Intelligent resident information management system based on knowledge graph

Abstract

The invention discloses a resident information intelligent management system based on a knowledge graph, and the system comprises the following modules: a behavior pattern reasoning module which is used for generating a behavior change trend based on a reasoning model of SPARQL query; the health risk assessment module is used for analyzing health data and reasoning results of residents and generating personalized health management suggestions; the resource allocation optimization module is used for generating a resource allocation optimization scheme according to the demand analysis and behavior mode of the resident; the personalized service recommendation module is used for generating personalized service recommendation according to the behavior pattern and health risk analysis of the resident; the decision support result display module is used for displaying a group of decision support results and displaying the results through an interface or in a visual mode; and the data sharing and interaction module is used for carrying out data interaction and sharing with an external system through an open API (Application Program Interface). The efficient resident information intelligent management system is constructed by combining the knowledge graph and the intelligent reasoning technology.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent management and data analysis, and particularly relates to a resident information intelligent management system based on a knowledge graph. BACKGROUND

[0002] With the continuous improvement of social informatization, the collection and management of resident information has gradually become an important part of urban governance and social services. Traditional resident information management systems are usually based on relational databases and managed through manual entry and manual updates. The data is in an isolated state and cannot be effectively integrated and associated. Such systems have certain limitations in data storage and management, especially when dealing with large-scale resident information and implementing dynamic analysis of data, they often face problems of low efficiency and untimely response.

[0003] Most current resident information management systems rely on manual intervention and lack intelligent data processing and analysis methods. The main operations that the system can perform include basic queries, statistics, and report generation, and lack intelligent analysis and prediction of deep information such as resident behavior patterns, health risks, and social needs. The fragmentation, isolated storage, and dynamic changes of resident information make traditional methods weak in prediction, decision-making, and resource optimization allocation. In addition, existing systems usually do not have effective cross-department and cross-system data sharing mechanisms, resulting in the existence of information silos, making it difficult to provide comprehensive support for decision-makers.

[0004] With the continuous development of artificial intelligence and big data technology, intelligent decision support systems have emerged. In particular, the introduction of knowledge graphs provides a new solution for the integration and intelligent analysis of resident information. Knowledge graphs can associate various scattered resident information through graph structures, providing a rich relationship network between entities, and enabling more accurate reasoning and prediction based on this. However, the existing knowledge graph technology still has some limitations in the application of resident information management, especially in terms of reasoning accuracy, resource optimization allocation, and personalized service recommendation.

[0005] Traditional rule-based reasoning methods often rely on manually defined rules, lack flexibility and adaptability, and are difficult to automate efficient reasoning. In addition, existing intelligent management systems often fail to effectively combine dynamic behavior data of residents, especially lacking deep analysis of time series data. Traditional methods often ignore the distinction between long-term trends and short-term fluctuations, resulting in low accuracy in predicting resident health risks, social needs, and behavior patterns. Existing technologies also have bottlenecks in data sharing and cross-system collaboration, and the interconnection between systems is weak, limiting the collaborative work of multi-party data.

[0006] Therefore, how to provide a resident information intelligent management system based on a knowledge graph is a problem that those skilled in the art need to solve urgently. SUMMARY

[0007] One object of the present application is to provide a resident information intelligent management system based on a knowledge graph. The present application combines knowledge graphs, intelligent reasoning, and long short-term memory networks to construct an efficient resident information intelligent management system. The system can break down data silos, integrate multi-source data, and accurately predict resident behavior patterns, health risks, and social needs through reasoning calculations. The system uses LSTM to process time series data, identify long-term trends and short-term fluctuations, and improve prediction accuracy. In addition, the system shares data and collaborates with external systems through open API interfaces, providing scientific and real-time decision support for decision makers.

[0008] According to the resident information intelligent management system based on the knowledge graph, the following modules are included:

[0009] A behavior pattern reasoning module is used to generate resident future behavior pattern predictions based on resident behavior data and social needs in the knowledge graph using a reasoning model based on SPARQL queries, and to generate behavior change trends in combination with health risk analysis.

[0010] A health risk assessment module is used to analyze resident health data and reasoning results, identify high-risk groups, and generate personalized health management recommendations based on prediction results.

[0011] A resource allocation optimization module is used to automatically generate resource allocation optimization schemes based on resident demand analysis and behavior patterns, in combination with health risk assessment, and to ensure that resources can be efficiently and accurately allocated according to demand.

[0012] A personalized service recommendation module is used to generate personalized service recommendations based on resident behavior patterns and health risk analysis, ensuring that residents can receive health services, social activities, or social resources that meet their needs.

[0013] A decision support result display module is used to present a set of decision support results based on generated health management recommendations, resource allocation optimization schemes, and personalized service recommendations for decision makers to reference, and to display the results through an interface or visual means.

[0014] A data sharing and interaction module is used to interact and share data with external systems through open API interfaces, support data integration with medical systems, social platforms, and other public service systems, and realize multi-party data collaboration to promote cross-system data integration and management.

[0015] According to the embodiment of the present application, a knowledge graph-based intelligent management method for resident information comprises the following steps:

[0016] Optionally, the modules are connected through the following method:

[0017] By collecting multi-dimensional information from multiple information sources, and converting the multi-dimensional information into nodes and edges in the knowledge graph according to the correlation between the multi-dimensional information, wherein the nodes represent various entities, and the edges represent the relationship between the entities;

[0018] Using natural language processing technology and ontology construction method, the various attributes and relationships of the entity nodes in the knowledge graph are identified and labeled;

[0019] Using the improved AMIE+ algorithm, a set of logical rules is generated from the constructed knowledge graph;

[0020] According to the set of logical rules, using the inference model based on SPARQL query, combining the inference rules and the entity relationship in the knowledge graph, the inference calculation is performed to predict the future behavior pattern, health risk and social demand of the residents;

[0021] Based on the long short-term memory network, the time series data of the resident behavior is processed to identify the long-term trend and short-term fluctuation, and the prediction of the health risk and social demand of the residents is optimized;

[0022] According to the behavior pattern, health risk and social demand of the residents obtained by reasoning in the knowledge graph, a set of decision support results is generated, and data interaction and sharing are performed with external systems through an open API interface.

[0023] Optionally, the multi-dimensional information collected from multiple information sources is converted into nodes and edges in the knowledge graph according to the correlation between the multi-dimensional information, wherein the nodes represent various entities, and the edges represent the relationship between the entities, and specifically comprises:

[0024] The multi-dimensional information of the residents is collected from multiple information sources, and the multi-dimensional information includes personal information, health data, social relationship, behavior record, medical record and consumption behavior data of the residents;

[0025] The collected multi-dimensional information is preprocessed, and the preprocessing includes data cleaning, missing value filling, data standardization and data format conversion to ensure the quality and consistency of the information;

[0026] The multi-dimensional information is mapped to entity nodes in the knowledge graph through RDF knowledge graph construction technology, and the entity nodes include residents, diseases, health indicators, social relationships and behavior types;

[0027] Identify and extract the relationship between various types of entity nodes, including social relationships between residents, health status association and behavior pattern similarity, and convert the relationship between various types of entity nodes into edges in the knowledge graph, each edge representing the specific association between two entity nodes.

