A hospital grid risk analysis method and system based on a knowledge graph

Abstract

The application discloses a hospital grid risk analysis method and system based on a knowledge graph, which comprises the following steps: collecting multi-source data such as medical equipment, spatial areas and personnel roles, extracting core entities by using a BERT model and constructing correlation, generating a hospital safety knowledge graph; mining potential risk correlation paths among entities by using a Dijkstra shortest path algorithm; extracting multi-dimensional risk features to construct a feature vector and inputting the feature vector into a logistic regression model to classify risk levels; generating a targeted early warning scheme according to the risk level and pushing the early warning scheme to responsible personnel; through the deep integration of the knowledge graph, the graph algorithm and the machine learning model, the application realizes the hidden correlation mining and accurate grading of multi-dimensional risk factors in the hospital, solves the problems of incomplete risk identification, low analysis accuracy and poor response efficiency in the prior art, and significantly improves the intelligent level of the hospital grid safety management.

Description

Technical Field

[0001] This invention relates to the field of hospital safety management technology, and in particular to a hospital grid risk analysis method and system based on knowledge graphs. Background Technology

[0002] As public service venues with high population density, numerous medical equipment, complex functional areas, and 24-hour continuous operation, hospitals' safety management level directly affects patients' lives and health, medical staff's occupational safety, and the normal operational order of medical institutions. In recent years, with the continuous expansion of hospital scale and the advancement of intelligent construction, the safety risks faced by hospitals have become increasingly diversified and concealed, specifically manifested in the following aspects: (1) Risk factors are multi-dimensional and intertwined; hospital safety involves multiple dimensions such as the operating status of medical equipment, storage of special drugs, functional attributes of spatial areas, personnel operating procedures, and implementation of systems. There are complex hidden connections between these dimensions. For example, minor malfunctions of medical equipment may have a coupling effect with the environmental temperature and humidity of a specific area, the qualification level of operators, and the implementation of relevant systems, eventually evolving into serious safety accidents. Existing technical means are difficult to effectively explore and express this multi-dimensional and cross-entity risk transmission relationship. (2) Traditional risk analysis methods have inherent limitations. Currently, hospital safety risk analysis mainly relies on two models: one is periodic inspection based on human experience, which is greatly affected by subjective factors and is difficult to achieve objective quantification of risks; the other is data monitoring based on single indicators (such as equipment operating status and environmental monitoring data), which only focuses on isolated indicators and ignores the impact of the synergistic effect of multiple factors on risk evolution, resulting in a high misjudgment rate of risk level and a lack of targeted early warning schemes. The analysis efficiency is difficult to meet the needs of dynamic management. The hospital's grid management requires differentiated and dynamic risk monitoring in different areas, but the traditional manual analysis model is time-consuming and slow to respond, making it difficult to meet the real needs of rapid changes in hospital risk factors. At the same time, historical risk event data is mostly stored in different business systems, failing to form effective integration and in-depth utilization, resulting in a waste of data resources. Summary of the Invention

[0003] In view of this, the present invention provides a hospital grid risk analysis method and system based on knowledge graphs, which solves the problems of insufficient association identification, low analysis accuracy and poor response efficiency in existing hospital risk analysis. It integrates multi-source data, mines hidden risk associations, and realizes accurate risk classification and rapid early warning in hospital grid risk analysis, thereby improving the intelligent level of hospital safety management.

[0004] To achieve the above objectives, this invention provides a hospital grid risk analysis method based on knowledge graphs, comprising the following steps: S1. Collect multi-source data related to hospital safety management, including medical equipment data, spatial area data, personnel role data, system and standard data, and risk event data; S2. Construct a knowledge graph by using the named entity recognition method in natural language processing to extract core entities from the collected multi-source data, build the relationships between entities, store the entity and relationship data in a graph database, and generate a hospital safety knowledge graph. S3. Use Dijkstra's shortest path method to mine potential risk association paths between entities in the hospital safety knowledge graph; S4. Extract multi-dimensional risk features to construct feature vectors, input them into the trained risk level classification model, and obtain the risk level of the hospital grid area; S401. Extract multi-dimensional risk features and construct the input feature vector. ,in Indicates the feature dimension; S402, Introducing the Category Weight Vector ,in, Corresponding to low risk, Corresponding to medium risk, Corresponding to high risk, For significant risks, the risk level model outputs the probabilities for each category as follows: in, Indicates risk level label, This represents the transpose of the weight vector; S403. Train the risk level model using historical risk data, and optimize the model parameters by maximizing the log-likelihood function, as expressed in: in, Indicates the number of training samples. The indicator function is used; the model parameters are optimized using gradient descent, with a learning rate of 0.01 and 500 training iterations until the model converges. S404. Input the feature vector of the grid area extracted in real time into the trained model, and select the risk level with the highest probability as the final risk level of the grid area. S5. Generate a targeted early warning plan based on the risk level and push the early warning information to the relevant responsible personnel.

