or remotely to understand the patient's medical history, rehabilitation progress, mental state, as well as the care burden and emotional reactions of family members. The family records are collected by gathering document materials such as family daily records, medical records, and rehabilitation logs. When processing text data such as clinical interview records and family records, use natural language processing technology to extract key information and conduct sentiment analysis.

[0007] Furthermore, the specific steps for constructing the knowledge graph of family resilience in the knowledge graph construction layer are as follows: Data preprocessing: Clean the collected data, remove invalid, redundant, or incorrect information, and standardize the data to ensure consistent data formats for subsequent processing; Natural Language Processing: Using natural language processing techniques to extract key information from text data, such as the roles of family members, the strength of family relationships, resilience strategies, etc.; Knowledge Graph Construction: Based on the extracted key information, construct a knowledge graph of family resilience, where the knowledge graph includes nodes (such as family members, resilience strategies, etc.) and edges (such as family relationships, the corresponding relationships between strategies and challenges, etc.); Semantic Rule Introduction: Introduce semantic rules to describe the relationships between family members and the corresponding relationships between resilience strategies and challenges; Graph Optimization and Verification: Optimize the constructed knowledge graph, such as removing redundant nodes and edges, adjusting the weights of nodes and edges, etc., and verify the optimized graph to ensure the accuracy and reliability of the graph.

[0008] Furthermore, using natural language processing techniques to extract key information from text data, the processing process includes the following steps: Text Preprocessing: Clean and format the original text, remove irrelevant characters (such as punctuation marks, numbers, etc.) from the text, perform word segmentation, and split the text into individual words or phrases; Part-of-Speech Tagging: Determine the part of speech of each word (such as noun, verb, adjective, etc.); Named Entity Recognition: Identify specific entities in the text (such as person names, place names, organization names, etc.); Relationship Extraction: Through rule matching, identify the semantic relationships between entities in the text; Key Information Determination: Define the criteria for key information, where the criteria for key information include family member roles, family relationship strength, and resilience strategies. Filter out relationships that are not directly related to family resilience assessment according to the criteria, and integrate the filtered relationships into a knowledge graph.

[0009] Furthermore, the evaluation model layer processes the individual's questionnaire survey data using logistic regression, and quantifies the assessment of family resilience at the individual level through feature selection and weight assignment, including the following steps: Feature Selection: Screen out key features related to family resilience assessment from the individual's questionnaire survey data, preprocess the questionnaire survey data, use correlation analysis to screen out key features, and further verify and screen features based on relevant theories and empirical studies of family resilience; Weight Assignment: Assign a weight to each key feature to reflect its importance in family resilience assessment, and use a logistic regression model for training. The formula of the logistic regression model is , where Y represents family resilience at the individual level, usually taking values of 0 or 1, representing low resilience or high resilience, and X represents the feature vector. represents the intercept term, represents the weights of each feature. The model automatically assigns a weight to each feature. Based on the training results of the model, the weight values of each feature are extracted. Combining with the relevant theories and empirical studies of family resilience, the weight values are interpreted and adjusted; Quantitative evaluation: Based on the results of feature selection and weight assignment, quantitatively evaluate the family resilience at the individual level. Convert the individual's questionnaire survey data into a feature vector, and use a logistic regression model to predict the feature vector to obtain the family resilience score at the individual level. According to the score, quantitatively evaluate the family resilience at the individual level.

