Medical service item compliance dynamic monitoring method and system based on artificial intelligence

Abstract

The invention provides a medical service item compliance dynamic monitoring method and system based on artificial intelligence, and relates to the technical field of medical service management.The method comprises the steps that a data incidence relation matrix is established through the unbalanced state entropy theory, and risk features are extracted through a chaotic neural network; a risk prediction spectrum is generated in combination with deep generative adversarial diagnosis and treatment modeling and heterogeneous medical network neural evolution analysis, and a causal reasoning decision tree is adopted to construct a time sequence evolution model to position abnormal risk points. The medical service project compliance real-time monitoring and risk early warning can be realized, the supervision efficiency is improved, and the medical risk is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of medical service management, and in particular to a method and system for dynamically monitoring the compliance of medical service items based on artificial intelligence. Background Art

[0002] As the complexity of medical service projects continues to increase, medical service compliance monitoring has become a key link in medical quality management. Existing medical service compliance monitoring mainly adopts manual review and rule matching methods, and compliance is determined by setting fixed monitoring indicators and thresholds. At the same time, some medical institutions have begun to use artificial intelligence technologies such as machine learning to analyze medical service data to identify and warn of non-compliant behaviors.

[0003] However, the existing technologies have the following problems: traditional rule matching methods are difficult to cope with the dynamic changes of medical service items, and fixed monitoring indicators cannot accurately reflect the actual operational risks; existing machine learning methods mainly focus on static features and lack the ability to dynamically analyze the evolution of risks in the medical service process; the correlation and timing characteristics between medical service items are insufficiently considered, making it difficult to achieve systematic risk prediction; there is a lack of timely intervention mechanism for abnormal risks, and the risk prevention and control effect is limited.

[0004] In summary, it is urgent to solve the technical problems in the dynamic monitoring of medical service compliance. Specifically, by establishing a data association relationship matrix to achieve dynamic knowledge mapping, using deep generative adversarial networks and heterogeneous medical networks to extract diagnosis and treatment features, and building a risk evolution model based on a causal reasoning decision tree, a complete set of risk identification, prediction and intervention mechanisms is formed, thereby improving the accuracy and real-time performance of medical service compliance monitoring. The present invention can solve the problems in the prior art. Summary of the invention

[0005] The embodiments of the present invention provide a method and system for dynamically monitoring the compliance of medical service items based on artificial intelligence, which can solve the problems in the prior art.

[0006] According to a first aspect of the embodiments of the present invention,

[0007] Provided is an artificial intelligence-based dynamic monitoring method for compliance of medical service items, including:

[0008] The non-equilibrium entropy theory is used to dynamically map the medical service project data, establish a data association relationship matrix, and obtain project knowledge association data; based on the data association relationship matrix, a chaotic neural network is used to extract the risk feature distribution in the project knowledge association data; the Lyapunov stability analyzer is used to predict the long-term trend of the risk feature distribution, and a risk benchmark model containing risk assessment weights and dynamic monitoring thresholds is generated;

[0009] The deep generative adversarial diagnosis and treatment modeler is used to extract diagnosis and treatment feature sequences from project operation data and construct a knowledge representation model. Based on the knowledge representation model, a heterogeneous medical network neural evolution device is used to analyze the correlation strength and timing characteristics between various diagnosis and treatment activities, and a risk prediction spectrum is generated after comparison with historical compliance records. Combined with the risk benchmark model, the risk prediction spectrum is dynamically evolved and analyzed through the medical value dynamics framework to output real-time compliance analysis results.

[0010] A causal reasoning decision tree is used to build a time-series evolution model of real-time compliance assessment results, calculate the risk evolution gradient, dynamically decompose the risk status matrix based on the risk evolution gradient, locate abnormal risk points, and generate risk warning information and risk intervention strategies based on the spatiotemporal distribution characteristics of abnormal risk points; the spatiotemporal distribution characteristics, risk warning information and risk intervention strategies are fed back to the data association relationship matrix to update the knowledge association data of medical service items.

[0011] In an optional embodiment,

[0012] The non-equilibrium entropy theory is used to dynamically map the medical service project data, establish a data association relationship matrix, and obtain project knowledge association data; based on the data association relationship matrix, a chaotic neural network is used to extract the risk feature distribution in the project knowledge association data; the Lyapunov stability analyzer is used to predict the long-term trend of the risk feature distribution, and a risk benchmark model containing risk assessment weights and dynamic monitoring thresholds is generated, including:

[0013] The medical service project data is constructed into an initial data set according to the service type, execution time, operator, service object, and operation record; the non-equilibrium entropy value between any two data items is obtained according to the probability distribution of each data item in the initial data set on the corresponding dimension, and the association strength between the data items is calculated according to the non-equilibrium entropy value and the smoothing factor to construct a data association relationship matrix; a time window mechanism is introduced into the data association relationship matrix, and the data association relationship matrix of the current time window and the historical data association relationship matrix are weightedly combined through the forgetting factor to generate project knowledge association data of the medical service project;

[0014] The association strength in the project knowledge association data is used as the neuron connection weight to construct a chaotic neural network. The nonlinear iterative mapping is used to process the neuron input signal to keep the chaotic neural network in a chaotic critical state. The output of the chaotic neural network is reduced in dimension by the local sensitive hashing algorithm to obtain the risk characteristic distribution of medical service items.

[0015] The risk characteristic distribution is converted into a dynamic system equation, and the Lyapunov index of the dynamic system equation in each characteristic dimension is calculated, and the Lyapunov index represents the degree of trajectory divergence of the characteristic dimension; the Lyapunov index is subjected to exponential transformation and normalization to generate a risk assessment weight; the historical data statistical value of each characteristic dimension is combined with the Lyapunov index of the corresponding dimension to generate a dynamic monitoring threshold, and the risk assessment weight and the dynamic monitoring threshold are combined to construct a risk benchmark model for medical service projects.

[0016] In an optional embodiment,

[0017] The association strength in the project knowledge association data is used as the neuron connection weight to construct a chaotic neural network. The nonlinear iterative mapping is used to process the neuron input signal to maintain the chaotic critical state of the chaotic neural network. The output of the chaotic neural network is reduced in dimension by the local sensitive hashing algorithm. The risk characteristic distribution of medical service items is obtained, including:

[0018] The correlation strength of project knowledge association data is constructed as the neuron connection weight, and the chaotic critical state of the network is controlled by nonlinear iterative mapping. The weight is dynamically updated through the chaotic adaptive adjustment mechanism, the network synchronization is monitored by the chaotic resonance detection mechanism, and the risk feature distribution is obtained by local similarity mapping, which includes:

[0019] The initial connection weight matrix of the chaotic neural network is constructed according to the correlation strength in the project knowledge association data; the input signal of the adjacent neurons and the initial connection weight matrix are input into the signal accumulation unit, the cumulative state value at the current moment is calculated according to the attenuation coefficient, the cumulative state value is processed by nonlinear iterative mapping and compared with the activation threshold, and when the cumulative state value is greater than the activation threshold, the neuron activation is triggered and enters the transient inhibition period; the positive activation energy and the negative inhibition energy are calculated and converted into a chaotic regulation signal, so that the chaotic neural network maintains a chaotic critical state;

[0020] Construct a chaotic adaptive adjustment module to obtain the activation timing information of adjacent neurons, calculate the weight adjustment increment and dynamically merge it with the initial connection weight matrix, and update the connection weight matrix; adjust the corresponding sensitivity threshold according to the activation state of the neuron, and adjust the signal transmission intensity based on the overall synchronization degree of the chaotic neuron network;

[0021] A chaotic resonance detection module is set up to calculate the synchronization degree based on the activation mode of neurons in the chaotic neural network, and when the synchronization degree is greater than a preset steady-state threshold, a feature extraction state is entered to generate a feature vector;

[0022] The local sensitive hashing algorithm is used to perform dimension reduction mapping on the feature vector to obtain the risk feature distribution of medical service items.

[0023] In an optional embodiment,

[0024] The deep generative adversarial diagnosis and treatment modeler is used to extract diagnosis and treatment feature sequences from project operation data and build a knowledge representation model. Based on the knowledge representation model, the heterogeneous medical network neural evolution device is used to analyze the correlation strength and time series characteristics between various diagnosis and treatment activities, and a risk prediction spectrum is generated after comparison with historical compliance records. Combined with the risk benchmark model, the risk prediction spectrum is dynamically evolved and analyzed through the medical value dynamics framework, and the real-time compliance analysis results are output, including:

[0025] Construct a deep generative adversarial diagnosis and treatment modeler, wherein a residual connection structure is set in the generator network of the deep generative adversarial diagnosis and treatment modeler, and a feature fusion gating unit is set in each connection layer of the residual connection structure, and the features of different levels are adaptively fused through the feature fusion gating unit to obtain fusion features; a plurality of convolution kernels of different scales are set in the discriminator network of the deep generative adversarial diagnosis and treatment modeler, and the fusion features are multi-scale decomposed by group convolution to obtain decomposed features; the operation data of medical service items are input into the deep generative adversarial diagnosis and treatment modeler, and a diagnosis and treatment feature sequence is extracted based on the decomposition feature, and the diagnosis and treatment feature sequence is grouped according to time series features, behavior features and resource features to construct a knowledge representation model;

[0026] The diagnosis and treatment feature sequence in the knowledge representation model is mapped to the feature nodes of the heterogeneous medical network, and the associated edges are constructed based on the temporal relationship, behavior association and resource dependency of the feature nodes; the node importance of the feature nodes is calculated as the starting probability of the random walk, and the temporal walk is executed to obtain the multi-hop walk path; the node transfer probability matrix is ​​calculated according to the multi-hop walk path, and the diagnosis and treatment activity association intensity matrix is ​​constructed;

[0027] The medical treatment activity association intensity matrix is ​​aligned with the historical compliance records in time series to obtain aligned data, and the aligned data is input into a time series convolution network with a dilated convolution structure to extract multi-scale time series features and calculate the risk propagation coefficient to generate a risk prediction spectrum; a medical value dynamics equation group is constructed based on the risk prediction spectrum, and a risk propagation function and a value loss function are set in the medical value dynamics equation group, the risk propagation function describes the diffusion process of risk between the feature nodes according to the risk propagation coefficient, and the value loss function quantifies the impact of risk on the value of medical services; the medical value dynamics equation group is solved to obtain a dynamic evolution trajectory of the risk state, and a real-time compliance analysis result is output based on the dynamic evolution trajectory.