[0028] Optionally, the natural language processing technology and ontology construction method are used to identify and label various attributes and relationships of entity nodes in the knowledge graph, specifically including:

[0029] Using the word segmentation, part-of-speech tagging and syntax analysis in natural language processing technology, the text data in the resident information is processed to identify the entity attributes and relationships in the text data, and the corresponding relationship between the entity and the attribute is constructed;

[0030] Using the ontology construction method, by defining various concepts, attributes and relationships in the resident information, the concepts include disease, resident and health, the attributes include disease type and age range, and the relationships include kinship, a unified semantic standard is established, so that the entities and relationships in the knowledge graph can conform to the predefined ontology model;

[0031] Each entity node in the knowledge graph is labeled, including assigning a corresponding category label to each entity node, the category label includes resident, disease and hospital, and the attributes of each entity node are semantically labeled, the attributes include birth date and disease time, ensuring clear information and standardized semantics;

[0032] Using relationship extraction technology, the relationship between entities in the resident information is extracted and labeled, including social relationship, health status association and behavior pattern similarity, the relationship between entities is labeled as an edge in the knowledge graph, ensuring that the semantics of each relationship is clear and accurately reflects the association between entities in the graph;

[0033] The labeled knowledge graph is verified and quality checked to ensure that the labeling of all entity nodes and edges conforms to the predefined ontology model and semantic standard, ensuring consistent and complete information in the knowledge graph, and supporting further data processing, reasoning and analysis.

[0034] Optionally, the improved AMIE+ algorithm specifically includes:

[0035] The improved AMIE+ algorithm uses a multi-layer nonlinear transformation function to perform high-dimensional embedding representation on entity nodes and relationship edges in the knowledge graph, and converts each entity node and relationship edge into a point in a low-dimensional vector space through nonlinear mapping to generate a nonlinear embedding vector; so that each relationship in the knowledge graph not only retains the direct relationship between entities, but also can capture implicit and complex nonlinear dependency relationships. Through this method, the semantic expression ability between entity nodes is enhanced, especially for complex multi-level relationships, providing deeper structural information. This nonlinear embedding result provides a basis for subsequent implicit association reasoning and rule generation.

[0036] Based on the nonlinear embedding vector, the improved AMIE+ algorithm uses an implicit association reasoning method to reason about potential complex associations between entity nodes, identify and extract potential deep association relationships from the knowledge graph, and generate a preliminary candidate logic rule set by evaluating the similarity and co-occurrence patterns between different entity pairs in the high-dimensional embedding space; such as residents with similar health conditions may have similar lifestyles, behavior patterns and potential associations with health conditions, etc. This method further explores the potential unexpressed associations through similarity measurement and mutual influence calculation in the embedding space, and the generated candidate logic rule set has high potential in this step and provides a basis for causal reasoning.

[0037] The improved AMIE+ algorithm introduces a causal graph reconstruction method, which performs causal analysis on entities and relationships in the knowledge graph and generates causal effect weights;

[0038] In combination with the nonlinear embedding vector, the candidate logic rule set and the causal effect weight, the improved AMIE+ algorithm prioritizes the candidate logic rule set, preferentially selects candidate logic rules with clear causal relationships and accurate behavior pattern prediction effects according to the causal effect weight of the candidate logic rule, and simultaneously performs weighted optimization of the candidate logic rule using the causal effect weight;

[0039] By calculating the reasoning accuracy of each candidate logic rule under different situations, candidate logic rules with high causal explanatory power and reasoning accuracy are selected, and a final logic rule set is generated. The final logic rule set can not only efficiently predict resident behavior patterns, health risks and social needs, but also adapt to dynamic changes in knowledge graph data and diverse needs.

[0040] Optionally, the causal graph reconstruction method introduced by the improved AMIE+ algorithm performs causal analysis on entities and relationships in the knowledge graph and generates causal effect weights, and specifically includes:

[0041] The entity nodes and relationship edges in the knowledge graph are structurally represented, and the relationships in the knowledge graph are abstracted as causal relationship chains;

[0042] Based on the causal chain, the improved AMIE+ algorithm constructs a causal relationship graph according to the semantic and structural information of each entity and relationship in the knowledge graph through dynamic causal analysis, wherein the nodes in the causal relationship graph represent entities, and the edges represent the causal relationship between entities;

[0043] Based on the causal relationship graph, the improved AMIE+ algorithm evaluates the changes of each candidate logical rule under different hypothetical conditions by setting different hypothetical conditions, generates counterfactual scenarios, simulates the changes of candidate logical rules under different counterfactual scenarios, and analyzes the potential influence of each candidate logical rule on entity behavior and relationship under hypothetical conditions;

[0044] The improved AMIE+ algorithm quantifies the causal reasoning effect of each candidate logical rule, measures the causal dependence relationship between entity nodes in the causal relationship graph, calculates the influence strength of each candidate logical rule on behavior prediction, generates causal effect weight, and the causal effect weight reflects the influence degree of candidate logical rule on resident behavior, health risk and social demand under different conditions.

[0045] The selected candidate logical rule is bound to the causal relationship between the actual behavior and entity in the knowledge graph, ensuring that each candidate logical rule can not only capture the correlation between entities in the graph, but also reveal the potential causal chain and causal influence, thereby improving the explainability and reasoning accuracy of the candidate logical rule.

[0046] Optionally, according to the set of logical rules, an inference model based on SPARQL query is used to combine the inference rules and the entity relationship in the knowledge graph to perform inference calculation to predict the future behavior pattern, health risk and social demand of the residents, specifically including:

[0047] The condition part and the conclusion part of each logical rule are converted into SPARQL query statements to facilitate matching and reasoning with entity nodes and relationship edges in the knowledge graph;

[0048] A SPARQL query template is constructed, which forms a query request by dynamically inserting the generated condition part of the logical rule, and each SPARQL query template can query the relationship between specific entities, specifically including finding entities and their attributes, behavior patterns and relationships associated with them in the knowledge graph according to the condition description in the rule;

[0049] The inference calculation based on the SPARQL query language is performed, the entities and relationships in the knowledge graph are matched through the SPARQL query template, the entity data meeting the logical rule condition is extracted, the inference result of each entity is calculated in combination with the condition of the logical rule and the relationship between the entity nodes in the knowledge graph, and the related behavior mode, health risk and social demand information are obtained.