[0005] Preferably, a named entity recognition algorithm based on a BERT pre-trained model is used for entity extraction, and the fine-tuning learning rate of the BERT pre-trained model is set to... The training iterations were set to 100 rounds, the batch size to 32, a function optimization model was used, and the convergence threshold was set to... .

[0006] Preferably, the relationships between entities include "medical equipment - location", "personnel - responsible grid", "risk event - associated equipment" and "system / standard - applicable area"; The Neo4j graph database is used to store entity and relationship data, and unique indexes are created for the entity ID, entity type, and relationship type fields.

[0007] Preferably, the Dijkstra shortest path method for mining potential risk association paths between entities in the hospital safety knowledge graph specifically includes: S301. Assigning each edge in the knowledge graph an attribute based on the degree of influence of the association on risk transmission. Weight values ​​within the range; S302, Taking the target risk entity as the starting point The grid area entities are the target nodes. The shortest path from the starting node to each target node is calculated by updating the distance, and the expression is: in, Represents a node To the node edge weights, Indicates the starting node To the node The current shortest distance, initially, Other nodes Iterate and update until the distances between all nodes are stable; S303. Select the risk-related paths whose shortest path length is less than a preset threshold as potential risk-related results.

[0008] Preferably, the early warning scheme generation step specifically includes: S501. Classify warning levels: Set the corresponding risk levels as four warning levels, with low risk corresponding to blue warning, medium risk corresponding to yellow warning, high risk corresponding to orange warning, and major risk corresponding to red warning. S502. Generate early warning information: Generate early warning information based on the core entities in the risk association path, and clarify the source of risk, scope of impact and related personnel; S503. Pushing early warning information: Pushing early warning information to the relevant responsible personnel through SMS, mobile APP push, and smart display screen sound and light alarm. The push priority follows the principle of "major risk > high risk > medium risk > low risk".

[0009] This invention also provides a knowledge graph-based hospital grid risk analysis system, comprising: The data acquisition module is used to collect various types of data related to hospital safety management and provides multi-source data access interfaces; Knowledge graph construction module: used for entity extraction, relation construction and knowledge storage, generating a hospital safety knowledge graph; Risk Association Mining Module: Employs Dijkstra's algorithm to mine potential risk association paths between entities; Risk level classification module: Based on a logistic regression model, it enables accurate judgment of the risk level of grid areas; Early warning plan generation module: Generates targeted early warning plans based on risk levels and pushes early warning information; Data storage module: Neo4j graph database is used to store knowledge graph data, and MySQL database is used to store risk analysis results and early warning records.

[0010] Compared with the prior art, the beneficial effects of the present invention are: This invention constructs a hospital safety knowledge graph that integrates multiple entities such as medical equipment, spatial areas, personnel roles, institutional standards, and risk events. It also introduces Dijkstra's shortest path algorithm to mine implicit risk transmission paths between entities. This breaks through the limitations of traditional methods that can only analyze single dimensions or explicit relationships. The method provided by this invention can automatically discover complex risk chains that cross entities and levels, such as "equipment failure - regional environment - operator - patient safety". It effectively identifies potential risks that are difficult to discover through traditional manual investigation, and significantly improves the comprehensiveness and foresight of risk identification. This invention employs a logistic regression classification model that integrates multi-dimensional features. The model's input features not only include topological information of risk-related paths (such as the number of paths and the average edge weights), but also integrate real-time status information such as equipment operating parameters and environmental monitoring data to construct a multi-dimensional comprehensive feature vector. Through deep learning training on historical risk data, the model can more accurately assess the true risk level under the coupling effect of multiple factors, thereby generating differentiated and precise early warning schemes for different risk scenarios and effectively reducing false alarms and false negatives. This invention constructs a fully automated closed loop from data collection, knowledge graph construction, association mining to risk classification and early warning push. It can access dynamic data from various hospital information systems in real time and automatically complete risk analysis and level determination. It completely changes the lagging mode of relying on manual periodic investigation, and can quickly respond to the dynamic changes in the security situation within the hospital, providing real-time decision support for the hospital. Attached Figure Description

