Medical institution safety production risk reasoning method, device and equipment and storage medium

Abstract

The application provides a medical institution safety production risk reasoning method, device and equipment and a storage medium, relates to the technical field of intelligent medical treatment, and comprises the following steps: inputting real-time monitoring data into a medical institution safety production knowledge graph, mining hidden risk relations in the medical institution safety production knowledge graph, and calculating time-varying risk transmission weights; outputting a risk transmission path and a medical institution safety production risk priority based on the time-varying risk transmission weights; wherein the medical institution safety production knowledge graph is constructed based on triple structure extracted from multi-source data, the triple includes entities, relations and attributes, hidden risk associations are found through graph reasoning, and indirect risk transmission paths are dynamically output; rules do not need to be manually written, and the method is suitable for regulation updates and scene changes.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smart medical care, and in particular to a medical institution safety production risk reasoning method, device and equipment and a storage medium. BACKGROUND

[0002] The application of the current medical institution safety production field knowledge graph mainly focuses on static data storage and simple query, such as department fire-fighting equipment list retrieval, and lacks complex logical reasoning function. The reasoning method is based on rules such as "fire-fighting equipment exceeding the service life corresponds to high risk", which is difficult to handle dynamic changing risk scenarios, has high rule maintenance cost, is difficult to cover all scenarios, and cannot adapt to the logical changes after the update of new regulations or equipment. The rule-based reasoning method cannot identify the existence of implicit association relationship, causing hidden danger discovery lag; in addition, the medical institution safety production data is scattered in unstructured documents such as regulation files, inspection records and accident cases, and has not formed a structured knowledge network, lacking the ability of cross-source data association and integration, and being unable to mine deep risk association. SUMMARY

[0003] The present application provides a medical institution safety production risk reasoning method, device, equipment and storage medium, which solves the defects of complex reasoning missing, high rule maintenance cost and hidden danger discovery lag of the traditional medical institution safety production risk reasoning method.

[0004] The present application provides a medical institution safety production risk reasoning method, which comprises:

[0005] Input real-time monitoring data into the medical institution safety production knowledge graph, mine the implicit risk relationship in the medical institution safety production knowledge graph, and calculate the time-varying risk transmission weight;

[0006] Output the risk transmission path and the medical institution safety production risk priority based on the time-varying risk transmission weight;

[0007] Wherein, the medical institution safety production knowledge graph is constructed based on the triple structure extracted from multi-source data, and the triple includes entity, relationship and attribute.

[0008] According to the medical institution safety production risk reasoning method provided by the present application, the construction method of the medical institution safety production knowledge graph comprises:

[0009] Extract entities and relationships from multi-source data, including medical regulations, equipment inspection records and accident reports;

[0010] Input the entity and the relationship into a graph database, input the entity as a node in the graph database, and input the relationship as an edge in the graph database, to generate a medical institution safety production knowledge graph, wherein the node includes a department, equipment, and a regulation clause, and the edge includes a membership relationship, a risk association, and a space-time constraint relationship.

[0011] According to the medical institution safety production risk reasoning method provided by the application, the implicit risk relationship in the medical institution safety production knowledge graph is mined, and the time-varying risk transmission weight is calculated, including:

[0012] The features of the neighbor nodes of the target entity are aggregated through a graph neural network model;

[0013] The time-varying risk transmission weight between entities is calculated based on the features of the neighbor nodes of the target entity.

[0014] According to the medical institution safety production risk reasoning method provided by the application, the training method of the graph neural network model includes:

[0015] The historical accident data is used as a positive sample, and a negative sample is generated by randomly masking the graph edge and the node attribute;

[0016] The positive sample and the negative sample are input into the graph neural network model to obtain a prediction result;

[0017] A contrast learning loss function is constructed based on the prediction result and the actual result, the model parameters are optimized based on the contrast learning loss function, and a trained graph neural network model is obtained.

[0018] According to the medical institution safety production risk reasoning method provided by the application, the time-varying risk transmission weight between entities is calculated based on the features of the neighbor nodes of the target entity, including:

[0019] The neighbor nodes of the target entity are obtained, the key features of the neighbor nodes are extracted, the key features include the equipment state, the historical failure frequency, and the regulation compliance, and the key features are encoded into a feature vector of a unified dimension through a neural network;

[0020] The node feature similarity is obtained based on the feature vector;

[0021] The space-time attenuation factor is calculated based on the deviation degree of the equipment detection time difference and the standard period;

[0022] The time-varying risk transmission weight between the target entity and each neighbor node is calculated based on the node feature similarity and the space-time attenuation factor through a graph attention mechanism.