[0050] Optionally, the long short-term memory network is used to process the time series data of the resident behavior, identify the long-term trend and short-term fluctuation, and optimize the prediction of the health risk and social demand of the resident, specifically including:

[0051] The time series data of the resident behavior is collected, the time series data including the health record, consumption behavior and social activity of the resident, and the time series data is cleaned to remove abnormal values;

[0052] The processed time series data is input into the long short-term memory network, the long short-term memory network processes the long-term dependence in the time series data through a gating mechanism, and learns and records the dynamic mode of the time series data changing over time;

[0053] According to the output of the long short-term memory network, the long-term trend analysis and short-term fluctuation detection are performed on the time series data, and through the state updating mechanism of the long short-term memory network, the long-term change trend of the resident behavior is extracted, and the short-term fluctuation abnormality or burst mode is identified;

[0054] According to the identified long-term trend and short-term fluctuation, the parameters of the long short-term memory network are updated, so that the long short-term memory network can adapt to new data mode, and the prediction result is continuously improved over time, so that the prediction of the resident behavior is more accurate and timely;

[0055] The prediction result output by the long short-term memory network is weighted in multiple dimensions, and the historical data and real-time data are compared to further optimize the prediction of the health risk and social demand of the resident, so that the change of the resident behavior under different situations can be accurately adjusted and predicted.

[0056] The optimized prediction result is used for subsequent decision support system, and the health risk assessment and social demand prediction result based on the time series data are provided, so that the decision support system can respond in time and effectively under the changing resident behavior mode

[0057] Optionally, according to the resident behavior mode, health risk and social demand inferred from the knowledge graph, a set of decision support results are generated, and data interaction and sharing are performed with external systems through an open API interface, specifically including:

[0058] According to the behavior patterns, health risks and social needs of the residents, the analysis reasoning result is analyzed to identify the key needs and potential problems of each resident, especially for the groups of residents with high health risks or strong social needs.

[0059] Based on the reasoning result, a set of decision support results are generated, including health management suggestions, resource allocation optimization schemes and personalized service recommendations. The health management suggestions include preventive measures for health risks, lifestyle adjustment suggestions and regular health checks to ensure that the suggestions can help residents reduce health risks and improve health conditions. The resource allocation optimization scheme allocates medical resources and social service resources based on the demand evaluation of different groups in the reasoning result to ensure that resources can be efficiently and accurately allocated to the residents who need them. The personalized service recommendation recommends personalized services according to the behavior patterns and social needs of each resident, including social activity recommendations, health consultation services, and improves the quality of life and happiness of residents.

[0060] The generated decision support results are interacted and shared with external systems through open API interfaces. Data docking with medical systems, social platforms and other public service systems is supported to realize data interconnection and sharing, and to provide decision-making basis for relevant departments and service providers to ensure that they can take appropriate actions based on the generated results.

[0061] Through cross-system data integration and management, data collaboration between different systems is promoted. By integrating real-time data from multiple external data sources such as medical, social and behavioral data, a comprehensive resident data view is formed to provide sufficient data support for decision makers, ensuring the scientificity and operability of the decision.

[0062] The beneficial effects of the present application are:

[0063] 1、The present application constructs a resident information intelligent management system based on knowledge graph, effectively solves the problem of fragmentation and isolated storage of resident information in the prior art. Traditional management systems often rely on manual input and maintenance, and data cannot be efficiently integrated and associated. The present application uses knowledge graph to unify modeling of multi-source, multi-dimensional resident data, and integrates through entity nodes and relationship edges, breaking the data silos, so that various types of information can realize efficient interaction and sharing, thereby improving the integration and accuracy of data.

[0064] 2、The application realizes efficient prediction of resident behavior patterns, health risks and social needs by combining a SPARQL query-based reasoning model. Traditional systems often only provide static query and statistical functions, lacking intelligent analysis and prediction capabilities. However, the application can automatically reason according to the association between entities in the knowledge graph, predict future behavior patterns and health risks of residents, and adjust the prediction results in real time to adapt to changes in resident behavior. This not only improves the accuracy of prediction, but also provides decision makers with more reliable decision-making basis.

[0065] 3、The application uses long short-term memory network (LSTM) to process time series data of resident behavior, which can accurately identify long-term trends and short-term fluctuations, and effectively optimize the prediction of health risks and social needs. Compared with traditional methods that ignore time dependence, LSTM has strong memory ability and processing advantages for time series data, so that the dynamic changes of resident behavior can be tracked and optimized in real time, greatly improving the flexibility and timeliness of prediction.

[0066] 4、The application promotes data integration between different systems through cross-system data sharing and collaborative work. Through open API interface, the system can interact and share data with medical systems, social platforms and other public service systems, realizing collaborative work of multi-party data. This not only improves the utilization efficiency of data, but also strengthens the interconnection between systems, avoids the existence of data silos, so that decision makers can make more accurate decisions based on comprehensive data. BRIEF DESCRIPTION OF DRAWINGS

[0067] The accompanying drawings are used to provide a further understanding of the application, and constitute a part of the specification, together with embodiments of the application, to explain the application, and do not constitute a limitation on the application. In the drawings:

[0068] Figure 1 A structural schematic diagram of a resident information intelligent management system based on a knowledge graph is provided for the application;

[0069] Figure 2 A whole flow chart of a resident information intelligent management method based on a knowledge graph is provided for the application. DETAILED DESCRIPTION

[0070] The application will now be further described in detail in conjunction with the drawings. These drawings are simplified schematic diagrams, and only illustrate the basic structure of the application in a schematic manner, so they only show the components related to the application.

[0071] REFERENCE Figure 1 A resident information intelligent management system based on a knowledge graph includes the following modules:

[0072] a behavior pattern inference module for generating future behavior pattern predictions of residents based on resident behavior data and social needs in the knowledge graph using a SPARQL query-based inference model, and generating behavior change trends in combination with health risk analysis;

[0073] a health risk assessment module for analyzing health data and inference results of residents, identifying high-risk groups, and generating personalized health management recommendations according to prediction results;

[0074] a resource allocation optimization module for automatically generating resource allocation optimization schemes according to demand analysis and behavior patterns of residents in combination with health risk assessment, and ensuring efficient and accurate allocation of resources according to demand;

[0075] a personalized service recommendation module for generating personalized service recommendations according to behavior patterns and health risk analysis of residents, ensuring that residents can receive health services, social activities or social resources that meet their needs;

[0076] a decision support result display module for presenting a set of decision support results according to generated health management recommendations, resource allocation optimization schemes and personalized service recommendations for decision makers to refer, and displaying the results through an interface or a visual way;

[0077] a data sharing and interaction module for data interaction and sharing with external systems through an open API interface, supporting data connection with medical systems, social platforms and other public service systems, realizing multi-party data collaboration, and promoting cross-system data integration and management.