[0011] Figure 1 is a flowchart of the hospital grid risk analysis method based on knowledge graph of the present invention; Figure 2 This is a schematic diagram of the entities and relationships in the hospital safety knowledge graph of this invention; Figure 3 This is a system module architecture diagram of the present invention; Figure 4 This is an interface for the division of secondary and tertiary grids and risk point information in an embodiment of the present invention; Figure 5 This is the interface for details of preventive measures in an embodiment of the present invention; Figure 6 This is the inspection details interface according to an embodiment of the present invention; Figure 7 This is the interface showing the details of the early warning classification and response measures in an embodiment of the present invention; Figure 8 This is the emergency response plan details interface according to an embodiment of the present invention. Detailed Implementation

[0012] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0013] Example 1

[0014] This embodiment provides a hospital grid risk analysis method based on knowledge graphs, including the following steps: S1. Collect multi-source data related to hospital security management through various methods such as web scraping, calling interfaces of existing hospital information systems, and manual data entry, including: Medical equipment data, including equipment model, operating parameters, maintenance records, and fault history; Spatial area data, including grid division information, regional functional attributes, distribution of hazard sources, evacuation routes, etc. Personnel role data, including personnel name, position, scope of responsibilities, skills and qualifications, training records, etc. Institutional and standard data, including safety management systems, operating procedures, and risk threshold standards; Risk event data, including historical safety hazard records, accident handling results, and emergency response processes.

[0015] S2. Construct a knowledge graph. Use Named Entity Recognition (NER) from the collected multi-source data, defining 12 core entity types, including: medical equipment entities, spatial region entities, personnel role entities, institutional standard entities, risk event entities, and environmental parameter entities. Employ a BERT-based pre-trained model-based NER algorithm for entity extraction. The fine-tuning learning rate of the BERT pre-trained model is set to... The training iterations were set to 100 rounds, the batch size to 32, a function optimization model was used, and the convergence threshold was set to... ; Based on entity attributes and hospital safety management business logic, eight core relationships between entities are constructed, including "medical equipment - location", "personnel - responsible grid", "risk event - associated equipment" and "system standard - applicable area". The relationships between entities are explored and constructed through entity co-occurrence analysis, business rule matching and other methods. The Neo4j graph database is used to store entity and relationship data, generating a hospital safety knowledge graph. Graph database indexing optimization technology is used to improve the efficiency of entity relationship queries. The index creation rules are as follows, and unique indexes are created for the entity ID, entity type, and relationship type fields.

[0016] S3. Use Dijkstra's shortest path method to mine potential risk association paths between entities in the hospital safety knowledge graph; S301. Assigning each edge in the knowledge graph an attribute based on the degree of influence of the association on risk transmission. The weight value within the scope is greater if the degree of influence is higher. For example, the weight of the relationship between "medical equipment failure - location" is 8, and the weight of the relationship between "lack of personnel training - risk event" is 9. S302, Starting with the target risk entity (such as a medical device failure entity). The grid area entities are the target nodes. The shortest path from the starting node to each target node is calculated by updating the distance, and the expression is: in, Represents a node To the node edge weights, Indicates the starting node To the node The current shortest distance, initially, Other nodes Iterate and update until the distances between all nodes are stable; S303. Filter out the risk association paths whose shortest path length is less than a preset threshold (set to 15 in this embodiment) as potential risk association results, such as the path "MRI equipment failure - radiology area - radiology medical staff - patient".