[0010] Furthermore, the evaluation model layer uses the Graph Convolutional Network (GCN) to evaluate the resilience at the family level. The GCN captures the relationships between family members and evaluates the overall resilience level of the family through the feature representations of nodes and edges, including the following steps: Feature extraction: A knowledge graph containing family relationship data. The nodes in the knowledge graph represent family members, and the edges represent the relationships between members (such as father-son, husband-wife, etc.). Extract features for each node (family member) and edge (relationship). The features include individual features such as age, gender, education level, occupation, health status, etc., and relationship features such as relationship type and relationship strength; Definition of the GCN layer: Define the GCN layer for capturing the relationship features between family members. The GCN layer updates the representation of the current node by aggregating the information of neighboring nodes. The input of the GCN layer is the node feature matrix X and the adjacency matrix A, and the output is the updated node representation matrix H. The update formula of the GCN layer is expressed as , where, is the node representation matrix of the l-th layer, is the weight matrix of the l-th layer, is the activation function; Information propagation: Perform information propagation on the graph, that is, each node updates its own representation by aggregating the information of its neighboring nodes. The information propagation is iterated multiple times to capture the information of farther neighbors; Feature fusion: Integrate the individual features and relationship features into the node representation to form a new node vector containing rich information; Aggregation of node vectors: For the evaluation of family resilience at the family level, aggregate the node vectors of all members within the family to form an overall representation of the family; Resilience calculation: Based on the aggregated family representation, use an evaluation function (such as linear regression, neural network, etc.) to calculate the family resilience level. The aggregated family representation is , then the family resilience level R is expressed as , where f is the evaluation function.

[0011] Furthermore, the evaluation model layer uses the Graph Convolutional Network (GCN) to combine the social relationship data in the knowledge graph to evaluate the impact of social relationships on family resilience. By the connection strength and importance between nodes, it quantifies the resilience contribution at the social relationship level, including the following steps: Social relationship data collection: Extract the social relationship data between family members from the knowledge graph, including relationship types, relationship strengths, etc.; Family information integration: Integrate the social relationship data with the individual information of family members to form a complete family dataset; Graph structure definition: Represent the family dataset as a graph structure, where nodes represent family members and edges represent the relationships between members; Feature vector initialization: Initialize the feature vectors for each node (family member), and these feature vectors contain basic individual information, questionnaire survey results, etc.; GCN layer stacking: Construct multiple layers of GCN, and each layer updates the feature vector of the current node by aggregating the features of neighbor nodes; Relationship strength calculation: Use the edge weights in GCN to represent the relationship strength, which can be learned or set based on the prior information in the knowledge graph; Node importance evaluation: Through the output of GCN, evaluate the importance of each node (family member) in the social relationship network, which reflects the contribution degree of the member in family resilience; Resilience calculation at the social relationship level: Based on the connection strength and importance between nodes, calculate the resilience contribution at the social relationship level, and the calculation formula is , where represents considering the feature vector of node i and its relationship matrix function.

[0012] Furthermore, the evaluation model layer combines the relevant theories and empirical studies of family resilience, and determines the weights of the individual, family, and social relationship levels in the evaluation model through regression analysis, including the following steps: Construct a correlation matrix: Calculate the correlation coefficients between variables and construct a correlation matrix, which is represented as , where represents the correlation coefficient between variable i and variable j; Extract factors: Use principal component analysis to extract factors, and determine the number of factors and the variables included in each factor; Factor rotation: Rotate the extracted factors to better explain the meaning of the factors; Calculate factor scores: According to the factor loading matrix, the factor loading matrix is represented as , where represents the loadings on variables i and factor j, calculates the scores of each household on each factor, and the factor scores are expressed as , where is the original data matrix, is the transpose of the factor loading matrix, is the inverse matrix of the correlation matrix.