[0078] The present application effectively integrates multi-dimensional information of residents by constructing a resident information intelligent management system based on a knowledge graph, solving the problems of data silos and fragmentation. The system can accurately predict the behavior patterns, health risks and social needs of residents through intelligent inference and health risk assessment, and provide personalized service recommendations and resource optimization allocation schemes. Through open API interface and external systems, data sharing and collaborative work are realized, improving the real-time and accuracy of decision support. The system provides an efficient and accurate solution for resident health management, resource allocation and personalized services, optimizes the allocation of social resources, and improves management efficiency.

[0079] Reference Figure 2 A resident information intelligent management method based on a knowledge graph, comprising the following steps:

[0080] In this embodiment, the modules are realized through the following methods:

[0081] By collecting multi-dimensional information from multiple information sources, and transforming the multi-dimensional information into nodes and edges in the knowledge graph according to the relevance between the multi-dimensional information, wherein the nodes represent various entities, and the edges represent the relationships between the entities;

[0082] Adopting natural language processing technology and ontology construction method, identifying and labeling various attributes and relationships of entity nodes in the knowledge graph;

[0083] Using the improved AMIE+ algorithm, a set of logical rules is generated from the constructed knowledge graph;

[0084] According to the set of logical rules, using a reasoning model based on SPARQL query, combining the reasoning rules and the entity relationships in the knowledge graph, inference calculation is carried out to predict the future behavior patterns, health risks and social needs of the residents;

[0085] Based on the long short-term memory network, the time series data of the resident behavior is processed, the long-term trend and short-term fluctuation are identified, and the prediction of the health risks and social needs of the residents is optimized;

[0086] According to the behavior patterns, health risks and social needs of the residents obtained by reasoning in the knowledge graph, a set of decision support results is generated, and data interaction and sharing with external systems are realized through an open API interface.

[0087] The present application realizes the comprehensive integration and efficient management of resident information by integrating multi-dimensional data from multiple information sources to construct a knowledge graph. Through natural language processing and ontology construction method, the standardization and labeling of entity nodes and relationships in the graph are ensured, making the data processing more accurate and consistent. Using the improved AMIE+ algorithm and SPARQL query reasoning model, the system can automatically generate an efficient rule set to accurately predict the future behavior patterns, health risks and social needs of the residents. Through the processing of time series data by LSTM, the long-term trend and short-term fluctuation are effectively identified and optimized, further improving the accuracy of health risk and social demand prediction. Finally, the system generates decision support results and realizes data sharing and external system interaction, providing scientific and real-time management basis for decision makers, promoting cross-system data collaboration and resource optimization allocation.

[0088] In the embodiment, the multi-dimensional information from multiple information sources is collected, and the multi-dimensional information is transformed into nodes and edges in the knowledge graph according to the relevance between the multi-dimensional information, wherein the nodes represent various entities, and the edges represent the relationships between the entities, specifically including:

[0089] Multi-dimensional information of residents is collected from multiple information sources, including personal information, health data, social relationships, behavior records, medical records and consumption behavior data of residents;

[0090] The collected multi-dimensional information is preprocessed, and the preprocessing includes data cleaning, missing value filling, data standardization and data format conversion, to ensure the quality and consistency of the information;

[0091] Through the RDF knowledge graph construction technology, the multi-dimensional information is mapped into entity nodes in the knowledge graph, and the entity nodes include residents, diseases, health indicators, social relationships and behavior types;

[0092] The relationships between various entity nodes are identified and extracted, including social relationships between residents, health condition correlations and behavior pattern similarities, and the relationships between various entity nodes are converted into edges in the knowledge graph, and each edge represents a specific correlation between two entity nodes.

[0093] The present application collects multi-dimensional information of residents from multiple information sources, constructs a comprehensive and dynamically updated knowledge graph, and solves the problems of traditional information silos and fragmentation. Through data cleaning, missing value filling, standardization and format conversion, the quality and consistency of the information are ensured, thereby providing reliable basic data for subsequent reasoning and analysis. Using RDF knowledge graph construction technology, the multi-dimensional information of residents, such as personal information, health data and social relationships, is converted into entity nodes and relationships in the knowledge graph, enhancing the expression ability and relevance of the data. By identifying and extracting the relationships between entity nodes, the system can accurately describe the health status, social network and behavior pattern similarity of residents, providing more accurate support for intelligent reasoning and personalized service recommendation.

[0094] In the embodiment, the natural language processing technology and ontology construction method are used to identify and label various attributes and relationships of entity nodes in the knowledge graph, specifically including:

[0095] The text data in the resident information is processed using the word segmentation, part-of-speech tagging and syntax analysis in the natural language processing technology, the entity attributes and relationships in the text data are identified, and the corresponding relationship between the entity and the attribute is constructed;

[0096] Using the ontology construction method, various concepts, attributes and relationships in the resident information are defined, the concepts include diseases, residents and health, the attributes include disease types and age groups, and the relationships include kinship, a unified semantic standard is established, so that the entities and relationships in the knowledge graph can conform to the predefined ontology model;

[0097] Each entity node in the knowledge graph is labeled, including assigning a corresponding category label to each entity node, the category label includes residents, diseases and hospitals, and the attributes of each entity node are semantically labeled, the attributes include birth date and disease time, to ensure that the information is clear and conforms to the standardized semantics;

[0098] The relationship extraction technology is used to extract and label the relationship between entities in the resident information, including social relationship, health condition association and behavior pattern similarity, the relationship between entities is labeled as an edge in the knowledge graph, and the semantic of each relationship is clear, and the association between entities in the graph can be accurately reflected;

[0099] The labeled knowledge graph is verified and quality checked to ensure that the labeling of all entity nodes and edges conforms to the predefined ontology model and semantic standard, and the information in the knowledge graph is consistent, complete, and supports further data processing, reasoning and analysis.

[0100] The natural language processing technology and ontology construction method are used to effectively improve the labeling accuracy of entity nodes and the standardization of information in the knowledge graph. Through word segmentation, part-of-speech tagging and syntax analysis, the text data in the resident information is processed, the clear relationship between entities and attributes is identified and constructed, and the consistency and clarity of the information are ensured. By using the ontology construction method, the system defines a unified concept, attribute and relationship, so that the entities and relationships in the knowledge graph conform to the predefined ontology model, thereby realizing semantic standardization, enhancing the scalability and adaptability of the graph. In addition, through the relationship extraction technology, the system can accurately identify the relationship between entities in the resident information, such as social relationship, health condition association, etc., and ensure that the semantics of these relationships are clear and can accurately reflect the actual situation. Finally, the labeled knowledge graph is verified and quality checked to ensure the integrity and consistency of the information in the graph, supporting subsequent data processing, reasoning and analysis, and improving the intelligent decision-making ability and data reliability of the system.