[0017] S4. A risk level classification model is constructed using a logistic regression algorithm to accurately determine the risk level of the hospital grid area; multi-dimensional risk features are extracted to construct feature vectors, which are then input into the trained risk level classification model to obtain the risk level of the hospital grid area. S401. Extract multi-dimensional risk features and construct the input feature vector. ,in The feature dimensions, in this embodiment, include: number of risk-related paths, risk level of key entities, mean edge weight, deviation of environmental parameters, and device runtime, etc. S402, Introducing the Category Weight Vector ,in, Corresponding to low risk, Corresponding to medium risk, Corresponding to high risk, For significant risks, the risk level model outputs the probabilities for each category as follows: in, Indicates risk level label, This represents the transpose of the weight vector; S403. Train the risk level model using historical risk data, and optimize the model parameters by maximizing the log-likelihood function, as expressed in: in, Indicates the number of training samples. Indicates the indicator function (when the condition is met) ,otherwise The model parameters are optimized using gradient descent, with a learning rate of 0.01 and 500 training iterations, until the model converges (in this embodiment, the loss function value only needs to be stable below 0.05). S404. Input the feature vector of the grid area extracted in real time into the trained model, and select the risk level with the highest probability as the final risk level of the grid area. S5. Generate a targeted early warning plan based on the risk level and push the early warning information to the relevant responsible personnel.

[0018] S501. Classify warning levels: Set the corresponding risk levels as four warning levels, with low risk corresponding to blue warning, medium risk corresponding to yellow warning, high risk corresponding to orange warning, and major risk corresponding to red warning. S502. Generate early warning information: Generate early warning information based on the core entities in the risk association path, clarifying the source of risk, scope of impact and related personnel. For example, for the high risk caused by "radiology department MRI equipment failure", the early warning information is "the radiology department area MRI equipment failure risk is under orange warning, which may affect the examination safety of patients in the area and the occupational safety of medical staff. Please have the equipment maintenance personnel arrive at the scene immediately and the head of the radiology department should do a good job of on-site control." S503. Pushing early warning information: Pushing early warning information to the relevant responsible personnel through SMS, mobile APP push, and smart display screen sound and light alarm. The push priority follows the principle of "major risk > high risk > medium risk > low risk".

[0019] Example 2

[0020] This embodiment provides a hospital grid risk analysis system based on knowledge graphs, including: The data acquisition module is used to collect various types of data related to hospital safety management and provides multi-source data access interfaces; Knowledge graph construction module: used for entity extraction, relation construction and knowledge storage, generating a hospital safety knowledge graph; Risk Association Mining Module: Employs Dijkstra's algorithm to mine potential risk association paths between entities; Risk level classification module: Based on a logistic regression model, it enables accurate judgment of the risk level of grid areas; Early warning plan generation module: Generates targeted early warning plans based on risk levels and pushes early warning information; Data storage module: Neo4j graph database is used to store knowledge graph data, and MySQL database is used to store risk analysis results and early warning records.

[0021] Example 3

[0022] This embodiment uses a hospital's safety production system information platform as an example to demonstrate the complete implementation process of the method and system of the present invention. The relevant data for this grid is acquired from the hospital's safety production system information platform, equipment management system, and manual data entry terminal through the data acquisition module of the present invention. Specifically, this includes: like Figure 4 As shown, the system displays grid ID, responsible area, grid leader, grid member, and risk point information, including three main risk points / hazard sources (heptafluoropropane gas cylinder, fire control panel, and electrical distribution box). Based on historical assessments, the system classifies the overall risk of the grid as "high risk," with possible accident types being "gas fire" and "gas leak." like Figure 5 The preventative measures details interface shows that "daily safety checks" are performed within the grid once a day. The checks include electrical equipment, fire safety facilities, chemical reagent storage, and other items. Figure 6 The inspection of the distribution box is detailed down to the appearance and internal components, such as the condition of the box, wiring terminals, and cleanliness. like Figure 7 The system shown has a built-in three-level early warning system: high-risk warning (immediately stop work and evacuate personnel), medium-risk warning (strengthen monitoring and rectify within a time limit), and low-risk reminder (daily attention). like Figure 8 As shown, the emergency plans that have been activated include the "Fire Emergency Plan" and the "Electrical Accident Emergency Plan". The Fire Emergency Plan specifies the procedures for alarm, evacuation and fire extinguishing, and the emergency contacts are the alarm center and the grid leader.