[0028] In an optional embodiment,

[0029] Calculating the node importance of the feature node as the starting probability of random walk, and executing the time series walk to obtain the multi-hop walk path includes:

[0030] The comprehensive importance is calculated by integrating the community structure characteristics and historical activity sequence characteristics of the feature nodes as the starting probability of the random walk; in the process of executing the temporal walk, the transition probability is dynamically updated based on the local importance and temporal importance of the node, and the path sampling and skeleton extraction mechanism is introduced to generate a multi-hop walk path that meets the temporal transition constraints, including:

[0031] Obtaining a connection relationship matrix of the characteristic nodes, decomposing the connection relationship matrix using a spectral clustering method to obtain a characteristic vector matrix, identifying a node community structure based on the characteristic vector matrix; calculating the structural density between node pairs within the node community structure to obtain a density matrix, calculating the propagation capacity in combination with the node's in-and-out degree distribution to obtain a propagation matrix; calculating the local importance of the node based on the density matrix and the propagation matrix;

[0032] Collecting the historical activity sequence of the feature node, performing time series decomposition on the historical activity sequence to obtain periodic components, constructing a timing template based on the periodic components, matching the current activity sequence of the feature node with the timing template to obtain a matching degree distribution, and calculating the timing importance of the feature node according to the matching degree distribution;

[0033] The local importance and the temporal importance are input into a feature fusion network, the feature fusion network dynamically adjusts the fusion weight based on an error feedback mechanism, outputs the comprehensive importance of the feature node, and performs normalization to obtain a starting probability distribution of a random walk;

[0034] The forward transfer probability and the backward transfer probability between nodes are calculated according to the initial probability distribution, and a timing transfer probability matrix is ​​constructed based on the forward transfer probability and the backward transfer probability; a non-homogeneous timing walk strategy is initialized using the timing transfer probability matrix, and in each walk step, the timing transfer probability matrix is ​​updated based on the local importance and timing importance of the current node; the non-homogeneous timing walk strategy is executed to obtain an initial walk path, the initial walk path is segmented and sampled to obtain a sampling path set, timing key points are extracted from the sampling path set to determine a path skeleton, and a multi-hop walk path is generated based on the path skeleton, wherein the multi-hop walk path includes a node jump sequence that satisfies the timing transfer constraint.

[0035] In an optional embodiment,

[0036] Use causal reasoning decision trees to build a time-series evolution model of real-time compliance assessment results, calculate the risk evolution gradient, dynamically decompose the risk state matrix based on the risk evolution gradient, locate abnormal risk points, and generate risk warning information and risk intervention strategies based on the spatiotemporal distribution characteristics of abnormal risk points; feed back the spatiotemporal distribution characteristics, risk warning information and risk intervention strategies to the data association relationship matrix, and update the medical service project knowledge association data including:

[0037] Acquire risk status data in the medical service process and extract risk status features, construct causal reasoning decision tree nodes according to the hierarchical relationship of the risk status features, establish transfer connections between nodes based on the temporal change relationship of the risk status data, and generate a multi-level causal reasoning decision tree; calculate the state transfer weights between adjacent nodes in the multi-level causal reasoning decision tree, normalize the state transfer weights to obtain the state transfer probability, and construct a risk state transfer matrix;

[0038] Obtain the risk state transfer matrix of adjacent time steps, calculate the difference of the risk state transfer matrix, and obtain the risk evolution gradient according to the ratio of the difference to the time interval; perform singular value decomposition on the risk evolution gradient to obtain a feature matrix, and identify abnormal risk points of sudden changes in risk situation based on the singular values ​​in the feature matrix;

[0039] Extracting the time span and spatial distribution information of the abnormal risk point, constructing a spatiotemporal feature vector in combination with the mutation degree of the abnormal risk point, determining the risk warning level according to the spatiotemporal feature vector, and generating a risk intervention strategy based on the risk warning level and the spatiotemporal feature vector;

[0040] The spatiotemporal feature vector, the risk warning level and the risk intervention strategy are input into a feature fusion function, a data association relationship matrix is ​​updated according to an output result of the feature fusion function, and knowledge association data of medical service items is optimized based on the data association relationship matrix.

[0041] In an optional embodiment,

[0042] The risk status data in the medical service process is obtained and the risk status features are extracted. According to the hierarchical relationship of the risk status features, a causal reasoning decision tree node is constructed. Based on the temporal change relationship of the risk status data, a transfer connection between nodes is established. The generation of a multi-level causal reasoning decision tree includes:

[0043] By standardizing and reducing the dimension of medical service risk status data, using the autoencoder network to extract potential feature vectors and calculate the risk status similarity, clustering the risk status nodes based on similarity, establishing transfer connections between nodes based on the time series change relationship, and using the maximum information coefficient to optimize the connection structure of the decision tree, specifically including:

[0044] Acquire risk status data in the medical service process and extract risk status features, perform standardization on the risk status features to obtain standardized features, and perform dimensionality reduction on the standardized features using a principal component analysis method to obtain reduced-dimensionality risk status features;

[0045] Inputting the reduced-dimensionality risk state feature into an autoencoder network, wherein the encoder part of the autoencoder network maps the reduced-dimensionality risk state feature into a latent feature vector, and the decoder part of the autoencoder network reconstructs the reduced-dimensionality risk state feature based on the latent feature vector, and calculates the Euclidean distance between risk states according to the latent feature vector to obtain the risk state similarity;

[0046] A clustering operation is performed based on the risk state similarity, and the corresponding risk states are classified into a risk state cluster according to a preset similarity threshold, so as to obtain a plurality of risk state clusters at different levels, and each of the risk state clusters is mapped to a risk state node of a causal reasoning decision tree; based on the temporal changes of the risk states in the risk state clusters, the temporal correlation strength between the risk state nodes is calculated, and the temporal correlation strength is determined according to the jump frequency of the risk state in the time series;

[0047] Establishing a directed connection relationship between the risk status nodes according to the temporal association strength, establishing a transfer connection between risk status node pairs whose temporal association strength is greater than a preset strength threshold, and constructing an initial structure of a multi-level causal reasoning decision tree based on the risk status nodes and the transfer connections;

[0048] The maximum information coefficient method is used to calculate the causal relationship strength between the risk status nodes, and the initial structure of the multi-level causal reasoning decision tree is optimized based on the causal relationship strength. The transfer connections with causal relationship strength lower than the preset causal threshold are deleted, and the transfer connections with strength higher than the preset causal threshold are retained to obtain an optimized multi-level causal reasoning decision tree.

[0049] According to a second aspect of the embodiments of the present invention,

[0050] Provide a medical service project compliance dynamic monitoring system based on artificial intelligence, including:

[0051] The first unit is used to use the non-equilibrium entropy theory to dynamically map the medical service project data, establish a data association relationship matrix, and obtain project knowledge association data; based on the data association relationship matrix, a chaotic neural network is used to extract the risk feature distribution in the project knowledge association data; the Lyapunov stability analyzer is used to predict the long-term trend of the risk feature distribution and generate a risk benchmark model containing risk assessment weights and dynamic monitoring thresholds;

[0052] The second unit is used to use the deep generative adversarial diagnosis and treatment modeler to extract diagnosis and treatment feature sequences from project operation data and build a knowledge representation model; based on the knowledge representation model, a heterogeneous medical network neural evolution device is used to analyze the correlation strength and timing characteristics between various diagnosis and treatment activities, and a risk prediction spectrum is generated after comparison with historical compliance records. Combined with the risk benchmark model, the risk prediction spectrum is dynamically evolved and analyzed through the medical value dynamics framework, and real-time compliance analysis results are output;

[0053] The third unit is used to use causal reasoning decision trees to build a time-series evolution model of real-time compliance assessment results, calculate the risk evolution gradient, dynamically decompose the risk status matrix based on the risk evolution gradient, locate abnormal risk points, and generate risk warning information and risk intervention strategies based on the spatiotemporal distribution characteristics of abnormal risk points; feed back the spatiotemporal distribution characteristics, risk warning information and risk intervention strategies to the data association relationship matrix, and update the knowledge association data of medical service items.