[0023] According to the medical institution safety production risk reasoning method provided by the application, the time-varying risk transmission weight is output based on the time-varying risk transmission weight, including:

[0024] generating a multi-hop risk conduction path according to a time-varying risk conduction weight between the entities based on a path search algorithm.

[0025] According to the medical institution safety production risk reasoning method provided by the application, the time-varying risk conduction weight is used to output a medical institution safety production risk priority, which comprises the following steps:

[0026] The risk features of the neighbor nodes are weighted and summed according to the attention weights, and the weighted sum is superimposed on the target entity to generate a comprehensive risk score.

[0027] According to the comprehensive risk score, a corresponding medical institution safety production risk priority is output.

[0028] The application further provides a medical institution safety production risk reasoning device based on a knowledge graph, which comprises the following steps:

[0029] The mining module is used for inputting real-time monitoring data into a medical institution safety production knowledge graph, mining an implied risk relationship in the medical institution safety production knowledge graph, and calculating a time-varying risk conduction weight.

[0030] The output module is used for outputting a risk conduction path and a medical institution safety production risk priority based on the time-varying risk conduction weight.

[0031] The medical institution safety production knowledge graph is constructed based on a triple structure extracted from multi-source data, and the triple comprises an entity, a relationship and an attribute.

[0032] The application further provides an electronic device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the medical institution safety production risk reasoning method according to any one of the above.

[0033] The application further provides a non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program is executable by a processor to implement the medical institution safety production risk reasoning method according to any one of the above.

[0034] The medical institution safety production risk reasoning method, device, equipment and storage medium provided by the application, by inputting real-time monitoring data into a medical institution safety production knowledge graph, mining the implied risk relationship in the medical institution safety production knowledge graph, and calculating the time-varying risk transmission weight; outputting the risk transmission path and the medical institution safety production risk priority based on the time-varying risk transmission weight; wherein the medical institution safety production knowledge graph is constructed based on the triple structure extracted from multi-source data, the triple includes entities, relationships and attributes, the implied risk association is found through graph reasoning, and the indirect risk transmission path is dynamically output; without manually writing rules, the method is suitable for regulation updates and scene changes. BRIEF DESCRIPTION OF DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0036] Figure 1 is a flowchart of the medical institution safety production risk reasoning method provided by the embodiment of the application;

[0037] Figure 2 is a functional structure schematic diagram of the medical institution safety production risk reasoning device based on a knowledge graph provided by the embodiment of the application;

[0038] Figure 3 is a functional structure schematic diagram of an electronic device provided by the embodiment of the application. DETAILED DESCRIPTION

[0039] In order to make the objects, technical solutions and advantages of the application clearer, the technical solutions in the application will be described clearly and completely in the following with reference to the drawings in the application. Obviously, the described embodiments are some embodiments of the application, not all embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor belong to the protection scope of the application.

[0040] Figure 1 The flowchart of the medical institution safety production risk reasoning method provided by the embodiment of the application is shown in Figure 1 The medical institution safety production risk reasoning method provided by the embodiment of the application includes:

[0041] Step 101, input real-time monitoring data into a medical institution safety production knowledge graph, mine the implied risk relationship in the medical institution safety production knowledge graph, and calculate the time-varying risk transmission weight;

[0042] Step 102, outputting a risk transmission path and a medical institution safety production risk priority based on the time-varying risk transmission weight;

[0043] The medical institution safety production knowledge graph is constructed based on triples extracted from multi-source data, and the triples include entities, relationships and attributes.

[0044] The application of the traditional medical institution safety production field knowledge graph is mainly concentrated in static data storage and simple query, such as department fire-fighting equipment list retrieval, and lacks complex logical reasoning function. The reasoning method is based on rules, which is difficult to handle dynamic changing risk scenarios, has high rule maintenance cost, is difficult to cover all scenarios, and cannot adapt to the logical changes after the update of new regulations or equipment. The rule-based reasoning method cannot identify the existence of implicit association relationship, causing hidden danger discovery lag; in addition, the medical institution safety production data is scattered in unstructured documents such as regulation files, inspection records and accident cases, and has not formed a structured knowledge network, lacking the ability of cross-source data association and integration, and being unable to mine deep risk association.