[0023] based on Figures 4-8 The data presented in this invention demonstrates the following steps performed by the method provided: S1. Construct a hospital safety knowledge graph; Natural language processing (NLP) technology was used to extract entities and construct relationships from the collected data. In this embodiment, the core entities extracted include: the third-level grid entity "Second Zone Fire Control Room," the grid leader entity "A," the grid member entity "B," risk point entities "Hepflucopropane gas cylinder," "fire control panel," and "distribution box," accident type entities "gas fire" and "gas leak," preventative measure entity "daily safety inspection," inspection item entity "distribution box inspection," and emergency plan entity "fire emergency plan," etc. Simultaneously, relationships between entities were constructed, such as: "Second Zone Fire Control Room - contains - Heptafluoropropane gas cylinder," "Xie Ke - serves as grid leader - Second Zone Fire Control Room," "Hepflucopropane gas cylinder - may cause - gas fire," "Second Zone Fire Control Room - performs - daily safety inspection," and "Fire emergency plan - applies to - Second Zone Fire Control Room," etc. All entities and relationships are stored in the Neo4j graph database to form a hospital safety knowledge graph.

[0024] S2. Risk association mining based on knowledge graphs; Starting with the risk point "heptafluoropropane gas cylinder", Dijkstra's shortest path algorithm was used to explore potential risk transmission paths between it and other entities such as grid areas, personnel, and emergency plans. Edge weights were assigned according to the degree of risk impact. For example, the relationship "gas cylinder - located in - Zone 2 fire control room" was weighted at 7, and the relationship "gas cylinder - may cause - gas fire" was weighted at 9. The calculated shortest paths include: "heptafluoropropane gas cylinder → Zone 2 fire control room → grid length A → fire emergency plan", with a path length of 6 (less than the preset threshold of 15), indicating that the risk may trigger the emergency plan response through the grid length. Another path is: "heptafluoropropane gas cylinder → gas fire → Zone 2 fire control room → routine safety inspection → electrical box inspection", with a path length of 8, indicating the correlation between the gas fire risk and the electrical box inspection.

[0025] S3. Multi-dimensional feature fusion and risk level classification; Multi-dimensional risk features of the grid area are extracted to construct a 12-dimensional feature vector. In this embodiment, the extracted features include: number of risk-related paths (3), risk level of key entities (high risk), mean edge weight (7.2), deviation of environmental parameters (normal temperature and humidity), equipment running time (gas cylinder last inspection 30 days ago), and inspection completion rate (100%). The feature vector is input into a pre-trained logistic regression risk level classification model, and the model outputs the probabilities of each category: low risk 0.05, medium risk 0.12, high risk 0.78, and major risk 0.05. Therefore, the current risk level of the grid area is determined to be "high risk".

[0026] S4. Generate and push out early warning plans; Based on the risk level of "high risk," the warning level is matched with "orange warning." A warning message is generated based on the core entities in the risk-related path: "[Orange Warning] There is a risk of leakage or fire from the heptafluoropropane gas cylinders in the fire control room of Zone 2, which may affect the fire safety and personnel safety of the area. Grid leader Xie Ke is requested to immediately organize an inspection, contact professionals to conduct gas cylinder testing, and activate the fire emergency plan." The warning message is pushed to grid leader A and grid member B via a mobile app, and simultaneously displays an orange warning notification on the smart display screen in the fire control room of Zone 2, triggering an audible and visual alarm.

[0027] S5. Early warning response and follow-up; The system records the early warning push time, response time, handling process and results; after receiving the early warning, grid leader A immediately arranged for personnel to conduct a special inspection of the gas cylinder, found that the pressure gauge reading was close to the critical value, and then contacted the maintenance unit to refill and replace it, and filled in the handling record in the system. The system stores the closed-loop data of this risk event in the MySQL database for subsequent model optimization and historical analysis.

[0028] As can be seen from this embodiment, the method and system of the present invention can effectively integrate the hospital's existing grid-based safety management data into a knowledge graph, automatically mine hidden risk associations, accurately determine the risk level, and generate targeted early warnings and contingency plan suggestions, significantly improving the intelligence level and response efficiency of hospital safety management.