[0054] According to a third aspect of the embodiments of the present invention,

[0055] An electronic device is provided, comprising:

[0056] processor;

[0057] a memory for storing processor-executable instructions;

[0058] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0059] A fourth aspect of the embodiments of the present invention is:

[0060] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0061] In an embodiment of the present invention, through dynamic association mapping and risk feature extraction, the compliance of medical service items can be monitored in real time, and the safety and effectiveness of medical services can be improved; by utilizing deep generative adversarial modeling and heterogeneous medical network analysis, the correlation and risks between diagnosis and treatment activities can be accurately identified, the allocation of medical resources can be optimized, and compliance risks can be reduced; through time-series evolution models and risk warning mechanisms, abnormal risk points can be discovered in a timely manner and corresponding intervention strategies can be formulated, thereby enhancing the resilience and continuous improvement capabilities of medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 This is a flow chart of a method for dynamic monitoring of compliance of medical service items based on artificial intelligence according to an embodiment of the present invention;

[0063] Figure 2This is a schematic diagram of the structure of a medical service item compliance dynamic monitoring system based on artificial intelligence according to an embodiment of the present invention. DETAILED DESCRIPTION

[0064] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0066] Figure 1 FIG. 1 is a flow chart of a method for dynamically monitoring compliance of medical service items based on artificial intelligence according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0067] S101. Use the non-equilibrium entropy theory to dynamically map the medical service project data, establish a data association relationship matrix, and obtain project knowledge association data; based on the data association relationship matrix, use a chaotic neural network to extract the risk feature distribution in the project knowledge association data; use the Lyapunov stability analyzer to predict the long-term trend of the risk feature distribution and generate a risk benchmark model containing risk assessment weights and dynamic monitoring thresholds;

[0068] In this embodiment, the data association matrix is ​​constructed through the non-equilibrium entropy theory, which can capture the complex dynamic relationships hidden in medical service items and improve data utilization efficiency; the chaotic neural network is used to accurately extract the risk feature distribution and enhance the detection ability of nonlinear and complex risk patterns; long-term trend prediction is achieved through Lyapunov stability analysis, and a benchmark model including risk weights and dynamic monitoring thresholds is constructed to provide timely warning and decision-making support for medical risk management.

[0069] S102. Use the deep generative adversarial diagnosis and treatment modeler to extract the diagnosis and treatment feature sequence from the project operation data and build a knowledge representation model; based on the knowledge representation model, use the heterogeneous medical network neural evolution device to analyze the correlation strength and time series characteristics between various diagnosis and treatment activities, generate a risk prediction spectrum after comparing with historical compliance records, combine the risk benchmark model, and dynamically analyze the risk prediction spectrum through the medical value dynamics framework to output real-time compliance analysis results;

[0070] In this embodiment, a deep generative adversarial diagnosis and treatment modeler is used to extract diagnosis and treatment feature sequences from project data, and a unified knowledge representation model is constructed, thereby improving the ability to capture and express key diagnosis and treatment information; a heterogeneous medical network neural evolver is used to deeply analyze the correlation strength and timing characteristics between various diagnosis and treatment activities, which can more comprehensively reflect the complex interactive relationship between medical activities; a risk prediction spectrum is generated in combination with historical compliance records, and connected to a risk benchmark model, thereby realizing a dynamic evolution analysis of medical risks and providing forward-looking data support for risk warning; and finally, real-time compliance analysis results are output through the medical value dynamics framework, which helps to detect anomalies in a timely manner and ensure the standardization and safety of medical processes.

[0071] S103. Use the causal reasoning decision tree to build a time-series evolution model of real-time compliance assessment results, calculate the risk evolution gradient, dynamically decompose the risk status matrix based on the risk evolution gradient, locate abnormal risk points, and generate risk warning information and risk intervention strategies based on the spatiotemporal distribution characteristics of abnormal risk points; feed back the spatiotemporal distribution characteristics, risk warning information and risk intervention strategies to the data association relationship matrix, and update the knowledge association data of medical service items.

[0072] In this embodiment, the time-series evolution model constructed through the causal reasoning decision tree realizes real-time tracking of compliance assessment results and capture of risk evolution trends, thereby enhancing the dynamic risk monitoring capability; the risk state matrix is ​​dynamically decomposed using the risk evolution gradient, which can accurately locate abnormal risk points and provide a clear basis for early warning and intervention; risk warning information and intervention strategies are automatically generated based on the spatiotemporal distribution characteristics of abnormal risk points, thereby improving the response speed and accuracy of risk management; the spatiotemporal distribution characteristics, warning information and intervention strategies are fed back to the data association relationship matrix, thereby realizing the dynamic update of medical service project knowledge association data and promoting the continuous optimization and upgrading of the risk management system.

[0073] In an optional implementation, non-equilibrium entropy theory is used to dynamically associate and map medical service project data, establish a data association relationship matrix, and obtain project knowledge association data; based on the data association relationship matrix, a chaotic neural network is used to extract the risk feature distribution in the project knowledge association data; a Lyapunov stability analyzer is used to predict the long-term trend of the risk feature distribution, and a risk benchmark model containing risk assessment weights and dynamic monitoring thresholds is generated, including:

[0074] The medical service project data is constructed into an initial data set according to the service type, execution time, operator, service object, and operation record; the non-equilibrium entropy value between any two data items is obtained according to the probability distribution of each data item in the initial data set on the corresponding dimension, and the association strength between the data items is calculated according to the non-equilibrium entropy value and the smoothing factor to construct a data association relationship matrix; a time window mechanism is introduced into the data association relationship matrix, and the data association relationship matrix of the current time window and the historical data association relationship matrix are weightedly combined through the forgetting factor to generate project knowledge association data of the medical service project;

[0075] The association strength in the project knowledge association data is used as the neuron connection weight to construct a chaotic neural network. The nonlinear iterative mapping is used to process the neuron input signal to keep the chaotic neural network in a chaotic critical state. The output of the chaotic neural network is reduced in dimension by the local sensitive hashing algorithm to obtain the risk characteristic distribution of medical service items.

[0076] The risk characteristic distribution is converted into a dynamic system equation, and the Lyapunov index of the dynamic system equation in each characteristic dimension is calculated, and the Lyapunov index represents the degree of trajectory divergence of the characteristic dimension; the Lyapunov index is subjected to exponential transformation and normalization to generate a risk assessment weight; the historical data statistical value of each characteristic dimension is combined with the Lyapunov index of the corresponding dimension to generate a dynamic monitoring threshold, and the risk assessment weight and the dynamic monitoring threshold are combined to construct a risk benchmark model for medical service projects.

[0077] In a specific implementation, first, an initial data set of medical service items is constructed. Data related to medical service items are collected, including information such as service type, execution time, operator, service object, and detailed operation records. For example, all surgical records of a hospital in a year can be collected, including surgical type, operation time, surgeon, anesthesiologist, patient information, and various indicator records during the operation.

[0078] Next, the association strength between the data items is calculated and a data association relationship matrix is ​​constructed. According to the probability distribution of each data item in the initial data set in different dimensions, the non-equilibrium entropy value between any two data items is calculated. The non-equilibrium entropy value reflects the degree of difference between two data items. For example, the non-equilibrium entropy value between two surgical records of the same type of surgery will be relatively small, while the non-equilibrium entropy value between two surgical records of different types of surgery will be relatively large. Then, based on the non-equilibrium entropy value and the preset smoothing factor, the association strength between the data items is calculated. The smoothing factor is used to adjust the sensitivity of the association strength. The higher the association strength, the stronger the association between the two data items. The association strengths between all data items are combined into a matrix, namely the data association relationship matrix.

[0079] Introduce the time window mechanism to generate project knowledge association data. In order to capture the dynamic association relationship between medical service project data, the time window mechanism is introduced. The data association relationship matrix of the current time window is weighted and combined with the historical data association relationship matrix. For example, the association relationship matrix of the last month can be weighted and combined with the association relationship matrix of the past three months. The weight can be set according to the time distance. The closer the time, the greater the weight. The weighted combination matrix is ​​the project knowledge association data of the medical service project.

[0080] Construct a chaotic neural network and extract the risk feature distribution. Use the correlation strength in the project knowledge association data as the connection weight between neurons to construct a chaotic neural network. Input the medical service project data into the chaotic neural network, and use nonlinear iterative mapping to process the neuron input signal to keep the chaotic neural network in a chaotic critical state. The neural network in a chaotic critical state has higher sensitivity and information processing capabilities. Then, use the local sensitive hashing algorithm to reduce the dimension of the output of the chaotic neural network to obtain the risk feature distribution of the medical service project. For example, the high-dimensional output data can be reduced to two or three dimensions for subsequent analysis and processing.

[0081] Perform Lyapunov stability analysis to generate risk assessment weights and dynamic monitoring thresholds. Convert risk feature distribution into a dynamic system equation and calculate the Lyapunov index of the dynamic system equation on each feature dimension. The Lyapunov index characterizes the divergence of the trajectory on the feature dimension and can be used to assess the long-term trend of risk. Perform exponential transformation and normalization on the Lyapunov index to generate risk assessment weights. For example, the Lyapunov index can be normalized after taking the exponent to obtain the risk assessment weight for each feature dimension. Combine the historical data statistics of each feature dimension with the corresponding Lyapunov index to generate a dynamic monitoring threshold. For example, the dynamic monitoring threshold can be set based on the mean and standard deviation of the historical data combined with the Lyapunov index.

[0082] Finally, a risk benchmark model for medical service projects is constructed. The risk assessment weight and dynamic monitoring threshold are combined to construct a risk benchmark model for medical service projects. This model can be used to assess the risk level of medical service projects and monitor risks in real time.

[0083] In this embodiment, through non-equilibrium entropy theory and chaotic neural network, the risk characteristics in the medical service project data can be captured more comprehensively and the accuracy of risk assessment can be improved; through the time window mechanism and Lyapunov stability analysis, dynamic monitoring of medical service project risks can be achieved and potential risks can be discovered in time; the constructed risk benchmark model can provide medical institutions with decision-making support for risk management and help them formulate effective risk prevention and control measures.