[0013] The present invention has the following beneficial effects: In the present invention, the collected individual-level data is processed and analyzed using the logistic regression method. Based on the results of the logistic regression, the resilience level of each member is quantitatively evaluated. Combining the family relationship data in the knowledge graph, including the roles and relationship strengths of family members, a family-level evaluation framework is constructed. The graph neural network GCN is used to capture the relationships and interactions between family members. Through the feature representations of nodes and edges, the overall resilience level of the family is evaluated, and the family-level resilience is quantitatively evaluated. Combining the social relationship data in the knowledge graph, the graph neural network GCN is used to analyze the impact of social relationships on family resilience. Through the connection strength and importance between nodes, the resilience contribution at the social relationship level is quantitatively evaluated, and the evaluation results at the individual, family, and social relationship levels are integrated to form a comprehensive evaluation of family resilience. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 is the system block diagram of the family resilience evaluation system based on the knowledge graph proposed by the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0015] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0016] Please refer to Figure 1 shown. The present invention is a family resilience evaluation system based on the knowledge graph, including: Data acquisition layer: Collect the basic information, medical history, family member relationships, and social resources of families of stroke patients in the community; Knowledge Graph Construction Layer: Based on the collected data, construct a knowledge graph of family resilience. The knowledge graph includes relevant concepts, attributes, and relationships of family resilience, such as the roles of family members, the strength of family relationships, social resource support, etc. Semantic rules are introduced during the construction of the knowledge graph to describe the relationships between family members and the corresponding relationships between resilience strategies and challenges. Key information is extracted from text data using natural language processing to enrich the knowledge graph; Evaluation Model Layer: Based on three levels of individual, family, and social relationships, construct an evaluation model of family resilience. Conduct quantitative evaluation based on the data in the knowledge graph. Use logistic regression to process the individual's questionnaire survey data. Through feature selection and weight assignment, quantitatively evaluate the family resilience at the individual level. Combine the family relationship data in the knowledge graph and use the graph neural network GCN to evaluate the family-level resilience. The graph neural network GCN captures the relationships between family members and evaluates the overall resilience level of the family through the feature representation of nodes and edges. Use the graph neural network GCN and combine the social relationship data in the knowledge graph to evaluate the impact of social relationships on family resilience. Quantify the resilience contribution at the social relationship level through the connection strength and importance between nodes. Combine the relevant theories and empirical research of family resilience, and determine the weights of the three levels of individual, family, and social relationships in the evaluation model through regression analysis. Perform weighted summation according to the weights to obtain the comprehensive evaluation result of family resilience; Result Display and Application Layer: Display the evaluation results to users and provide personalized family intervention suggestions according to the evaluation results, such as family communication skills training, social resource connection, etc.

[0017] In one embodiment, the data collection methods in the data collection layer include questionnaire surveys, clinical interviews, and family records. Design a questionnaire covering multiple dimensions and distribute it to target families through online or offline methods. The questionnaire content covers basic family information (such as the age, gender, occupation, etc. of family members), medical history (including the diagnosis, treatment, and rehabilitation status of stroke patients, etc.), family member relationships (such as parent-child relationships, marital relationships, etc.), and social resources (such as social support networks, community participation, etc.). Encode the open-ended questions in the questionnaire to convert the text data into numerical data. At the same time, standardize each indicator to eliminate the dimensional differences between different indicators. The clinical interview is conducted by professional medical staff or psychologists to interview stroke patients and their family members face-to-face or remotely to understand the patient's medical history, rehabilitation progress, psychological state, as well as the care burden and emotional reactions of family members. The family records are collected by gathering document materials such as family daily records, medical records, and rehabilitation logs. When processing text data such as clinical interview records and family records, use natural language processing technology to extract key information and perform sentiment analysis.

[0018] In one embodiment, the specific steps for the knowledge graph construction layer to construct the knowledge graph of family resilience are as follows: Data preprocessing: Clean the collected data, remove invalid, redundant or incorrect information, and standardize the data to ensure consistent data formats for subsequent processing; Natural language processing: Use natural language processing techniques to extract key information from text data, such as the roles of family members, the strength of family relationships, resilience strategies, etc.; Knowledge graph construction: Construct a knowledge graph of family resilience based on the extracted key information. The knowledge graph includes nodes (such as family members, resilience strategies, etc.) and edges (such as family relationships, the corresponding relationships between strategies and challenges, etc.); Semantic rule introduction: Introduce semantic rules to describe the relationships between family members and the corresponding relationships between resilience strategies and challenges; Graph optimization and verification: Optimize the constructed knowledge graph, such as removing redundant nodes and edges, adjusting the weights of nodes and edges, etc., and verify the optimized graph to ensure the accuracy and reliability of the graph.