[0101] In this embodiment, the improved AMIE+ algorithm specifically includes:

[0102] The improved AMIE+ algorithm uses a multi-layer nonlinear transformation function to represent the entity nodes and relationship edges in the knowledge graph in a high-dimensional embedding, and converts each entity node and relationship edge into a point in a low-dimensional vector space through nonlinear mapping to generate a nonlinear embedding vector:

[0103] e i =σ3(W3·σ2(W2·σ1(W1·x i +b1)b2)b3)

[0104] Wherein, e i is a low-dimensional embedding vector after multi-layer nonlinear transformation, representing the final embedding of each entity node or relationship edge in the knowledge graph, x iis the input vector of the ith entity or relationship edge, W1, W2, W3 are the weight matrices of each layer, which control the mapping from the input vector to the embedding vector, b1, b2, b3 are the bias terms, each layer is used to adjust the output, and σ1, σ2, σ3 are Sigmoid nonlinear activation functions, which are used to capture the complex nonlinear relationship in the data. So that each relationship in the knowledge graph not only retains the direct relationship between entities, but also captures the implicit and complex nonlinear dependence relationship. Through this method, the semantic expression ability between entity nodes is enhanced, especially for complex multi-level relationships, providing deeper structural information. This nonlinear embedding result provides a basis for subsequent implicit association reasoning and rule generation.

[0105] Based on the nonlinear embedding vector, the improved AMIE+ algorithm uses an implicit association reasoning method to infer the potential complex association between entity nodes, identify and extract potential deep association relationships from the knowledge graph, and generate a preliminary candidate logic rule set by evaluating the similarity and co-occurrence pattern between different entity pairs in the high-dimensional embedding space. Such as residents with similar health conditions may have similar lifestyles, behavior patterns and potential associations with health conditions, etc. This method further explores the potential associations that have not yet emerged through similarity measurement and mutual influence calculation in the embedding space, and the generated candidate logic rule set has high potential in this step and provides a basis for causal reasoning.

[0106] The improved AMIE+ algorithm introduces a causal graph reconstruction method, which analyzes the entities and relationships in the knowledge graph and generates causal effect weights;

[0107] Combining the nonlinear embedding vector, the candidate logic rule set and the causal effect weight, the improved AMIE+ algorithm prioritizes the candidate logic rule set, selects the candidate logic rule with clear causal relationship and accurate behavior pattern prediction effect according to the causal effect weight of the candidate logic rule, and simultaneously performs weighted optimization of the candidate logic rule using the causal effect weight;

[0108] By calculating the reasoning accuracy of each candidate logic rule under different situations, the candidate logic rule with high causal explanatory power and reasoning accuracy is selected, and the final logic rule set is generated. The final logic rule set not only can efficiently predict the behavior patterns, health risks and social needs of residents, but also can adapt to the dynamic changes and diversified needs of the knowledge graph data.

[0109] The improved AMIE+ algorithm introduces a multi-layer nonlinear transformation function, generates high-dimensional embedding representations for entity nodes and relationship edges in the knowledge graph. This method not only preserves the direct relationship between entities, but also captures complex nonlinear dependency relationships, thereby enhancing the semantic expression ability of entity nodes in the graph, especially for complex multi-level relationship processing. This nonlinear embedding result provides a strong foundation for implicit association reasoning and rule generation, enabling the system to identify potential deep-level association relationships and effectively mine potential patterns in the graph that have not yet emerged.

[0110] By combining the implicit association reasoning method, the application can evaluate the similarity and co-occurrence patterns between different entities in the high-dimensional embedding space, generate a preliminary rule candidate set, and provide high-potential data support for causal reasoning. The improved AMIE+ algorithm introduces a causal graph reconstruction method to perform causal analysis on entities and relationships in the knowledge graph, generates causal effect weights, and selects rules with clear causal relationships and high prediction accuracy based on the reasoning effect of the rules. This process optimizes the generation and reasoning ability of the rules, enabling the final rule set to accurately predict resident behavior patterns, health risks, and social needs, and to adapt to dynamic changes and diverse needs of knowledge graph data, improving the flexibility and scalability of the system in practical applications.

[0111] In the embodiment, the introduced causal graph reconstruction method performs causal analysis on entities and relationships in the knowledge graph and generates causal effect weights, specifically including:

[0112] Structural representation of entity nodes and relationship edges in the knowledge graph, abstracting relationships in the knowledge graph as causal relationship chains;

[0113] Based on the causal relationship chain, the improved AMIE+ algorithm constructs a causal relationship graph based on the semantic and structural information of each entity and relationship in the knowledge graph through dynamic causal analysis, wherein the nodes in the causal relationship graph represent entities and the edges represent causal relationships between entities;

[0114] Based on the causal relationship graph, the improved AMIE+ algorithm evaluates the changes of each candidate logical rule under different hypothetical conditions by setting different hypothetical conditions, generates counterfactual scenarios, simulates the changes of candidate logical rules under different counterfactual scenarios, and analyzes the potential impact of each candidate logical rule on entity behavior and relationships under hypothetical conditions;

[0115] The improved AMIE+ algorithm quantifies the causal inference effect of each candidate logical rule, calculates the influence strength of each candidate logical rule on behavior prediction by measuring the causal dependence relationship between entity nodes in the causal relationship graph, generates a causal effect weight, and the causal effect weight reflects the influence degree of the candidate logical rule on the behavior of residents, health risks and social needs in different situations.

[0116] The selected candidate logical rule is bound to the causal relationship between the actual behavior and entity in the knowledge graph, ensuring that each candidate logical rule can not only capture the correlation between entities in the graph, but also reveal the potential causal chain and causal effect, thereby improving the explainability and reasoning accuracy of the candidate logical rule.

[0117] The present application effectively improves the causal explanation ability of the reasoning result in the knowledge graph by introducing a causal graph reconstruction method. By abstracting the relationships in the graph into causal relationship chains, the improved AMIE+ algorithm can construct a causal relationship graph based on the semantic and structural information in the graph, and reveal the deep causal relationship between entities. Through dynamic causal analysis and simulation of counterfactual scenarios, the system can evaluate the changes of candidate rules under different hypothetical conditions, further quantify the causal effect of each candidate rule, and generate a causal effect weight reflecting the influence of the rule on the behavior of residents, health risks and social needs. Finally, the selected candidate rule is bound to the causal relationship between the actual behavior and entity in the graph, ensuring that the generated logical rule can not only capture the correlation between entities, but also reveal the potential causal chain. This process improves the explainability and reasoning accuracy of the rule, significantly enhancing the adaptability and prediction ability of the system in a dynamic environment.