[0084] In an optional implementation, the association strength in the project knowledge association data is used as the neuron connection weight to construct a chaotic neural network, and a nonlinear iterative mapping is used to process the neuron input signal so that the chaotic neural network maintains a chaotic critical state. The output of the chaotic neural network is reduced in dimension by a local sensitive hashing algorithm, and the risk characteristic distribution of the medical service items is obtained, including:

[0085] The correlation strength of project knowledge association data is constructed as the neuron connection weight, and the chaotic critical state of the network is controlled by nonlinear iterative mapping. The weight is dynamically updated through the chaotic adaptive adjustment mechanism, the network synchronization is monitored by the chaotic resonance detection mechanism, and the risk feature distribution is obtained by local similarity mapping, which includes:

[0086] The initial connection weight matrix of the chaotic neural network is constructed according to the correlation strength in the project knowledge association data; the input signal of the adjacent neurons and the initial connection weight matrix are input into the signal accumulation unit, the cumulative state value at the current moment is calculated according to the attenuation coefficient, the cumulative state value is processed by nonlinear iterative mapping and compared with the activation threshold, and when the cumulative state value is greater than the activation threshold, the neuron activation is triggered and enters the transient inhibition period; the positive activation energy and the negative inhibition energy are calculated and converted into a chaotic regulation signal, so that the chaotic neural network maintains a chaotic critical state;

[0087] Construct a chaotic adaptive adjustment module to obtain the activation timing information of adjacent neurons, calculate the weight adjustment increment and dynamically merge it with the initial connection weight matrix, and update the connection weight matrix; adjust the corresponding sensitivity threshold according to the activation state of the neuron, and adjust the signal transmission intensity based on the overall synchronization degree of the chaotic neuron network;

[0088] A chaotic resonance detection module is set up to calculate the synchronization degree based on the activation mode of neurons in the chaotic neural network, and when the synchronization degree is greater than a preset steady-state threshold, a feature extraction state is entered to generate a feature vector;

[0089] The local sensitive hashing algorithm is used to perform dimension reduction mapping on the feature vector to obtain the risk feature distribution of medical service items.

[0090] In a specific implementation, the construction process of the chaotic neural network first needs to extract detailed association data between items from the medical service database. These data contain information such as the order of diagnosis and treatment operations, drug compatibility rules, and treatment plan combinations. The system calculates the association frequency between each pair of items and converts it into a standardized association strength value, which is filled into the initial connection weight matrix. After the weight matrix is ​​normalized, all weight values ​​are mapped to the range of zero to one as the initial connection strength of the neural network.

[0091] In the signal processing stage, each neuron continuously receives input signals from adjacent neurons, which are multiplied by the corresponding connection weights to obtain weighted inputs. The signal accumulation unit integrates these weighted inputs over time and introduces an attenuation factor to simulate the actual signal attenuation process. When the accumulated value exceeds the preset activation threshold, the neuron generates an output pulse and immediately enters the refractory period. The output pulse is transmitted to adjacent neurons through synaptic connections. The system records the activation time and pulse interval information of each neuron in detail for subsequent analysis.

[0092] In order to maintain the dynamic balance of the network, the system calculates the average activation frequency of all neurons in the network and the degree of synchronous activation of the neuron group in real time. Based on these indicators, the system adjusts the chaos control parameters and uses positive and negative feedback mechanisms to keep the network in a chaotic critical state. This state not only ensures the sensitivity of the network to the input signal, but also avoids the network from falling into a local optimum.

[0093] In the dynamic weight adjustment phase, the system continuously records the activation timing of each pair of adjacent neurons, and extracts pairs of neurons with significant timing correlation by calculating the correlation between the activation timings. For these highly correlated pairs of neurons, the system strengthens the connection weights between them. The weight update uses the Heb learning rule to calculate the adjustment amount. By setting the appropriate adjustment step size and learning rate, the adjustment amount is weighted and fused with the original weight, thereby achieving dynamic update of the weight matrix.

[0094] At the same time, the system monitors the changes in the activation frequency of each neuron and dynamically adjusts the sensitivity threshold of each neuron based on these changes. For those neurons that are frequently activated, the system will increase their sensitivity accordingly; while for neurons that are inactive for a long time, their sensitivity will be reduced. This adaptive mechanism ensures the network's ability to respond quickly to abnormal patterns.

[0095] In the feature extraction stage, the system evaluates the overall synchronization of the network by calculating the average phase difference of the neuron group and the temporal clustering of activation. When the synchronization reaches the preset conditions, the system begins to extract the activation pattern characteristics of the neurons, including the weight distribution of synaptic connections and the spatial correlation characteristics of the neuron group, and combines these features to form a high-dimensional feature vector. Finally, the feature vector is reduced in dimension by constructing a family of locally sensitive hash functions to generate a low-dimensional risk feature representation, and finally output a risk distribution map.

[0096] For example, the antibiotic use monitoring of a tertiary hospital is used as a specific application scenario: the system first collects all medical data related to the use of antibiotics, including doctor's order records, pharmacy dispensing records, nursing execution records, etc. By analyzing these data, the system constructs a neural network containing twenty key monitoring points and sets the initial connection weights based on business relevance. When the system detects frequent modifications to antibiotic orders at night, the activation frequency of the corresponding neurons increases significantly, triggering the weight adaptive adjustment mechanism. The system then increases the monitoring sensitivity of the doctor's order modification link and strengthens the connection weight with the doctor's order review link. By tracking the execution of modified orders, the system can analyze abnormal patterns of drug consumption and detect time series changes in nursing execution records. Finally, the system generates a detailed antibiotic use compliance risk distribution map based on the time synchronization, spatial aggregation, and continuity of process connection of each link, and accurately marks high-risk time periods and operation links.

[0097] In this embodiment, the initial weight matrix is ​​constructed using project knowledge association data, which provides a real dynamic association basis for the neural network, thereby more accurately reflecting the complex interactive relationship between medical projects; real-time monitoring of the neuron activation state is achieved through signal accumulation units and nonlinear iterative mapping, ensuring that when the accumulated signal exceeds the threshold, the activation is quickly triggered and transient inhibition is entered, thereby improving the system's sensitivity and response speed to risk characteristics; positive activation energy and negative inhibition energy are calculated and converted into chaos regulation signals, effectively keeping the network in a chaotic critical state, and enhancing the system's ability to process complex, nonlinear risk information; the chaos adaptive regulation module automatically adjusts the connection weights and sensitivity thresholds according to the activation timing information, so that the model has adaptive adjustment functions and improves overall robustness and adaptability; through chaos resonance detection and local sensitive hashing algorithms, the system can extract key features when the neurons reach a steady state synchronously, and perform dimensionality reduction mapping on high-dimensional feature vectors, thereby obtaining a clear risk feature distribution, providing a reliable basis for subsequent risk assessment

[0098] In an optional implementation, a deep generative adversarial diagnosis and treatment modeler is used to extract diagnosis and treatment feature sequences from project operation data to construct a knowledge representation model. Based on the knowledge representation model, a heterogeneous medical network neural evolution device is used to analyze the correlation strength and time series characteristics between various diagnosis and treatment activities, and a risk prediction spectrum is generated after comparison with historical compliance records. Combined with the risk benchmark model, the risk prediction spectrum is dynamically evolved and analyzed through the medical value dynamics framework, and the real-time compliance analysis results are output, including:

[0099] Construct a deep generative adversarial diagnosis and treatment modeler, wherein a residual connection structure is set in the generator network of the deep generative adversarial diagnosis and treatment modeler, and a feature fusion gating unit is set in each connection layer of the residual connection structure, and the features of different levels are adaptively fused through the feature fusion gating unit to obtain fusion features; a plurality of convolution kernels of different scales are set in the discriminator network of the deep generative adversarial diagnosis and treatment modeler, and the fusion features are multi-scale decomposed by group convolution to obtain decomposed features; the operation data of medical service items are input into the deep generative adversarial diagnosis and treatment modeler, and a diagnosis and treatment feature sequence is extracted based on the decomposition feature, and the diagnosis and treatment feature sequence is grouped according to time series features, behavior features and resource features to construct a knowledge representation model;

[0100] The diagnosis and treatment feature sequence in the knowledge representation model is mapped to the feature nodes of the heterogeneous medical network, and the associated edges are constructed based on the temporal relationship, behavior association and resource dependency of the feature nodes; the node importance of the feature nodes is calculated as the starting probability of the random walk, and the temporal walk is executed to obtain the multi-hop walk path; the node transfer probability matrix is ​​calculated according to the multi-hop walk path, and the diagnosis and treatment activity association intensity matrix is ​​constructed;

[0101] The medical treatment activity association intensity matrix is ​​aligned with the historical compliance records in time series to obtain aligned data, and the aligned data is input into a time series convolution network with a dilated convolution structure to extract multi-scale time series features and calculate the risk propagation coefficient to generate a risk prediction spectrum; a medical value dynamics equation group is constructed based on the risk prediction spectrum, and a risk propagation function and a value loss function are set in the medical value dynamics equation group, the risk propagation function describes the diffusion process of risk between the feature nodes according to the risk propagation coefficient, and the value loss function quantifies the impact of risk on the value of medical services; the medical value dynamics equation group is solved to obtain a dynamic evolution trajectory of the risk state, and a real-time compliance analysis result is output based on the dynamic evolution trajectory.

[0102] In a specific implementation, first, a deep generative adversarial diagnosis and treatment modeler is constructed. The modeler consists of two parts: a generator and a discriminator. A residual connection structure is set in the generator network, and a feature fusion gating unit is set in each connection layer. The function of the generator is to convert a random noise vector into a simulated diagnosis and treatment feature sequence. The residual connection structure helps to train deeper networks, while the feature fusion gating unit can adaptively fuse features at different levels, thereby improving the quality of the generated sequence. For example, the gating unit can learn a weight to control the fusion ratio between the features of the current layer and the features of the previous layer based on the features of the current layer and the features of the previous layer.