[0019] In one embodiment, the process of using natural language processing techniques to extract key information from text data includes the following steps: Text preprocessing: Clean and format the original text, remove irrelevant characters (such as punctuation marks, numbers, etc.) from the text, and perform word segmentation to split the text into individual words or phrases; Part-of-speech tagging: Determine the part of speech of each word (such as noun, verb, adjective, etc.); Named entity recognition: Identify specific entities in the text (such as person names, place names, organization names, etc.); Relationship extraction: Identify the semantic relationships between entities in the text through rule matching; Key information determination: Define the criteria for key information. The criteria for key information include the roles of family members, the strength of family relationships, and resilience strategies. Filter out relationships that are not directly related to family resilience assessment according to the criteria, and integrate the filtered relationships into a knowledge graph.

[0020] A multi-dimensional knowledge graph including family member roles, family relationship strength, and resilience strategies is constructed. This knowledge graph adopts a triple scoring system and configures the following quantitative indicators for each family member node: (1) Role scoring value, calculated based on the role positioning of family members in the family structure and the weight of their responsibilities; (2) Relationship strength scoring value, quantitatively evaluated by analyzing the tightness of the association between this member and other family member nodes; (3) Strategy effectiveness scoring value, objectively evaluated based on the implementation effect and adaptability of the resilience strategies adopted by the member. Through weighted calculation, the above scoring system can achieve a comprehensive quantitative evaluation of the family resilience level.

[0021] For example: , where is the weight coefficient, used to adjust the contribution of different scores to the comprehensive score. These weight coefficients can be adjusted and optimized according to the actual situation.

[0022] In actual operation, these weight coefficients and score values are determined through methods such as expert scoring, questionnaire surveys, or machine learning. Then, these scores and weights are used to calculate the comprehensive score of each family member, and the importance and priority of key information are determined according to the score level.

[0023] In one embodiment, the evaluation model layer processes the individual's questionnaire survey data using logistic regression, and through feature selection and weight assignment, quantitatively evaluates the family resilience at the individual level, including the following steps: Feature selection: Screen out the key features related to family resilience evaluation from the individual's questionnaire survey data, preprocess the questionnaire survey data, use correlation analysis to screen out the key features, and further verify and screen the features based on the relevant theories and empirical studies of family resilience; Weight assignment: Assign a weight to each key feature to reflect its importance in family resilience evaluation, and use a logistic regression model for training. The formula of the logistic regression model is , where Y represents the family resilience at the individual level, usually taking values of 0 or 1, indicating low resilience or high resilience, X represents the feature vector, represents the intercept term, represents the weights of each feature. The model automatically assigns a weight to each feature. According to the training results of the model, the weight values of each feature are extracted, and combined with the relevant theories and empirical studies of family resilience, the weight values are interpreted and adjusted; Quantitative evaluation: Based on the results of feature selection and weight assignment, quantitatively evaluate the family resilience at the individual level. Convert the individual's questionnaire survey data into a feature vector, use a logistic regression model to predict the feature vector, obtain the family resilience score at the individual level, and quantitatively evaluate the family resilience at the individual level according to the score.

[0024] Feature selection: Through correlation analysis and feature importance evaluation, we selected 4 key features: age, education level, family income, and family support level.

[0025] Weight assignment: Use a logistic regression model for training, and obtain the weight values of each feature as follows: Age: Education level: Family income: β3 = 0.4 Family support level: Quantitative evaluation: Convert the individual's questionnaire survey data into a feature vector. For example, the feature vector of a certain individual is . Use a logistic regression model for prediction, and obtain the family resilience score at this individual level as 0.8 (assuming the result after sigmoid function conversion). According to the score, we can judge that this individual has a relatively high family resilience.