[0118] In the embodiment, the inference model based on SPARQL query is used according to the logical rule set, the entity relationship in the knowledge graph is combined with the inference rule, and inference calculation is performed to predict the future behavior pattern, health risk and social demand of residents, specifically including:

[0119] The condition part and the conclusion part of each logical rule are converted into SPARQL query statements to facilitate matching and reasoning with entity nodes and relationship edges in the knowledge graph;

[0120] A SPARQL query template is constructed, the SPARQL query template is formed by dynamically inserting the condition part of the generated logical rule to form a query request, and each SPARQL query template can query the relationship between specific entities, specifically including finding entities associated with the condition description in the rule, their attributes, behavior patterns and relationships in the knowledge graph;

[0121] The inference calculation based on the SPARQL query language is performed, the entities and relations in the knowledge graph are matched through the SPARQL query template, the entity data meeting the logical rule conditions is extracted, the inference result of each entity is calculated combined with the conditions of the logical rules and the relations between the entity nodes in the knowledge graph, and the related behavior mode, health risk and social demand information is obtained.

[0122] By adopting the inference model based on the SPARQL query, the entity nodes and relation edges in the knowledge graph can be matched efficiently, and intelligent inference calculation is realized. By converting the condition part and the conclusion part of each logical rule into a SPARQL query statement and constructing a dynamic SPARQL query template, the system can accurately query the relations between specific entities according to the generated rule conditions and extract related entity data. This method makes the inference calculation process more flexible and accurate, and can effectively predict the future behavior mode, health risk and social demand of residents, thereby providing real-time and effective decision support for decision makers. Through the efficient calculation of the SPARQL query language, the system can quickly extract the behavior mode and health risk information meeting the conditions in large-scale data, improving the overall prediction accuracy and response speed.

[0123] In the embodiment, the long short-term memory network is used to process the time series data of the behavior of the residents, identify the long-term trend and short-term fluctuation, and optimize the prediction of the health risk and social demand of the residents, specifically including:

[0124] Collecting the time series data of the behavior of the residents, the time series data including the health records, consumption behavior and social activities of the residents, and cleaning the time series data to remove abnormal values;

[0125] Inputting the processed time series data into the long short-term memory network, the long short-term memory network processing the long-term dependence in the time series data through a gating mechanism, learning and recording the dynamic mode of the time series data changing over time;

[0126] According to the output of the long short-term memory network, the long-term trend analysis and short-term fluctuation detection are performed on the time series data, and through the state updating mechanism of the long short-term memory network, the long-term change trend of the behavior of the residents is extracted, and the abnormal or sudden mode of the short-term fluctuation is identified;

[0127] According to the identified long-term trend and short-term fluctuation, the parameters of the long short-term memory network are updated, so that the long short-term memory network can adapt to new data patterns, and the prediction result is continuously improved over time, thereby ensuring that the prediction of the behavior of the residents is more accurate and timely;

[0128] By weighting the prediction results output by the long short-term memory network in multiple dimensions, combining historical data and real-time data comparison, further optimizing the prediction of residents' health risks and social needs, ensuring that the behavior changes of residents in different situations can be accurately adjusted and predicted.

[0129] The optimized prediction results are used in the subsequent decision support system to provide health risk assessment and social demand prediction results based on time series data, ensuring that the decision support system can respond effectively and timely to changing resident behavior patterns

[0130] The present application can effectively identify long-term trends and short-term fluctuations in resident behavior by processing time series data of resident behavior based on long short-term memory network (LSTM). LSTM can handle long-term dependencies in data through its gating mechanism, accurately learn and record dynamic patterns of time series data over time, ensuring a deep understanding of changes in resident behavior. By analyzing long-term trends and detecting short-term fluctuations in output results, the system can identify abnormalities or sudden changes in behavior patterns in a timely manner. Further updating the parameters of the LSTM network enables the system to continuously optimize prediction results based on new data patterns, improving the accuracy and adaptability of predictions. In addition, by weighting the prediction results in multiple dimensions and combining historical and real-time data, the system can make accurate adjustments and optimizations in different situations, improving the accuracy of health risk and social demand predictions. Ultimately, these optimized prediction results are applied to the decision support system to ensure that decision-makers can respond accurately and in a timely manner when resident behavior patterns change.

[0131] In this embodiment, a set of decision support results is generated according to the resident behavior patterns, health risks and social needs inferred from the knowledge graph, and data interaction and sharing with external systems are performed through an open API interface, specifically including:

[0132] According to the resident behavior patterns, health risks and social needs, analyze the inference results, identify the key needs and potential problems of each resident, and pay special attention to the resident groups with high health risks or strong social needs;

[0133] Based on the inference result, a set of decision support results are generated, including health management suggestions, resource allocation optimization schemes and personalized service recommendations, the health management suggestions including preventive measures for residents' health risks, lifestyle adjustment suggestions and regular health checks, ensuring that the suggestions can help residents reduce health risks and improve health conditions. The resource allocation optimization scheme, based on the demand assessment of different groups in the inference result, allocates medical resources and social service resources, ensuring that resources can be efficiently and accurately allocated to the residents in need. The personalized service recommendation, according to the behavior pattern and social needs of each resident, recommends personalized services, including social activity recommendation, health consultation service, to improve the life quality and happiness of residents;

[0134] The generated decision support results are interacted and shared with external systems through open API interface. Data docking with medical systems, social platforms and other public service systems is supported to realize data interconnection and sharing, providing decision basis for relevant departments and service providers to ensure that they can take appropriate actions based on the generated results;

[0135] Through cross-system data integration and management, data collaboration between different systems is promoted. By integrating real-time data from multiple external data sources such as medical, social, behavior, etc., a comprehensive resident data view is formed to provide sufficient data support for decision makers, ensuring the scientificity and operability of the decision.

[0136] The present application generates precise decision support results based on the behavior patterns, health risks and social needs of residents obtained through knowledge graph reasoning, providing scientific management basis for decision makers. The system analyzes the inference results to identify the key needs and potential problems of each resident, especially focusing on the residents with high health risks or strong social needs, to ensure that resources can be accurately allocated. The generated health management suggestions help residents reduce health risks and improve health conditions; the resource allocation optimization scheme ensures that medical and social resources can be efficiently allocated; the personalized service recommendation provides residents with social activities and health consultation services that meet their needs, improving their life quality and happiness. Through open API interface, the system interacts and shares data with external systems, promoting cross-system data integration and collaboration, enabling decision support to be based on comprehensive and real-time data, ensuring the accuracy and effectiveness of the decision. This innovative solution significantly improves the intelligence and accuracy of resident information management.

[0137] Embodiment 1:

[0138] To verify the feasibility of the application in practice, the application is applied to the intelligent health management system of a certain city. The health data, social needs and behavior patterns of residents need to be integrated and analyzed through multiple data sources to provide accurate health management recommendations and resource allocation optimization schemes for decision makers. This embodiment shows how to solve the problem of resident data island in traditional information management systems through a resident information intelligent management system based on a knowledge graph, optimize resource allocation and provide personalized service recommendations.