[0103] Multiple convolution kernels of different scales are set in the discriminator network, and the fusion features output by the generator are decomposed at multiple scales by grouped convolution. The role of the discriminator is to distinguish between the real diagnosis and treatment feature sequence and the simulated sequence generated by the generator. Multi-scale convolution kernels can capture features of different granularities, while grouped convolution can reduce the amount of calculation and improve the generalization ability of the model. For example, convolution kernels of sizes 3×3, 5×5, and 7×7 can be used to extract features of different scales.

[0104] The real medical service project operation data is input into the deep generative adversarial diagnosis and treatment modeler for training. Through adversarial training, the generator and discriminator continuously improve their respective capabilities, and finally the generator can generate a diagnosis and treatment feature sequence that is indistinguishable from the real thing. After the training is completed, the discriminator is used to extract the decomposition features, and the diagnosis and treatment feature sequence is extracted based on the decomposition features.

[0105] The extracted diagnosis and treatment feature sequences are grouped according to time series features, behavior features, and resource features to construct a knowledge representation model. For example, time series features can include the time and duration of diagnosis and treatment activities; behavior features can include specific behaviors such as diagnosis, treatment, and examination; and resource features can include the drugs, equipment, and human resources used.

[0106] Heterogeneous medical network neural evolution device analyzes the correlation between diagnosis and treatment activities:

[0107] The diagnosis and treatment feature sequences in the knowledge representation model are mapped to feature nodes of the heterogeneous medical network. For example, each diagnosis and treatment activity can be used as a node, and the attributes of the node can include time series features, behavior features, and resource features.

[0108] The associated edges are constructed based on the temporal relationship, behavior association and resource dependency of the feature nodes. For example, if two diagnosis and treatment activities are closely connected in time, an edge can be established between them; if two diagnosis and treatment activities use the same resources, an edge can also be established between them.

[0109] Calculate the node importance of the feature node as the starting probability of the random walk. The node importance can be calculated based on the node degree, centrality and other indicators. Perform a time-series walk to obtain a multi-hop walk path. For example, you can start from an important node, perform a random walk along the edges in the network, and record the node sequence you pass through.

[0110] The node transition probability matrix is ​​calculated based on the multi-hop wandering path, and the diagnosis and treatment activity association strength matrix is ​​constructed. For example, if two nodes often appear together in the multi-hop wandering path, the association strength between them is high.

[0111] Risk prediction spectrum generation and medical value dynamics framework analysis:

[0112] Align the time sequence of the medical treatment activity association intensity matrix with the historical compliance records. For example, you can match the current medical treatment activity sequence with the sequence in the historical compliance records to find similar sequences.

[0113] The aligned data is input into a temporal convolutional network with a dilated convolutional structure. Dilated convolution can expand the receptive field to capture longer-term temporal dependencies. Multi-scale temporal features are extracted and the risk propagation coefficient is calculated. The risk propagation coefficient can represent the possibility of risk propagation between different diagnosis and treatment activities.

[0114] Generate a risk prediction spectrum. The risk prediction spectrum can indicate the risk level of different diagnosis and treatment activities.

[0115] The medical value dynamics equation group is constructed based on the risk prediction spectrum. The risk propagation function and the value loss function are set in the equation group. The risk propagation function describes the diffusion process of risk between characteristic nodes according to the risk propagation coefficient. The value loss function quantifies the impact of risk on the value of medical services. For example, if the risk of a diagnosis and treatment activity is high, it may reduce the value of the entire medical service.

[0116] Solving the medical value dynamics equations can obtain the dynamic evolution trajectory of the risk state. For example, the process of risk propagation in the network can be simulated to predict the risk level of different diagnosis and treatment activities in the future.

[0117] Output real-time compliance analysis results based on dynamic evolution trajectories. For example, it can determine whether the current diagnosis and treatment plan is compliant based on the risk status and give corresponding suggestions.

[0118] Data case: Assume that a patient undergoes three medical treatments: A, B, and C. A and B are closely connected in time and use the same resources; C has no direct association with A and B. The deep generative adversarial medical modeler is used to extract the medical feature sequence and build a knowledge representation model. Then, A, B, and C are mapped as nodes of a heterogeneous medical network, and an edge is established between A and B. The node transition probability matrix is ​​calculated through random walks, and the medical activity association strength matrix is ​​constructed. Finally, the association strength matrix is ​​compared with the historical compliance records to generate a risk prediction spectrum.

[0119] In this embodiment, by extracting diagnosis and treatment characteristics through a deep learning model and combining heterogeneous network analysis and the medical value dynamics framework, the risks of diagnosis and treatment activities can be predicted more accurately and the accuracy of compliance analysis can be improved; the compliance of diagnosis and treatment activities can be analyzed in real time, and risk warnings can be given in a timely manner, which helps medical institutions take timely measures to avoid compliance risks; the risk transmission path and impact degree between diagnosis and treatment activities can be clearly demonstrated, the interpretability of compliance analysis can be improved, and medical institutions can better understand the causes of risks.

[0120] In an optional implementation, calculating the node importance of the feature node as the starting probability of the random walk, and performing the time series walk to obtain the multi-hop walk path includes:

[0121] The comprehensive importance is calculated by integrating the community structure characteristics and historical activity sequence characteristics of the feature nodes as the starting probability of the random walk; in the process of executing the temporal walk, the transition probability is dynamically updated based on the local importance and temporal importance of the node, and the path sampling and skeleton extraction mechanism is introduced to generate a multi-hop walk path that meets the temporal transition constraints, including:

[0122] Obtaining a connection relationship matrix of the characteristic nodes, decomposing the connection relationship matrix using a spectral clustering method to obtain a characteristic vector matrix, identifying a node community structure based on the characteristic vector matrix; calculating the structural density between node pairs within the node community structure to obtain a density matrix, calculating the propagation capacity in combination with the node's in-and-out degree distribution to obtain a propagation matrix; calculating the local importance of the node based on the density matrix and the propagation matrix;

[0123] Collecting the historical activity sequence of the feature node, performing time series decomposition on the historical activity sequence to obtain periodic components, constructing a timing template based on the periodic components, matching the current activity sequence of the feature node with the timing template to obtain a matching degree distribution, and calculating the timing importance of the feature node according to the matching degree distribution;

[0124] The local importance and the temporal importance are input into a feature fusion network, the feature fusion network dynamically adjusts the fusion weight based on an error feedback mechanism, outputs the comprehensive importance of the feature node, and performs normalization to obtain a starting probability distribution of a random walk;

[0125] The forward transfer probability and the backward transfer probability between nodes are calculated according to the initial probability distribution, and a timing transfer probability matrix is ​​constructed based on the forward transfer probability and the backward transfer probability; a non-homogeneous timing walk strategy is initialized using the timing transfer probability matrix, and in each walk step, the timing transfer probability matrix is ​​updated based on the local importance and timing importance of the current node; the non-homogeneous timing walk strategy is executed to obtain an initial walk path, the initial walk path is segmented and sampled to obtain a sampling path set, timing key points are extracted from the sampling path set to determine a path skeleton, and a multi-hop walk path is generated based on the path skeleton, wherein the multi-hop walk path includes a node jump sequence that satisfies the timing transfer constraint.

[0126] In a specific implementation, a characteristic node temporal walk method is used to mine dynamic association patterns of nodes in a complex network. The method integrates static structural features and dynamic temporal behavior features of nodes to generate a multi-hop walk path with time constraints.

[0127] First, obtain the connection relationship of the feature nodes. For example, the connection relationship between nodes can be represented by an adjacency matrix. Assuming there are 5 feature nodes, the adjacency matrix can be represented as a 5×5 matrix, where the value of a matrix element is 1, indicating that there is a connection between two nodes, and the value is 0, indicating that there is no connection.

[0128] Then, the connection relationship matrix is ​​decomposed using the spectral clustering method to obtain the eigenvector matrix. The spectral clustering method can divide the nodes into different communities. Assume that the above five nodes are divided into two communities, {1, 2, 3} and {4, 5}, by spectral clustering.

[0129] Within each community, the structural density between node pairs is calculated to obtain a density matrix. The structural density can be used to measure the closeness of the connection between nodes, for example, it can be expressed by the number of common neighbors between node pairs. At the same time, the propagation capacity is calculated based on the in-and-out degree distribution of the nodes to obtain a propagation matrix. The propagation capacity can be used to measure the ability of a node to spread information in the network, for example, it can be expressed by the degree centrality of the node. Combining the density matrix and the propagation matrix, the local importance of the node is calculated.

[0130] Next, collect the historical activity sequence of the feature node. For example, you can collect the activity data of each node in the past period of time, such as the number of visits per hour. Perform time series decomposition on the historical activity sequence. For example, you can decompose the time series into trend components, seasonal components, and residual components. Construct a timing template based on periodic components, such as seasonal components. Match the current activity sequence of the feature node with the timing template to obtain a matching degree distribution. For example, you can use a dynamic time warping algorithm to calculate the similarity between the current activity sequence and the timing template. Based on the matching degree distribution, calculate the timing importance of the feature node.

[0131] The local importance and temporal importance are input into the feature fusion network. The network can be a multi-layer perceptron, which dynamically adjusts the fusion weights through the error feedback mechanism and outputs the comprehensive importance of the feature nodes. For example, the back propagation algorithm can be used to adjust the weights of the network. The comprehensive importance is normalized to obtain the starting probability distribution of the random walk.

[0132] According to the initial probability distribution, the forward transition probability and backward transition probability between nodes are calculated, and the timing transition probability matrix is ​​constructed. The forward transition probability represents the probability of transferring from one node to another node, and the backward transition probability represents the probability of returning from one node to the previous node. The non-homogeneous timing walk strategy is initialized using the timing transition probability matrix. In each walk step, the timing transition probability matrix is ​​updated based on the local importance and timing importance of the current node. For example, the transition probability can be weighted and adjusted according to the local importance and timing importance of the current node. The non-homogeneous timing walk strategy is executed to obtain the initial walk path.