[0045] The traditional medical institution safety production system performs risk judgment based on fixed rule chain, and the embodiment of the application realizes adaptive calculation of time-varying risk transmission weight through time-varying weight neural network, and dynamically generates a risk transmission path based on the time-varying risk transmission weight.

[0046] The medical institution safety production risk reasoning method provided by the embodiment of the application inputs real-time monitoring data into the medical institution safety production knowledge graph, mines implicit risk relationships in the medical institution safety production knowledge graph, and calculates time-varying risk transmission weights; outputs a risk transmission path and a medical institution safety production risk priority based on the time-varying risk transmission weights; wherein the medical institution safety production knowledge graph is constructed based on triples extracted from multi-source data, the triples include entities, relationships and attributes, implicit risk associations are found through graph reasoning, and indirect risk transmission paths are dynamically output; without manually writing rules, the method adapts to regulation updates and scenario changes.

[0047] Based on any of the above embodiments, the construction method of the medical institution safety production knowledge graph comprises:

[0048] Step 201, extracting entities and relationships from multi-source data, the multi-source data including but not limited to medical regulation texts, device inspection records and accident reports, etc.;

[0049] Step 202, input the entity and relationship into a graph database, the entity as a node in the graph database, the relationship as an edge in the graph database, generate a medical institution safety production knowledge graph, wherein the node includes department, equipment and regulation clause, and the edge includes affiliation, risk association and space-time constraint relationship.

[0050] For example, extract entities "safety exit", "fire extinguisher" and relationship "must be equipped with" "periodic inspection items" from regulations (such as "Nine Regulations on Fire Safety Management of Medical Institutions"); use a graph database such as Neo4j for storage, nodes are entities such as departments, equipment, and regulation clauses, and edges are relationships such as "belongs to" "associated risk".

[0051] In some embodiments of the present application, in the case of missing medical data (such as new equipment without historical detection records), the embodiments of the present application use common sense knowledge graph completion technology for reasoning enhancement:

[0052] External knowledge injection: extract general rules from medical industry knowledge base (such as clinical guidelines, equipment manuals), for example:

[0053] "Fire hydrant corrosion rate increases by 20% in humid environment" ->

[0054] Automatically generate hidden edges: (fire hydrant)-[environmental corrosion risk]->(high humidity area)

[0055] Or, use TransE and other embedding models to predict missing relationships, and the completed graph can be used for more comprehensive risk reasoning.

[0056] In some embodiments of the present application, to integrate unstructured data (such as equipment images, monitoring videos), the embodiments of the present application extend the knowledge graph to the multi-modal field:

[0057] Visual data fusion, use CNN to extract rust and damage features from fire-fighting equipment images, and convert them into graph node attributes; associate visual features with structured data (such as rust degree-> fault probability mapping) through graph attention mechanism. Real-time behavior recognition is performed on monitoring videos (such as "blocking fire exit with objects"), and dynamic risk event nodes are generated; and real-time reasoning is triggered in linkage with static graph, such as channel blockage + surgery-> emergency evacuation risk.

[0058] Based on any of the above embodiments, the method for mining implicit risk relationships in the medical institution safety production knowledge graph and calculating time-varying risk transmission weights comprises:

[0059] Step 301, aggregate features of neighbor nodes of a target entity through a graph neural network model;

[0060] Step 302, calculating the time-varying risk conduction weight between entities based on the characteristics of the target entity neighbor nodes.

[0061] In the embodiments of the present application, the inference process for mining the implicit risk relationship in the medical institution safety production knowledge graph comprises:

[0062] (1) input the data to be inferred (such as "fire hydrant in a certain department overdue and not checked");

[0063] (2) aggregate neighbor node features (such as historical accident records of the department, risk levels of adjacent departments) through a graph neural network model (such as GraphSAGE);

[0064] (3) calculate the risk conduction weight between entities based on the graph attention mechanism, and generate a risk probability distribution;

[0065] (4) output the inference result (such as "the fire risk level of the department is upgraded to intermediate") and the associated path (such as "overdue and not checked → equipment failure → fire spread risk").

[0066] The embodiments of the present application discover the implicit risk association through graph inference, such as the indirect risk conduction path between "surgery room circuit load exceeding standard" and "design defects of floor fire passage". Without manually writing rules, the algorithm automatically learns the risk logic, adapts to regulatory updates and changes in scenarios, and provides a visual risk conduction path to help managers locate the root cause (such as "institutional loopholes → execution bias → equipment maintenance → accident") and develop targeted measures.