[0139] In this scenario, the health data, social behavior and other life information of residents are collected and integrated into a knowledge graph. These data sources come from electronic health records (EHR) in hospitals, social platforms, consumer behavior analysis and daily activity records in communities. Through the collection and processing of multi-dimensional data, the system converts these data into a structured knowledge graph, in which each resident, disease, social relationship, etc. is mapped as a node, and the association between nodes is represented as an edge. For example, the health data node and the disease node, the social relationship node and the resident node establish various relationships, forming a comprehensive and multi-level resident information graph.

[0140] After these data enter the system, the system labels and semantically standardizes each entity node in the knowledge graph through natural language processing technology and ontology construction methods. This process ensures consistency between different data sources and avoids information ambiguity or loss due to inconsistent data formats. Next, using the improved AMIE+ algorithm, the system automatically generates a set of logical rules to predict residents' future behavior patterns, health risks and social needs through reasoning calculations.

[0141] A resident, Mr. Zhang, has health data indicating a high family history of diabetes and a high body mass index. Based on the reasoning results, the system predicts that Mr. Zhang may face a high risk of developing diabetes in the next six months. This reasoning result is combined with Mr. Zhang's social needs and behavior patterns to predict that he may have a behavior pattern of reducing social activities and reducing exercise. Based on these reasoning results, the system generates health management recommendations, recommending that Mr. Zhang undergo regular health checks, increase exercise, and recommend social activities to increase interaction with others to prevent social isolation.

[0142] At the same time, based on the resource allocation optimization module, the system generates a corresponding resource allocation scheme, recommending that more health management resources be allocated to high-risk groups in the area, including professional physician follow-up and health education activities. Through automated recommendations based on the reasoning model, the system ensures that medical resources can be accurately allocated, avoiding waste or uneven distribution of resources.

[0143] Moreover, in the personalized service recommendation module, the system generates personalized social activity recommendations based on Zhang's social needs, including local diabetes support group activities and fitness courses suitable for middle-aged people, ensuring that Zhang can receive precise services according to his personal needs. This not only improves Zhang's quality of life but also enhances his social interaction and reduces potential feelings of loneliness.

[0144] To verify the effectiveness of the system, the data of residents in a certain city were used as test samples in the embodiment. There are about 5000 residents in the city participating in the health management project. Through the reasoning calculation and personalized recommendation of the system, the system successfully predicts the high health risks of 600 residents and provides personalized health management suggestions and social activity recommendations based on the reasoning results.

[0145] Table 1: Comparison of health risk prediction and social needs satisfaction before and after the system

[0146]

[0147]

[0148] Table 1 is a comparison table of data before and after the system reasoning, showing the accuracy of the system's health risk prediction and the improvement of social needs satisfaction before and after the system implementation. As can be seen from Table 1, after the implementation of the system, the accuracy of the health risk prediction of the residents has increased from 62% to 89%, showing a significant improvement in the reasoning model and health risk assessment method. The identification success rate of the high-risk group has also significantly improved, from 60% to 85%. In addition, the participation rate of social activities, especially in the high-risk group, has jumped from 40% to 75%, which proves the effectiveness of the personalized service recommendation module, which helps high-risk residents find suitable social activities and improves their participation.

[0149] The acceptance rate of health management suggestions has also significantly increased from 55% to 92%, indicating that the personalized health management suggestions provided by the system are more targeted and practical, and the residents' acceptance of these suggestions has greatly improved. Finally, the accuracy of resource allocation optimization has also increased from 58% to 90%, which shows the efficiency of the resource optimization allocation scheme based on the reasoning results, ensuring that medical and social resources can be accurately allocated to the most needed resident groups.

[0150] In summary, through the implementation of the system, the health risk prediction of residents, the satisfaction of social needs, and the accuracy of resource allocation have been significantly improved, proving the effectiveness and feasibility of the intelligent management system based on knowledge graph in practical application.