[0133] The initial wandering path is sampled in segments to obtain a set of sampled paths. For example, the path can be sampled at fixed time intervals. Timing key points, such as nodes with high local importance or timing importance, are extracted from the set of sampled paths. A path skeleton is determined based on the timing key points. A multi-hop wandering path is generated based on the path skeleton, which contains a node jump sequence that satisfies the timing transfer constraint.

[0134] In this embodiment, the community structure characteristics and historical activity sequence characteristics of the nodes are integrated, which can more accurately characterize the importance of the nodes, thereby improving the accuracy of the generated wandering paths and more effectively capturing the dynamic association patterns between nodes; the transition probability is dynamically updated during the wandering process, and the path sampling and skeleton extraction mechanisms are introduced to generate multi-hop wandering paths that meet the timing transition constraints, better reflecting the dynamic evolution process of the nodes in the network; by extracting timing key points, the evolution process of the path can be more clearly displayed, the interpretability of the path is improved, and it helps users understand the dynamic association patterns of nodes in the network.

[0135] In an optional implementation, a causal reasoning decision tree is used to construct a temporal evolution model of real-time compliance assessment results, the risk evolution gradient is calculated, the risk state matrix is ​​dynamically decomposed based on the risk evolution gradient, abnormal risk points are located, and risk warning information and risk intervention strategies are generated according to the spatiotemporal distribution characteristics of abnormal risk points; the spatiotemporal distribution characteristics, risk warning information and risk intervention strategies are fed back to the data association relationship matrix, and the medical service item knowledge association data is updated, including:

[0136] Acquire risk status data in the medical service process and extract risk status features, construct causal reasoning decision tree nodes according to the hierarchical relationship of the risk status features, establish transfer connections between nodes based on the temporal change relationship of the risk status data, and generate a multi-level causal reasoning decision tree; calculate the state transfer weights between adjacent nodes in the multi-level causal reasoning decision tree, normalize the state transfer weights to obtain the state transfer probability, and construct a risk state transfer matrix;

[0137] Obtain the risk state transfer matrix of adjacent time steps, calculate the difference of the risk state transfer matrix, and obtain the risk evolution gradient according to the ratio of the difference to the time interval; perform singular value decomposition on the risk evolution gradient to obtain a feature matrix, and identify abnormal risk points of sudden changes in risk situation based on the singular values ​​in the feature matrix;

[0138] Extracting the time span and spatial distribution information of the abnormal risk point, constructing a spatiotemporal feature vector in combination with the mutation degree of the abnormal risk point, determining the risk warning level according to the spatiotemporal feature vector, and generating a risk intervention strategy based on the risk warning level and the spatiotemporal feature vector;

[0139] The spatiotemporal feature vector, the risk warning level and the risk intervention strategy are input into a feature fusion function, a data association relationship matrix is ​​updated according to an output result of the feature fusion function, and knowledge association data of medical service items is optimized based on the data association relationship matrix.

[0140] In a specific implementation, first, the system collects risk status data from all aspects of medical services, including all-round information such as patient diagnosis and treatment process, execution of medical orders, drug use, nursing records, etc. After data cleaning and standardization, the risk status features in these data are extracted, such as abnormal operation time, dosage deviation, execution delay, etc. The system analyzes the hierarchical dependencies between these risk status features and constructs the initial causal reasoning decision tree nodes. Subsequently, based on the evolution law of risk status data in the time dimension, the transfer connection relationship between nodes is established, and finally a causal reasoning decision tree structure containing multiple levels is formed.

[0141] After the decision tree is built, the system calculates the state transition weights between each pair of adjacent nodes in detail. This weight value reflects the tendency of one risk state to change to another risk state. By normalizing all transfer weights, the standardized state transition probability is obtained, and these probability values ​​are organized into a risk state transition matrix. This matrix fully records all possible risk state transition paths in the system and their occurrence probabilities.

[0142] The system continuously monitors the risk state transfer matrix in adjacent time windows, calculates the difference between the matrices at two time points, and divides the difference by the corresponding time interval to obtain the risk evolution gradient that reflects the risk change rate. Perform a singular value decomposition operation on this gradient matrix to obtain a feature matrix containing the main feature patterns. By analyzing the size and distribution of singular values ​​in the feature matrix, the system can accurately identify abnormal risk points where the risk situation has changed suddenly.

[0143] For each identified abnormal risk point, the system extracts the time span information of its occurrence in detail, including the start time, duration, periodic characteristics, etc., and analyzes its spatial distribution in different medical service links. The information of these time and space dimensions is combined with the degree index of risk mutation to construct a complete spatiotemporal feature vector. Based on the performance of each dimension of this feature vector, the system determines the corresponding risk warning level and formulates targeted risk intervention strategies based on the warning level and specific spatiotemporal characteristics.

[0144] Finally, the system inputs the three types of information, namely, spatiotemporal feature vectors, risk warning levels, and risk intervention strategies, into the feature fusion function for integration processing. The fusion results are used to update the data association matrix, thereby optimizing the knowledge association data of medical service items and forming a closed-loop risk management mechanism.

[0145] Specific application example: Take the anesthesia medication management in the operating room of a hospital as an example:

[0146] The system first collects all data related to surgical anesthesia, including medication orders issued by anesthesiologists, nurse execution records, and patient vital signs monitoring data. By analyzing this data, the system identifies multiple risk status characteristics, such as abnormal anesthetic drug dosage, improper medication intervals, and intraoperative vital sign fluctuations. Based on these characteristics, the system constructs a multi-level causal reasoning decision tree, with the bottom-level nodes representing specific risk manifestations, the middle-level nodes representing risk aggregation states, and the top-level nodes representing the overall risk level.

[0147] The system monitors the changes of various indicators during the anesthesia process in real time and calculates the transition probability between risk states. For example, when it is found that the drug dosage adjustment occurs repeatedly during the operation, the system will calculate the probability of switching from the normal drug administration state to the abnormal dosage state. By continuously monitoring the data for a week, the system found that the transition probability of abnormal dosage suddenly increased within a certain period of time, and confirmed that this was an abnormal risk point through singular value decomposition analysis.

[0148] The system further analyzed and found that this abnormal risk point mainly appeared during the night shift, concentrated in a few operating rooms, and involved a few specific medical staff. Based on these spatiotemporal characteristics, the system set the risk warning level to a higher level and generated specific intervention strategies, including recommendations to strengthen night shift staff training and adjust the operating room scheduling mechanism.

[0149] These findings and suggestions are input into the feature fusion function for integration and used to update the knowledge-related data of anesthesia medication management. The updated related data show that it is necessary to strengthen the supervision of night shift anesthesia operations and optimize the staffing plan to reduce the probability of similar risks. The whole process forms a complete closed loop from risk discovery to intervention improvement, continuously improving the safety of anesthesia medication management.

[0150] In this embodiment, risk status data can be obtained in real time from the medical service process, and key risk features can be extracted to achieve real-time monitoring of risk status and changing trends; by constructing a multi-level causal reasoning decision tree and a state transition matrix, the hierarchical relationship and time-series evolution between risk states are systematically revealed, providing an accurate quantitative description of risk changes; using risk evolution gradients and singular value decomposition, the mutation points of risk situations can be effectively captured, abnormal risk states can be discovered in a timely manner, and the sensitivity of early warning responses can be improved; the time span and spatial distribution information of abnormal risk points can be extracted to generate spatiotemporal feature vectors, thereby determining the warning level and formulating targeted intervention strategies to enhance the pertinence and effectiveness of risk disposal; the data association relationship matrix is ​​updated through feature fusion feedback, and the knowledge association data of medical service items is continuously optimized to achieve dynamic adaptation and continuous improvement of risk assessment models.

[0151] In an optional implementation, risk status data in the medical service process is obtained and risk status features are extracted, causal reasoning decision tree nodes are constructed according to the hierarchical relationship of the risk status features, and transfer connections between nodes are established based on the temporal change relationship of the risk status data. Generating a multi-level causal reasoning decision tree includes:

[0152] By standardizing and reducing the dimension of medical service risk status data, using the autoencoder network to extract potential feature vectors and calculate the risk status similarity, clustering the risk status nodes based on similarity, establishing transfer connections between nodes based on the time series change relationship, and using the maximum information coefficient to optimize the connection structure of the decision tree, specifically including:

[0153] Acquire risk status data in the medical service process and extract risk status features, perform standardization on the risk status features to obtain standardized features, and perform dimensionality reduction on the standardized features using a principal component analysis method to obtain reduced-dimensionality risk status features;

[0154] Inputting the reduced-dimensionality risk state feature into an autoencoder network, wherein the encoder part of the autoencoder network maps the reduced-dimensionality risk state feature into a latent feature vector, and the decoder part of the autoencoder network reconstructs the reduced-dimensionality risk state feature based on the latent feature vector, and calculates the Euclidean distance between risk states according to the latent feature vector to obtain the risk state similarity;

[0155] A clustering operation is performed based on the risk state similarity, and the corresponding risk states are classified into a risk state cluster according to a preset similarity threshold, so as to obtain a plurality of risk state clusters at different levels, and each of the risk state clusters is mapped to a risk state node of a causal reasoning decision tree; based on the temporal changes of the risk states in the risk state clusters, the temporal correlation strength between the risk state nodes is calculated, and the temporal correlation strength is determined according to the jump frequency of the risk state in the time series;

[0156] Establishing a directed connection relationship between the risk status nodes according to the temporal association strength, establishing a transfer connection between risk status node pairs whose temporal association strength is greater than a preset strength threshold, and constructing an initial structure of a multi-level causal reasoning decision tree based on the risk status nodes and the transfer connections;

[0157] The maximum information coefficient method is used to calculate the causal relationship strength between the risk status nodes, and the initial structure of the multi-level causal reasoning decision tree is optimized based on the causal relationship strength. The transfer connections with causal relationship strength lower than the preset causal threshold are deleted, and the transfer connections with strength higher than the preset causal threshold are retained to obtain an optimized multi-level causal reasoning decision tree.