[0026] In one embodiment, the evaluation model layer uses the graph neural network GCN to evaluate the resilience at the family level. The graph neural network GCN captures the relationships between family members and evaluates the overall resilience level of the family through the feature representations of nodes and edges, including the following steps: Feature extraction: A knowledge graph containing family relationship data. The nodes in the knowledge graph represent family members, and the edges represent the relationships between members (such as father-son, husband-wife, etc.). Extract features for each node (family member) and edge (relationship). The features include individual features such as age, gender, education level, occupation, and health status, as well as relationship features such as relationship type and relationship strength; Definition of the GCN layer: Define the GCN layer for capturing the relationship features between family members. The GCN layer updates the representation of the current node by aggregating the information of neighboring nodes. The input of the GCN layer is the node feature matrix X and the adjacency matrix A, and the output is the updated node representation matrix H. The update formula of the GCN layer is expressed as , where, is the node representation matrix of the l-th layer, is the weight matrix of the l-th layer, is the activation function; Information dissemination: Information is disseminated on the graph, that is, each node updates its own representation by aggregating the information of its neighbor nodes. The information dissemination is iterated multiple times to capture the information of more distant neighbors; Feature fusion: The individual features and relationship features are fused into the node representation to form a new node vector containing rich information; Node vector aggregation: For the assessment of family-level resilience, the node vectors of all members within the family are aggregated to form an overall representation of the family; Resilience calculation: Based on the aggregated family representation, an evaluation function (such as linear regression, neural network, etc.) is used to calculate the family resilience level. The aggregated family representation is , then the family resilience level R is expressed as , where f is the evaluation function.

[0027] In one embodiment, the evaluation model layer uses a graph neural network GCN, combines the social relationship data in the knowledge graph, evaluates the impact of social relationships on family resilience, and quantifies the resilience contribution at the social relationship level through the connection strength and importance between nodes, including the following steps: Social relationship data collection: Extract the social relationship data between family members from the knowledge graph, including relationship types, relationship strengths, etc.; Family information integration: Integrate the social relationship data with the individual information of family members to form a complete family data set; Graph structure definition: Represent the family data set as a graph structure, where nodes represent family members and edges represent the relationships between members; Feature vector initialization: Initialize the feature vectors for each node (family member), and these feature vectors contain individual basic information, questionnaire survey results, etc.; GCN layer stacking: Build multiple layers of GCN, and each layer updates the feature vector of the current node by aggregating the features of neighbor nodes; Relationship strength calculation: Use the edge weights in GCN to represent the relationship strength, which can be learned or set based on the prior information in the knowledge graph; Node importance evaluation: Through the output of GCN, evaluate the importance of each node (family member) in the social relationship network, which reflects the contribution degree of the member in family resilience; Social relationship level resilience calculation: Based on the connection strength and importance between nodes, calculate the resilience contribution at the social relationship level. The calculation formula is , where represents considering the feature vector of node i and its relationship matrix function.

[0028] In one embodiment, the evaluation model layer combines the relevant theories and empirical studies of family resilience, and determines the weights of the individual, family, and social relationship levels in the evaluation model through regression analysis, including the following steps: Construct a correlation matrix: Calculate the correlation coefficients between variables and construct a correlation matrix, which is expressed as , where represents the correlation coefficient between variable i and variable j; Extract factors: Use principal component analysis to extract factors and determine the number of factors and the variables included in each factor; Factor rotation: Rotate the extracted factors to better explain the meaning of the factors; Calculate factor scores: According to the factor loading matrix, which is expressed as , where represents the loading of variable i on factor j, calculate the scores of each family on each factor, and the factor scores are expressed as , where is the original data matrix, is the transpose of the factor loading matrix, is the inverse matrix of the correlation matrix.