[0151] The above merely describes preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art, according to the technical solution and inventive concept of the present application, makes equivalent replacement or change within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A knowledge graph-based resident information intelligent management system, characterized in that, The method comprises the following modules: a behavior pattern inference module for generating a prediction of future behavior patterns of residents based on resident behavior data and social needs in a knowledge graph, using a SPARQL query-based inference model, and generating a trend of behavior changes in combination with health risk analysis; a health risk assessment module for analyzing health data and inference results of residents, identifying high-risk groups, and generating personalized health management recommendations according to prediction results; a resource allocation optimization module for automatically generating resource allocation optimization schemes according to demand analysis and behavior patterns of residents in combination with health risk assessment; a personalized service recommendation module for generating personalized service recommendations according to behavior patterns and health risk analysis of residents; a decision support result display module for presenting a set of decision support results according to generated health management recommendations, resource allocation optimization schemes, and personalized service recommendations, for reference by decision makers, and displaying the results through an interface or a visual method; a data sharing and interaction module for data interaction and sharing with external systems through an open API interface. 2.The intelligent management system of resident information based on knowledge graph according to claim 1, characterized in that, The modules are implemented through the following methods: by collecting multi-dimensional information from multiple information sources and converting the multi-dimensional information into nodes and edges in a knowledge graph according to the relevance between the multi-dimensional information, wherein the nodes represent various entities and the edges represent the relationships between the entities; using natural language processing technology and ontology construction methods to identify and label various attributes and relationships of entity nodes in the knowledge graph; using an improved AMIE+ algorithm to generate a set of logical rules from the constructed knowledge graph; using a SPARQL query-based inference model to perform inference calculation based on the set of logical rules and the entity relationships in the knowledge graph, to predict future behavior patterns, health risks, and social needs of residents; processing time series data of resident behavior based on a long short-term memory network to identify long-term trends and short-term fluctuations, and optimizing the prediction of health risks and social needs of residents; generating a set of decision support results based on the inferred behavior patterns, health risks, and social needs of residents in the knowledge graph, and performing data interaction and sharing with external systems through an open API interface. 3.The knowledge graph-based intelligent management system for resident information according to claim 2, characterized in that, The method of collecting multi-dimensional information from multiple information sources and converting the multi-dimensional information into nodes and edges in a knowledge graph according to the relevance between the multi-dimensional information, wherein the nodes represent various entities and the edges represent the relationships between the entities, specifically comprises: collecting multi-dimensional information of residents from multiple information sources, the multi-dimensional information including personal information, health data, social relationships, behavior records, medical records, and consumption behavior data of residents; preprocessing the collected multi-dimensional information, the preprocessing including data cleaning, missing value filling, data standardization, and data format conversion; mapping the multi-dimensional information into entity nodes in a knowledge graph through RDF knowledge graph construction technology, the entity nodes including residents, diseases, health indicators, social relationships, and behavior types; Identify and extract the relationship between various types of entity nodes, including social relationships between residents, health status association and behavior pattern similarity, and convert the relationship between various types of entity nodes into edges in the knowledge graph, each edge representing the specific association between two entity nodes. 4.The intelligent management system of resident information based on knowledge graph according to claim 2, characterized in that, The natural language processing technology and ontology construction method are used to identify and label various attributes and relationships of entity nodes in the knowledge graph, specifically including: Using the word segmentation, part-of-speech tagging and syntax analysis in natural language processing technology, the text data in the resident information is processed to identify the entity attributes and relationships in the text data, and the corresponding relationship between the entity and the attribute is constructed; Using the ontology construction method, define various concepts, attributes and relationships in the resident information, the concepts include disease, resident and health, the attributes include disease type and age range, and the relationships include kinship, and establish a unified semantic standard; Each entity node in the knowledge graph is labeled, including assigning a corresponding category label to each entity node, the category label includes resident, disease and hospital, and performing semantic labeling on the attributes of each entity node, the attributes include birth date and disease time; Using the relationship extraction technology, the relationship between entities is extracted and labeled from the resident information, including social relationship, health status association and behavior pattern similarity, and the relationship between entities is labeled as an edge in the knowledge graph. 5.The intelligent management system of resident information based on knowledge graph according to claim 2, characterized in that, The improved AMIE+ algorithm, specifically including: The improved AMIE+ algorithm uses a multi-layer nonlinear transformation function to represent the entity nodes and relationship edges in the knowledge graph in a high-dimensional embedding, and converts each entity node and relationship edge into a point in a low-dimensional vector space through nonlinear mapping to generate a nonlinear embedding vector; Based on the nonlinear embedding vector, the improved AMIE+ algorithm uses an implicit association reasoning method to reason about the potential complex associations between entity nodes, identify and extract potential deep association relationships from the knowledge graph, and generate a preliminary candidate logic rule set by evaluating the similarity and co-occurrence patterns between different entity pairs in the high-dimensional embedding space; The improved AMIE+ algorithm introduces a causal graph reconstruction method, which analyzes the entities and relationships in the knowledge graph and generates causal effect weights; Combining the nonlinear embedding vector, the candidate logic rule set and the causal effect weight, the improved AMIE+ algorithm prioritizes the candidate logic rule set, selects the candidate logic rule with clear causal relationship and accurate behavior pattern prediction effect according to the causal effect weight of the candidate logic rule, and performs weighted optimization of the candidate logic rule using the causal effect weight; By calculating the reasoning accuracy of each candidate logic rule in different situations, the candidate logic rule with high causal explanation and reasoning accuracy is selected, and the final logic rule set is generated. 6.The knowledge graph-based intelligent management system for resident information according to claim 5, characterized in that, The introduction of the causal graph reconstruction method, which analyzes the entities and relationships in the knowledge graph and generates causal effect weights, specifically includes: Structurally represent the entity nodes and relationship edges in the knowledge graph, and abstract the relationships in the knowledge graph as causal relationship chains; Based on the causal chain, the improved AMIE+ algorithm constructs a causal relationship graph according to the semantic and structural information of each entity and relationship in the knowledge graph, wherein the nodes in the causal relationship graph represent entities, and the edges represent the causal relationship between entities; Based on the causal relationship graph, the improved AMIE+ algorithm sets different hypothetical conditions, evaluates the changes of each candidate logical rule in these situations, generates counterfactual scenarios, simulates the changes of candidate logical rules in different counterfactual scenarios, and analyzes the potential impact of each candidate logical rule on entity behavior and relationship under the hypothetical situation; The improved AMIE+ algorithm quantifies the causal reasoning effect of each candidate logical rule, calculates the influence strength of each candidate logical rule on behavior prediction by measuring the causal dependence relationship between entity nodes in the causal relationship graph, and generates causal effect weights, which reflect the influence degree of candidate logical rules on resident behavior, health risk and social demand under different situations. 7.The intelligent management system of resident information based on knowledge graph according to claim 2, characterized in that, According to the set of logical rules, the inference model based on SPARQL query is used to combine the inference rules and the entity relationships in the knowledge graph to perform inference calculation and predict the future behavior patterns, health risks and social demands of residents, specifically including: Convert the condition part and conclusion part of each logical rule into a SPARQL query statement; Build a SPARQL query template, which forms a query request by dynamically inserting the generated logical rule condition part; Perform inference calculation based on the SPARQL query language, match the SPARQL query template with the entities and relationships in the knowledge graph, extract entity data that meets the logical rule conditions, combine the conditions of the logical rules with the relationships between entity nodes in the knowledge graph, calculate the inference results of each entity, and obtain related behavior patterns, health risks and social demand information. 8.The intelligent management system of resident information based on knowledge graph according to claim 2, characterized in that, The long short-term memory network is used to process the time series data of resident behavior, identify long-term trends and short-term fluctuations, and optimize the prediction of resident health risks and social demands, specifically including: Collect time series data of resident behavior, including resident health records, consumption behavior and social activities, and clean the time series data to remove outliers; Input the processed time series data into the long short-term memory network, which processes the long-term dependencies in the time series data through a gating mechanism, learns and records the dynamic patterns of time series data over time; According to the output of the long short-term memory network, analyze the long-term trend and short-term fluctuation of the time series data, and extract the long-term change trend of resident behavior and identify the abnormal or sudden pattern of short-term fluctuation through the state update mechanism of the long short-term memory network; Update the parameters of the long short-term memory network according to the identified long-term trend and short-term fluctuation; Further optimize the prediction of resident health risks and social demands by weighting the prediction results output by the long short-term memory network in multiple dimensions and comparing historical data and real-time data. 9.The intelligent management system of resident information based on knowledge graph according to claim 2, characterized in that, According to the resident behavior pattern, health risk and social demand obtained by reasoning in the knowledge graph, a set of decision support results are generated, and data interaction and sharing are carried out with external systems through an open API interface, specifically including: According to the resident behavior pattern, health risk and social demand, the reasoning results are analyzed to identify the key needs and potential problems of each resident; Based on the reasoning results, a set of decision support results are generated, including health management suggestions, resource allocation optimization schemes and personalized service recommendations, the health management suggestions include preventive measures for resident health risks, lifestyle adjustment suggestions and regular health checks, the resource allocation optimization scheme allocates medical resources and social service resources based on the demand assessment of different groups in the reasoning results, and the personalized service recommendation recommends personalized services according to the behavior pattern and social demand of each resident, including social activity recommendation, health consultation service; The generated decision support results are interacted and shared with external systems through an open API interface.

Images