[0067] Based on any of the above embodiments, the training method of the graph neural network model comprises:

[0068] Step 401, using historical accident data as positive samples, generating negative samples by randomly masking graph edges and node attributes;

[0069] Step 402, inputting the positive samples and the negative samples into the graph neural network model to obtain a prediction result;

[0070] Step 403, constructing a contrastive learning loss function based on the prediction result and the actual result, optimizing the model parameters based on the contrastive learning loss function, and obtaining a trained graph neural network model.

[0071] In some embodiments of the present application, dynamic risk inference can also be implemented based on the fusion of rule engines and machine learning. In the medical institution safety production risk inference, interpretability and accuracy are equally important. The present embodiment introduces an expert rule engine as a constraint condition based on the original graph neural network inference model, and the specific implementation is as follows:

[0072] Expert rule embedding: Convert hard requirements in medical institution safety production regulations (such as "Medical Institution Fire Safety Specification") into logical constraints;

[0073] Constraint optimization training: Introduce rule constraints through loss function during graph neural network model training to ensure that model output conforms to industry specifications.

[0074] Based on any of the above embodiments, the time-varying risk transmission weight between entities is calculated based on the features of the target entity neighbor nodes, comprising:

[0075] Step 501, obtaining the neighbor nodes of the target entity, extracting the key features of the neighbor nodes, the key features including device state, historical failure times and regulatory compliance, and encoding the key features into a unified dimension feature vector through a neural network;

[0076] Step 502, obtaining node feature similarity based on the feature vector;

[0077] Step 503, calculating the space-time attenuation factor based on the deviation of the device detection time difference from the standard period;

[0078] Step 504, using a graph attention mechanism to calculate the time-varying risk transmission weight between the target entity and each neighbor node based on the node feature similarity and the space-time attenuation factor.

[0079] Based on any of the above embodiments, the time-varying risk transmission weight is based on the output of the risk transmission path, comprising:

[0080] Based on the path search algorithm, a multi-hop risk transmission path is generated according to the time-varying risk transmission weight between entities.

[0081] The multi-hop risk transmission path is, for example, device aging → process vulnerability → personnel injury (3-hop transmission).

[0082] The path reasoning process includes:

[0083] Initial event: The sensor detects that the operating room distribution box temperature exceeds the standard (node A);

[0084] First jump reasoning: The graph associates the distribution box → power supply line → breathing machine (node B), and calculates the overload risk probability (aggregated electrical load feature);

[0085] Second jump reasoning: It is found that the breathing machine backup battery (node C) is not tested every month according to "Medical Device Maintenance Specification", triggering a system vulnerability label;

[0086] Third jump reasoning: Combined with scheduling data, it is predicted that the in-operation power failure + backup failure (node D) will cause the risk of patient hypoxia;

[0087] Output result: risk path: distribution box overload → breathing machine power supply interruption → backup battery failure → patient hypoxia risk; risk level: urgent (red); disposal suggestion: immediately overhaul the distribution box, test the backup battery and revise the monthly inspection system.

[0088] Based on any of the above embodiments, the output of the time-varying risk conduction weight based on the time-varying risk conduction weight includes:

[0089] Step 601, the risk features of the neighbor nodes are weighted and summed according to the attention weight, and the weighted sum is superimposed on the target entity to generate a comprehensive risk score;

[0090] Step 602, output the safety production risk priority of the corresponding medical institution according to the comprehensive risk score, and the safety production risk priority of the medical institution is shown in Table 1.

[0091] Table 1 Safety production risk priority of medical institutions

[0092]

[0093] The medical institution safety production risk reasoning method provided by the embodiment of the application, by constructing a medical institution safety production knowledge graph: integrating multiple source data such as regulations, standards, equipment information, and accident cases, constructing a knowledge network containing "entity - relationship - attribute" (such as entity: fire hydrant, relationship: installed in department, attribute: detection time, model); using graph neural network (GNN) or path search algorithm (such as PRA), mining implicit risk relationships (such as the association path of "equipment overdue inspection" and "historical fire accident") in the knowledge graph; input real-time monitoring data (such as equipment status, inspection results), predict potential risks through graph reasoning, output risk conduction path and priority. Through the deep combination of knowledge graph and dynamic reasoning, the medical institution safety production management paradigm upgrade from "passive response" to "active prevention" is realized.