[0029] Although the embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The family resilience assessment system based on knowledge graph is characterized by: include: Data collection layer: Collect basic information, medical history, family member relationships, and social resources of the families of community stroke patients; Knowledge graph construction layer: Based on the collected data, a knowledge graph of family resilience is constructed. The knowledge graph includes relevant concepts, attributes, and relationships of family resilience. Semantic rules are introduced in the process of constructing the knowledge graph to describe the relationships between family members and the correspondence between resilience strategies and challenges. Natural language processing is used to extract key information from text data to enrich the knowledge graph. Evaluation model layer: Based on the three levels of individual, family, and social relations, an evaluation model for family resilience is constructed. A quantitative evaluation is performed based on the data in the knowledge graph. Logistic regression is used to process the individual questionnaire data. Through feature selection and weight allocation, the family resilience at the individual level is quantitatively evaluated. In combination with the family relationship data in the knowledge graph, the family resilience is evaluated using the graph neural network GCN. The graph neural network GCN captures the relationship between family members and evaluates the overall family resilience level through the feature representation of nodes and edges. The graph neural network GCN is used in combination with the social relationship data in the knowledge graph to evaluate the impact of social relationships on family resilience. The contribution of resilience at the social relationship level is quantified through the connection strength and importance between nodes. In combination with relevant theories and empirical studies on family resilience, the weights of the three levels of individual, family, and social relations in the evaluation model are determined through regression analysis. The weighted sum is performed according to the weights to obtain a comprehensive evaluation result of family resilience. Result presentation and application layer: The evaluation results are presented to users, and personalized family intervention recommendations are provided based on the evaluation results.

2. The family resilience assessment system based on knowledge graph according to claim 1 is characterized in that: The data collection layer uses methods including questionnaire surveys, clinical interviews, and family records. A questionnaire with multiple dimensions is designed and distributed to target families online or offline. The questionnaire covers basic family information, medical history, relationships among family members, and social resources. The open-ended questions in the questionnaire are coded, and text data is converted into numerical data. At the same time, each indicator is standardized to eliminate the dimensional differences between different indicators. The clinical interview is a face-to-face or remote interview conducted by professional medical staff or psychological counselors with stroke patients and their family members to understand the patient's medical history, rehabilitation progress, psychological state, and the care burden and emotional response of family members. The family records are collected by collecting family document materials. When it comes to text data, natural language processing technology is used to extract key information and perform sentiment analysis.

3. The family resilience assessment system based on knowledge graph according to claim 1 is characterized in that: The specific steps of the knowledge graph construction layer to construct the knowledge graph of family resilience are as follows: Data preprocessing: clean the collected data, remove invalid, redundant or erroneous information, and standardize the data; Natural Language Processing: Use natural language processing technology to extract key information from text data; Knowledge graph construction: constructing a knowledge graph of family resilience based on the extracted key information, wherein the knowledge graph includes nodes and edges; Introduction of semantic rules: Introduce semantic rules to describe the relationships between family members and the correspondence between resilience strategies and challenges; Graph optimization and verification: Optimize the constructed knowledge graph and verify the optimized graph.

4. The family resilience assessment system based on knowledge graph according to claim 1, characterized in that: Using natural language processing technology to extract key information from text data, the processing process includes the following steps: Text preprocessing: cleaning and formatting the original text, removing irrelevant characters from the text, performing word segmentation, and splitting the text into individual words or phrases; Part-of-speech tagging: determine the part of speech of each word; Named Entity Recognition: Identifying specific entities in text; Relation extraction: Identify the semantic relationship between entities in the text through rule matching; Key information determination: define the criteria for key information, which include family member roles, family relationship strength, and resilience strategies. Filter out relationships that are not directly related to family resilience assessment based on the criteria, and integrate the filtered relationships into a knowledge graph.

5. The family resilience assessment system based on knowledge graph according to claim 4 is characterized in that: The assessment model layer processes the individual questionnaire data using logistic regression, and quantitatively assesses the family resilience at the individual level through feature selection and weight allocation, including the following steps: Feature selection: Screen out key features related to family resilience assessment from individual questionnaire survey data, pre-process the questionnaire survey data, use correlation analysis to screen out key features, and further verify and screen features based on relevant theories and empirical research on family resilience; Weight assignment: A weight is assigned to each key feature to reflect its importance in the assessment of family resilience. The logistic regression model is used for training. The formula of the logistic regression model is: , where Y represents the family resilience at the individual level, usually taking the value of 0 or 1, indicating low resilience or high resilience, and X represents the eigenvector. represents the intercept term, Represents the weight of each feature. The model automatically assigns a weight to each feature. Based on the model training results, the weight value of each feature is extracted. Combined with relevant theories and empirical research on family resilience, the weight value is explained and adjusted; Quantitative assessment: Based on the results of feature selection and weight allocation, quantitative assessment of family resilience at the individual level is conducted. The individual questionnaire data is converted into feature vectors, and the feature vectors are predicted using a logistic regression model to obtain the individual-level family resilience score. Based on the scores, a quantitative assessment of family resilience at the individual level is conducted.