[0158] In a specific implementation, the implementation process first comprehensively collects risk status data from the medical service system. These data cover risk information in various links such as the diagnosis and treatment process, execution of medical orders, use of equipment, and personnel operation. The system extracts features from the collected raw data to obtain multi-dimensional risk status features including time features, spatial features, and operation features. These features are standardized to eliminate dimensional differences between different dimensions so that all features are mapped to the same numerical range. Then, the principal component analysis method is used to reduce the dimension of the standardized features, retain the main feature information, and reduce the redundancy of the data.

[0159] The risk status features after dimensionality reduction are input into the pre-trained autoencoder network. In the encoder part, the high-dimensional risk status features are compressed and mapped into low-dimensional latent feature vectors through a multi-layer neural network. This vector contains the core feature information of the original risk status. The decoder part attempts to reconstruct the original risk status features through this latent feature vector to ensure that the latent feature vector retains enough information. Based on these latent feature vectors, the system calculates the Euclidean distance between different risk states and obtains a similarity matrix that reflects the similarity of risk states.

[0160] The system uses a clustering algorithm to group risk states. According to the pre-set similarity threshold, risk states with similarity higher than the threshold are classified into the same risk state cluster. By adjusting the similarity threshold, risk state clusters of different levels are formed to reflect the hierarchical structure of risk states. Each risk state cluster is mapped to a node in the causal reasoning decision tree. The system analyzes the change pattern of risk states in each risk state cluster over time, counts the jump frequency between risk states, and calculates the temporal correlation strength between different nodes based on this.

[0161] Based on the calculated temporal correlation strength, the system begins to build the connection relationship between nodes. When the temporal correlation strength between two nodes exceeds the preset strength threshold, the system establishes a directed transfer connection between the two nodes. Transfer connections are established between all node pairs that meet the conditions, thus forming the initial structure of a multi-level causal reasoning decision tree.

[0162] Finally, the system uses the maximum information coefficient method to deeply analyze the causal relationship between nodes. By calculating the causal relationship strength between each pair of connected nodes, the initial structure of the decision tree is optimized. The system deletes the transfer connections whose causal relationship strength is lower than the preset threshold, and only retains the connections with significant causal relationships, and finally obtains an optimized multi-level causal reasoning decision tree.

[0163] For example, take the management of the emergency infusion room of a hospital as an example:

[0164] The system first collects all the business data of the infusion room, including doctor's order information, medication records, infusion execution records, nursing records, etc. Multiple risk status features are extracted from these data, such as medication time delay, abnormal infusion rate, drug compatibility risk, etc. After standardization, these feature data are input into the principal component analysis model, and the main feature dimensions that can explain 85% of the variation are retained after dimensionality reduction.

[0165] The autoencoder network receives these reduced features and compresses the feature data originally containing more than ten dimensions into three-dimensional latent feature vectors. The decoder successfully reconstructs the original features with these latent feature vectors, and the reconstruction error is within an acceptable range. The system calculates the Euclidean distance between the latent feature vectors corresponding to all infusion events and constructs a complete similarity matrix.

[0166] By setting different similarity thresholds, the system divides the infusion risk status into three levels. The bottom layer contains specific risk manifestations, such as abnormal single drug administration time; the middle layer represents the concentrated occurrence of a certain type of risk, such as abnormal multiple consecutive drug administration times; the top layer represents systemic risks, such as overall infusion process disorder. Each risk status cluster corresponds to a node in the decision tree.

[0167] The system analysis found that when a single medication time anomaly occurs, there is a probability of more than 60% that continuous medication anomalies will occur within the next two hours. The strength of this temporal association exceeds the preset threshold, so a transfer connection is established between the corresponding nodes. Similarly, the system establishes connections between all nodes with significant temporal associations, forming an initial decision tree structure.

[0168] After analysis using the maximum information coefficient method, it was found that although some transfer connections have temporal correlation, the causal relationship is weak. For example, the causal relationship between the abnormal time of a single administration and the subsequent drug compatibility risk is low, and this connection is deleted by the system. The final decision tree clearly shows the true causal relationship between various risk states in the infusion room, providing a reliable decision-making basis for risk management.

[0169] In this embodiment, the original risk status data is reduced in dimension and key potential features are extracted through standardization, principal component analysis and autoencoder network, effectively reducing noise and redundant information and improving the accuracy of feature expression; the Euclidean distance is used to calculate the similarity of risk status, and similar risk status are classified into the same cluster through clustering to form risk status nodes at multiple levels, laying a fine foundation for subsequent causal analysis; a multi-level causal reasoning decision tree is constructed based on the temporal changes and transfer connections of risk status nodes to capture the temporal correlation and dynamic evolution laws between risk states and enhance the understanding of risk change trends; the maximum information coefficient method is used to quantify the strength of causal relationships between nodes, and the decision tree structure is optimized to ensure that only high-intensity transfer connections are retained, thereby improving the robustness and predictive ability of the model; the overall process helps to accurately identify changes in risk status and their causal mechanisms, providing reliable and detailed data support and decision-making basis for risk warning and intervention strategy formulation in medical services.

[0170] Figure 2 FIG. 1 is a schematic diagram of a medical service item compliance dynamic monitoring system based on artificial intelligence according to an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0171] The first unit is used to use the non-equilibrium entropy theory to dynamically map the medical service project data, establish a data association relationship matrix, and obtain project knowledge association data; based on the data association relationship matrix, a chaotic neural network is used to extract the risk feature distribution in the project knowledge association data; the Lyapunov stability analyzer is used to predict the long-term trend of the risk feature distribution and generate a risk benchmark model containing risk assessment weights and dynamic monitoring thresholds;

[0172] The second unit is used to use the deep generative adversarial diagnosis and treatment modeler to extract diagnosis and treatment feature sequences from project operation data and build a knowledge representation model; based on the knowledge representation model, a heterogeneous medical network neural evolution device is used to analyze the correlation strength and timing characteristics between various diagnosis and treatment activities, and a risk prediction spectrum is generated after comparison with historical compliance records. Combined with the risk benchmark model, the risk prediction spectrum is dynamically evolved and analyzed through the medical value dynamics framework, and real-time compliance analysis results are output;

[0173] The third unit is used to use causal reasoning decision trees to build a time-series evolution model of real-time compliance assessment results, calculate the risk evolution gradient, dynamically decompose the risk status matrix based on the risk evolution gradient, locate abnormal risk points, and generate risk warning information and risk intervention strategies based on the spatiotemporal distribution characteristics of abnormal risk points; feed back the spatiotemporal distribution characteristics, risk warning information and risk intervention strategies to the data association relationship matrix, and update the knowledge association data of medical service items.

[0174] According to a third aspect of the embodiments of the present invention,

[0175] An electronic device is provided, comprising:

[0176] processor;

[0177] a memory for storing processor-executable instructions;

[0178] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0179] A fourth aspect of the embodiments of the present invention is:

[0180] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0181] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0182] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A dynamic monitoring method for compliance of medical service items based on artificial intelligence, characterized in that: include: Use non-equilibrium entropy theory to dynamically map the medical service project data, establish a data association matrix, and obtain project knowledge association data; Based on the data association matrix, chaotic neural network is used to extract the risk feature distribution in project knowledge association data; Lyapunov stability analyzer is used to predict the long-term trend of risk feature distribution and generate a risk benchmark model including risk assessment weights and dynamic monitoring thresholds; Use the deep generative adversarial diagnosis and treatment modeler to extract diagnosis and treatment feature sequences from project operation data and build a knowledge representation model; Based on the knowledge representation model, a heterogeneous medical network neural evolution device is used to analyze the correlation strength and time series characteristics between various diagnosis and treatment activities, and a risk prediction spectrum is generated after comparison with historical compliance records. Combined with the risk benchmark model, the risk prediction spectrum is dynamically evolved and analyzed through the medical value dynamics framework to output real-time compliance analysis results; Use causal reasoning decision trees to build a time-series evolution model of real-time compliance assessment results, calculate the risk evolution gradient, dynamically decompose the risk state matrix based on the risk evolution gradient, locate abnormal risk points, and generate risk warning information and risk intervention strategies based on the spatiotemporal distribution characteristics of abnormal risk points; The spatiotemporal distribution characteristics, risk warning information and risk intervention strategies are fed back to the data association matrix to update the medical service item knowledge association data.