[0029] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A hospital grid risk analysis method based on knowledge graphs, characterized in that, Includes the following steps: S1. Collect multi-source data related to hospital safety management, including medical equipment data, spatial area data, personnel role data, system and standard data, and risk event data; S2. Construct a knowledge graph by using the named entity recognition method in natural language processing to extract core entities from the collected multi-source data, build the relationships between entities, store the entity and relationship data in a graph database, and generate a hospital safety knowledge graph. S3. Use Dijkstra's shortest path method to mine potential risk association paths between entities in the hospital safety knowledge graph; S4. Extract multi-dimensional risk features to construct feature vectors, input them into the trained risk level classification model, and obtain the risk level of the hospital grid area; S401. Extract multi-dimensional risk features and construct the input feature vector. ,in Indicates the feature dimension; S402, Introducing the Category Weight Vector ,in, Corresponding to low risk, Corresponding to medium risk, Corresponding to high risk, For significant risks, the risk level model outputs the probabilities for each category as follows: in, Indicates risk level label, This represents the transpose of the weight vector; S403. Train the risk level model using historical risk data, and optimize the model parameters by maximizing the log-likelihood function, as expressed in: in, Indicates the number of training samples. The indicator function is used; the model parameters are optimized using gradient descent, with a learning rate of 0.01 and 500 training iterations until the model converges. S404. Input the feature vector of the grid area extracted in real time into the trained model, and select the risk level with the highest probability as the final risk level of the grid area. S5. Generate a targeted early warning plan based on the risk level and push the early warning information to the relevant responsible personnel.

2. The hospital grid risk analysis method based on knowledge graphs according to claim 1, characterized in that, Entity extraction is performed using a named entity recognition algorithm based on a BERT pre-trained model, with the fine-tuning learning rate of the BERT pre-trained model set to... The training iterations were set to 100 rounds, the batch size to 32, a function optimization model was used, and the convergence threshold was set to... .

3. The hospital grid risk analysis method based on knowledge graphs according to claim 2, characterized in that, The relationships between entities include "medical equipment - location", "personnel - responsible grid", "risk event - associated equipment", and "system / standard - applicable area"; The Neo4j graph database is used to store entity and relationship data, and unique indexes are created for the entity ID, entity type, and relationship type fields.

4. The hospital grid risk analysis method based on knowledge graphs according to claim 1, characterized in that, The Dijkstra shortest path method for mining potential risk association paths between entities in the hospital safety knowledge graph specifically includes: S301. Assigning each edge in the knowledge graph an attribute based on the degree of influence of the association on risk transmission. Weight values ​​within the range; S302, Taking the target risk entity as the starting point The grid area entities are the target nodes. The shortest path from the starting node to each target node is calculated by updating the distance, and the expression is: in, Represents a node To the node edge weights, Indicates the starting node To the node The current shortest distance, initially, Other nodes Iterate and update until the distances between all nodes are stable; S303. Select the risk-related paths whose shortest path length is less than a preset threshold as potential risk-related results.

5. The hospital grid risk analysis method based on knowledge graph according to claim 1, characterized in that, The steps for generating the early warning scheme specifically include: S501. Classify warning levels: Set the corresponding risk levels as four warning levels, with low risk corresponding to blue warning, medium risk corresponding to yellow warning, high risk corresponding to orange warning, and major risk corresponding to red warning. S502. Generate early warning information: Generate early warning information based on the core entities in the risk association path, and clarify the source of risk, scope of impact and related personnel; S503. Pushing early warning information: Pushing early warning information to the relevant responsible personnel through SMS, mobile APP push, and smart display screen sound and light alarm. The push priority follows the principle of "major risk > high risk > medium risk > low risk".

6. A hospital grid risk analysis system based on knowledge graphs, employing the method described in any one of claims 1-5, characterized in that, include: The data acquisition module is used to collect various types of data related to hospital safety management and provides multi-source data access interfaces; Knowledge graph construction module: used for entity extraction, relation construction and knowledge storage, generating a hospital safety knowledge graph; Risk Association Mining Module: Employs Dijkstra's algorithm to mine potential risk association paths between entities; Risk level classification module: Based on a logistic regression model, it enables accurate judgment of the risk level of grid areas; Early warning plan generation module: Generates targeted early warning plans based on risk levels and pushes early warning information; Data storage module: Neo4j graph database is used to store knowledge graph data, and MySQL database is used to store risk analysis results and early warning records.

Images