6. The family resilience assessment system based on knowledge graph according to claim 1, characterized in that: The assessment model layer uses the graph neural network GCN to perform family-level resilience assessment. The graph neural network GCN captures the relationship between family members and assesses the overall family resilience level through node and edge feature representation, including the following steps: Feature extraction: A knowledge graph containing family relationship data, where nodes represent family members and edges represent relationships between members, and features are extracted for each node and edge; GCN layer definition: Define the GCN layer to capture the relationship characteristics between family members. The GCN layer updates the representation of the current node by aggregating the information of neighboring nodes. The input of the GCN layer is the node feature matrix X and the adjacency matrix A, and the output is the updated node representation matrix H. The update formula of the GCN layer is expressed as ,in, is the node representation matrix of the lth layer, is the weight matrix of the lth layer, is the activation function; Information propagation: Information propagation is performed on the graph, that is, each node updates its own representation by aggregating the information of its neighbor nodes. Information propagation is iterated multiple times to capture information about more distant neighbors. Feature fusion: individual features and relationship features are fused into node representation to form a new node vector; Node vector aggregation: For the resilience assessment at the family level, the node vectors of all family members are aggregated to form an overall representation of the family; Resilience calculation: Based on the aggregated household representation, an evaluation function is used to calculate the household resilience level. The aggregated household representation is , then the family resilience level R is expressed as , where f is the evaluation function.

7. The family resilience assessment system based on knowledge graph according to claim 1, characterized in that: The evaluation model layer uses the graph neural network GCN and combines the social relationship data in the knowledge graph to evaluate the impact of social relationships on family resilience. It quantifies the contribution of resilience at the social relationship level through the connection strength and importance between nodes, including the following steps: Social relationship data collection: extracting social relationship data between family members from the knowledge graph; Family information integration: Integrate social relationship data with individual information of family members to form a complete family data set; Graph structure definition: The family dataset is represented as a graph structure, where nodes represent family members and edges represent the relationships between members; Feature vector initialization: Initialize the feature vector for each node; GCN layer stacking: construct multiple layers of GCN, each layer updates the feature vector of the current node by aggregating the features of neighboring nodes; Relationship strength calculation: Use edge weights in GCN to represent relationship strength; Node importance evaluation: Through the output of GCN, the importance of each node in the social relationship network is evaluated, which reflects the contribution of members to family resilience; Calculation of resilience at the social relationship level: Based on the connection strength and importance between nodes, the resilience contribution at the social relationship level is calculated using the formula: ,in, Denotes the feature vector of node i And its relationship matrix function.

8. The family resilience assessment system based on knowledge graph according to claim 1, characterized in that: The assessment model layer combines relevant theories and empirical research on family resilience, and determines the weights of the three levels of individual, family, and social relationship in the assessment model through regression analysis, including the following steps: Construct a correlation matrix: Calculate the correlation coefficients between the variables and construct a correlation matrix, which is expressed as ,in represents the correlation coefficient between variable i and variable j; Extract factors: Use principal component analysis to extract factors and determine the number of factors and the variables contained in each factor; Factor rotation: Rotate the extracted factors; Calculate factor scores: According to the factor loading matrix, the factor loading matrix is ​​expressed as ,in represents the load on variable i and factor j, and calculates the score of each family on each factor. The factor score is expressed as ,in is the original data matrix, is the transpose of the factor loading matrix, is the inverse of the correlation matrix.

Images