[0094] The knowledge graph-based medical institution safety production risk reasoning device provided by the application is described below, and the knowledge graph-based medical institution safety production risk reasoning device described below can be correspondingly referred to the medical institution safety production risk reasoning method described above.

[0095] Figure 2 The structure diagram of the knowledge graph-based medical institution safety production risk reasoning device provided by the embodiment of the application is shown in Figure 2 The knowledge graph-based medical institution safety production risk reasoning device provided by the embodiment of the application includes:

[0096] The mining module 201 is configured to input real-time monitoring data into a medical institution safety production knowledge graph, mine an implied risk relationship in the medical institution safety production knowledge graph, and calculate a time-varying risk transmission weight.

[0097] The output module 202 is configured to output a risk transmission path and a medical institution safety production risk priority based on the time-varying risk transmission weight.

[0098] The medical institution safety production knowledge graph is constructed based on a triple structure extracted from multi-source data, and the triple includes an entity, a relationship, and an attribute.

[0099] The medical institution safety production risk reasoning device based on the knowledge graph provided in the embodiment of the application inputs real-time monitoring data into a medical institution safety production knowledge graph, mines an implied risk relationship in the medical institution safety production knowledge graph, and calculates a time-varying risk transmission weight. The device outputs a risk transmission path and a medical institution safety production risk priority based on the time-varying risk transmission weight. The medical institution safety production knowledge graph is constructed based on a triple structure extracted from multi-source data, and the triple includes an entity, a relationship, and an attribute. The device discovers an implied risk association through graph reasoning and dynamically outputs an indirect risk transmission path. The device does not need to manually write rules and is suitable for regulation updates and scene changes.

[0100] Figure 3 An example of an entity structure diagram of an electronic device is shown in Figure 3 The electronic device can include a processor 310, a communications interface 320, a memory 330, and a communications bus 340. The processor 310, the communications interface 320, and the memory 330 can communicate with each other through the communications bus 340. The memory 330 includes a computer program, an operating system, and acquired data. The processor 310 can invoke logical instructions in the memory 330 to execute a medical institution safety production risk reasoning method. The method includes inputting real-time monitoring data into a medical institution safety production knowledge graph, mining an implied risk relationship in the medical institution safety production knowledge graph, and calculating a time-varying risk transmission weight. The method also includes outputting a risk transmission path and a medical institution safety production risk priority based on the time-varying risk transmission weight. The medical institution safety production knowledge graph is constructed based on a triple structure extracted from multi-source data, and the triple includes an entity, a relationship, and an attribute.

[0101] Moreover, the logic instructions in the memory 330 described above can be implemented in the form of software functional units and sold or used as independent products, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that contribute to the related art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0102] In another aspect, the present application also provides a non-transitory computer readable storage medium having stored thereon a computer program, which, when executed by a processor, implements a medical institution safety production risk inference method provided by the above method, the method comprising: inputting real-time monitoring data into a medical institution safety production knowledge graph, mining implicit risk relationships in the medical institution safety production knowledge graph, and calculating time-varying risk transmission weights; outputting a risk transmission path and a medical institution safety production risk priority based on the time-varying risk transmission weights; wherein the medical institution safety production knowledge graph is constructed based on triples extracted from multi-source data, and the triples include entities, relationships and attributes.

[0103] The device embodiments described above are only schematic, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., they can be located in one place, or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the present embodiment according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0104] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be implemented by means of software plus necessary general hardware platforms, and of course can also be implemented by hardware. Based on such understanding, the above technical solutions essentially or the parts that contribute to the related art can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0105] Finally, it should be noted that the above examples are only used to illustrate the technical solutions of the present application, and are not intended to limit the same; although the present application has been described in detail with reference to the foregoing examples, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacements for some of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A medical institution safety production risk inference method, characterized in that, The method comprises the following steps: input real-time monitoring data into a medical institution safety production knowledge graph, mine implicit risk relationships in the medical institution safety production knowledge graph, and calculate time-varying risk transmission weights; the method of mining implicit risk relationships in the medical institution safety production knowledge graph and calculating time-varying risk transmission weights comprises the following steps: aggregate features of neighbor nodes of a target entity through a graph neural network model; the method of calculating time-varying risk transmission weights between entities based on the features of the neighbor nodes of the target entity comprises the following steps: obtain neighbor nodes of a target entity, extract key features of the neighbor nodes, the key features comprising device states, historical failure times, and regulatory compliance, encode the key features into feature vectors of a uniform dimension through a neural network; obtain node feature similarity based on the feature vectors; calculate a space-time attenuation factor based on a deviation of a device detection time difference from a standard period; calculate time-varying risk transmission weights between the target entity and each neighbor node based on the node feature similarity and the space-time attenuation factor through a graph attention mechanism; output a risk transmission path and a medical institution safety production risk priority based on the time-varying risk transmission weights; the method of outputting a risk transmission path based on the time-varying risk transmission weights comprises the following steps: generate a multi-hop risk transmission path according to the time-varying risk transmission weights between entities based on a path search algorithm; the method of outputting a medical institution safety production risk priority based on the time-varying risk transmission weights comprises the following steps: weight and sum risk features of neighbor nodes according to attention weights, superimpose the weighted sum result on a target entity, and generate a comprehensive risk score; output a corresponding medical institution safety production risk priority according to the comprehensive risk score. The medical institution safety production knowledge graph is constructed based on triples extracted from multi-source data, and the triples comprise entities, relationships, and attributes.

2. The medical facility safety production risk inference method of claim 1, wherein, The construction method of the medical institution safety production knowledge graph comprises the following steps: extract entities and relationships from multi-source data, the multi-source data comprising medical institution safety production-related regulation texts, business management records, device operation and maintenance records, and accident reports; input the entities and relationships into a graph database, take the entities as nodes in the graph database, take relationships as edges in the graph database, and generate a medical institution safety production knowledge graph, wherein the nodes comprise departments, devices, and regulation clauses, and the edges comprise affiliation relationships, risk associations, and space-time constraint relationships.

3. The medical facility safety production risk inference method of claim 1, wherein, The training method of the graph neural network model comprises the following steps: use historical accident data as positive samples, generate negative samples by randomly masking graph edges and node attributes; input the positive samples and the negative samples into a graph neural network model to obtain a prediction result; construct a contrastive learning loss function based on the prediction result and an actual result, optimize model parameters based on the contrastive learning loss function, and obtain a trained graph neural network model.

4. A knowledge graph-based medical institution safety production risk reasoning device, characterized in that, The method comprises the following steps: a mining module is configured to input real-time monitoring data into a medical institution safety production knowledge graph, mine implicit risk relationships in the medical institution safety production knowledge graph, and calculate time-varying risk transmission weights; The mining of the hidden risk relationship in the medical institution safety production knowledge graph and the calculation of the time-varying risk transmission weight include: The features of the target entity neighbor nodes are aggregated through a graph neural network model; The time-varying risk transmission weight between entities is calculated based on the features of the target entity neighbor nodes, specifically including: The neighbor nodes of the target entity are obtained, and the key features of the neighbor nodes are extracted, the key features including device state, historical failure times and regulatory compliance, the key features are encoded into a unified dimension feature vector through a neural network; The node feature similarity is obtained based on the feature vector; The spatio-temporal attenuation factor is calculated based on the deviation of the device detection time difference from the standard period; The graph attention mechanism is used to calculate the time-varying risk transmission weight between the target entity and each neighbor node based on the node feature similarity and the spatio-temporal attenuation factor An output module is configured to output a risk transmission path and a medical institution safety production risk priority based on the time-varying risk transmission weight, including: A multi-hop risk transmission path is generated based on the time-varying risk transmission weight between entities according to a path search algorithm; The medical institution safety production risk priority is outputted based on the time-varying risk transmission weight, including: The risk features of the neighbor nodes are weighted and summed according to the attention weight, and the weighted sum result is superimposed on the target entity to generate a comprehensive risk score; the corresponding medical institution safety production risk priority is outputted according to the comprehensive risk score; The medical institution safety production knowledge graph is constructed based on the triple structure extracted from multiple sources of data, and the triple includes entities, relationships and attributes.

5. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, The processor executes the program to realize the medical institution safety production risk reasoning method of any one of claims 1 to 3.

6. A non-transitory computer readable storage medium having stored thereon a computer program, the computer program comprising program instructions that, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 5. The computer program is executed by the processor to realize the medical institution safety production risk reasoning method of any one of claims 1 to 3.

Images