2. The method according to claim 1, characterized in that: Use non-equilibrium entropy theory to dynamically map the medical service project data, establish a data association matrix, and obtain project knowledge association data; Based on the data association matrix, chaotic neural network is used to extract the risk feature distribution in project knowledge association data; The Lyapunov stability analyzer is used to predict the long-term trend of risk feature distribution and generate a risk benchmark model containing risk assessment weights and dynamic monitoring thresholds, including: The medical service project data is constructed into an initial data set according to the service type, execution time, operator, service object, and operation record; the non-equilibrium entropy value between any two data items is obtained according to the probability distribution of each data item in the initial data set on the corresponding dimension, and the association strength between the data items is calculated according to the non-equilibrium entropy value and the smoothing factor to construct a data association relationship matrix; a time window mechanism is introduced into the data association relationship matrix, and the data association relationship matrix of the current time window and the historical data association relationship matrix are weightedly combined through the forgetting factor to generate project knowledge association data of the medical service project; The association strength in the project knowledge association data is used as the neuron connection weight to construct a chaotic neural network. The nonlinear iterative mapping is used to process the neuron input signal to keep the chaotic neural network in a chaotic critical state. The output of the chaotic neural network is reduced in dimension by the local sensitive hashing algorithm to obtain the risk characteristic distribution of medical service items. The risk characteristic distribution is converted into a dynamic system equation, and the Lyapunov index of the dynamic system equation in each characteristic dimension is calculated, and the Lyapunov index represents the degree of trajectory divergence of the characteristic dimension; the Lyapunov index is subjected to exponential transformation and normalization to generate a risk assessment weight; the historical data statistical value of each characteristic dimension is combined with the Lyapunov index of the corresponding dimension to generate a dynamic monitoring threshold, and the risk assessment weight and the dynamic monitoring threshold are combined to construct a risk benchmark model for medical service projects.

3. The method according to claim 2, characterized in that The association strength in the project knowledge association data is used as the neuron connection weight to construct a chaotic neural network. The nonlinear iterative mapping is used to process the neuron input signal to maintain the chaotic critical state of the chaotic neural network. The output of the chaotic neural network is reduced in dimension by the local sensitive hashing algorithm. The risk characteristic distribution of medical service items is obtained, including: The correlation strength of project knowledge association data is constructed as the neuron connection weight, and the chaotic critical state of the network is controlled by nonlinear iterative mapping. The weight is dynamically updated through the chaotic adaptive adjustment mechanism, the network synchronization is monitored by the chaotic resonance detection mechanism, and the risk feature distribution is obtained by local similarity mapping, which includes: The initial connection weight matrix of the chaotic neural network is constructed according to the correlation strength in the project knowledge association data; the input signal of the adjacent neurons and the initial connection weight matrix are input into the signal accumulation unit, the cumulative state value at the current moment is calculated according to the attenuation coefficient, the cumulative state value is processed by nonlinear iterative mapping and compared with the activation threshold, and when the cumulative state value is greater than the activation threshold, the neuron activation is triggered and enters the transient inhibition period; the positive activation energy and the negative inhibition energy are calculated and converted into a chaotic regulation signal, so that the chaotic neural network maintains a chaotic critical state; Construct a chaotic adaptive adjustment module to obtain the activation timing information of adjacent neurons, calculate the weight adjustment increment and dynamically merge it with the initial connection weight matrix, and update the connection weight matrix; adjust the corresponding sensitivity threshold according to the activation state of the neuron, and adjust the signal transmission intensity based on the overall synchronization degree of the chaotic neuron network; A chaotic resonance detection module is set up to calculate the synchronization degree based on the activation mode of neurons in the chaotic neural network, and when the synchronization degree is greater than a preset steady-state threshold, a feature extraction state is entered to generate a feature vector; The local sensitive hashing algorithm is used to perform dimension reduction mapping on the feature vector to obtain the risk feature distribution of medical service items.

4. The method according to claim 1, characterized in that: The deep generative adversarial diagnosis and treatment modeler is used to extract diagnosis and treatment feature sequences from project operation data and build a knowledge representation model. Based on the knowledge representation model, the heterogeneous medical network neural evolution device is used to analyze the correlation strength and time series characteristics between various diagnosis and treatment activities, and a risk prediction spectrum is generated after comparison with historical compliance records. Combined with the risk benchmark model, the risk prediction spectrum is dynamically evolved and analyzed through the medical value dynamics framework, and the real-time compliance analysis results are output, including: Construct a deep generative adversarial diagnosis and treatment modeler, wherein a residual connection structure is set in the generator network of the deep generative adversarial diagnosis and treatment modeler, and a feature fusion gating unit is set in each connection layer of the residual connection structure, and the features of different levels are adaptively fused through the feature fusion gating unit to obtain fusion features; a plurality of convolution kernels of different scales are set in the discriminator network of the deep generative adversarial diagnosis and treatment modeler, and the fusion features are multi-scale decomposed by group convolution to obtain decomposed features; the operation data of medical service items are input into the deep generative adversarial diagnosis and treatment modeler, and a diagnosis and treatment feature sequence is extracted based on the decomposition feature, and the diagnosis and treatment feature sequence is grouped according to time series features, behavior features and resource features to construct a knowledge representation model; The diagnosis and treatment feature sequence in the knowledge representation model is mapped to the feature nodes of the heterogeneous medical network, and the associated edges are constructed based on the temporal relationship, behavior association and resource dependency of the feature nodes; the node importance of the feature nodes is calculated as the starting probability of the random walk, and the temporal walk is executed to obtain the multi-hop walk path; the node transfer probability matrix is ​​calculated according to the multi-hop walk path, and the diagnosis and treatment activity association intensity matrix is ​​constructed; The medical treatment activity association intensity matrix is ​​aligned with the historical compliance records in time series to obtain aligned data, and the aligned data is input into a time series convolution network with a dilated convolution structure to extract multi-scale time series features and calculate the risk propagation coefficient to generate a risk prediction spectrum; a medical value dynamics equation group is constructed based on the risk prediction spectrum, and a risk propagation function and a value loss function are set in the medical value dynamics equation group, the risk propagation function describes the diffusion process of risk between the feature nodes according to the risk propagation coefficient, and the value loss function quantifies the impact of risk on the value of medical services; the medical value dynamics equation group is solved to obtain a dynamic evolution trajectory of the risk state, and a real-time compliance analysis result is output based on the dynamic evolution trajectory.

5. The method according to claim 4, characterized in that Calculating the node importance of the feature node as the starting probability of random walk, and executing the time series walk to obtain the multi-hop walk path includes: The comprehensive importance is calculated by integrating the community structure characteristics and historical activity sequence characteristics of the feature nodes as the starting probability of the random walk; in the process of executing the temporal walk, the transition probability is dynamically updated based on the local importance and temporal importance of the node, and the path sampling and skeleton extraction mechanism is introduced to generate a multi-hop walk path that meets the temporal transition constraints, including: Obtaining a connection relationship matrix of the characteristic nodes, decomposing the connection relationship matrix using a spectral clustering method to obtain a characteristic vector matrix, identifying a node community structure based on the characteristic vector matrix; calculating the structural density between node pairs within the node community structure to obtain a density matrix, calculating the propagation capacity in combination with the node's in-and-out degree distribution to obtain a propagation matrix; calculating the local importance of the node based on the density matrix and the propagation matrix; Collecting the historical activity sequence of the feature node, performing time series decomposition on the historical activity sequence to obtain periodic components, constructing a timing template based on the periodic components, matching the current activity sequence of the feature node with the timing template to obtain a matching degree distribution, and calculating the timing importance of the feature node according to the matching degree distribution; The local importance and the temporal importance are input into a feature fusion network, the feature fusion network dynamically adjusts the fusion weight based on an error feedback mechanism, outputs the comprehensive importance of the feature node, and performs normalization to obtain a starting probability distribution of a random walk; The forward transfer probability and the backward transfer probability between nodes are calculated according to the initial probability distribution, and a timing transfer probability matrix is ​​constructed based on the forward transfer probability and the backward transfer probability; a non-homogeneous timing walk strategy is initialized using the timing transfer probability matrix, and in each walk step, the timing transfer probability matrix is ​​updated based on the local importance and timing importance of the current node; the non-homogeneous timing walk strategy is executed to obtain an initial walk path, the initial walk path is segmented and sampled to obtain a sampling path set, timing key points are extracted from the sampling path set to determine a path skeleton, and a multi-hop walk path is generated based on the path skeleton, wherein the multi-hop walk path includes a node jump sequence that satisfies the timing transfer constraint.

6. The method according to claim 1, characterized in that Use causal reasoning decision trees to build a time-series evolution model of real-time compliance assessment results, calculate the risk evolution gradient, dynamically decompose the risk state matrix based on the risk evolution gradient, locate abnormal risk points, and generate risk warning information and risk intervention strategies based on the spatiotemporal distribution characteristics of abnormal risk points; Feedback the spatiotemporal distribution characteristics, risk warning information and risk intervention strategies to the data association matrix, and update the medical service project knowledge association data including: Acquire risk status data in the medical service process and extract risk status features, construct causal reasoning decision tree nodes according to the hierarchical relationship of the risk status features, establish transfer connections between nodes based on the temporal change relationship of the risk status data, and generate a multi-level causal reasoning decision tree; calculate the state transfer weights between adjacent nodes in the multi-level causal reasoning decision tree, normalize the state transfer weights to obtain the state transfer probability, and construct a risk state transfer matrix; Obtain the risk state transfer matrix of adjacent time steps, calculate the difference of the risk state transfer matrix, and obtain the risk evolution gradient according to the ratio of the difference to the time interval; perform singular value decomposition on the risk evolution gradient to obtain a feature matrix, and identify abnormal risk points of sudden changes in risk situation based on the singular values ​​in the feature matrix; Extracting the time span and spatial distribution information of the abnormal risk point, constructing a spatiotemporal feature vector in combination with the mutation degree of the abnormal risk point, determining the risk warning level according to the spatiotemporal feature vector, and generating a risk intervention strategy based on the risk warning level and the spatiotemporal feature vector; The spatiotemporal feature vector, the risk warning level and the risk intervention strategy are input into a feature fusion function, a data association relationship matrix is ​​updated according to an output result of the feature fusion function, and knowledge association data of medical service items is optimized based on the data association relationship matrix.

7. The method according to claim 6, characterized in that The risk status data in the medical service process is obtained and the risk status features are extracted. According to the hierarchical relationship of the risk status features, a causal reasoning decision tree node is constructed. Based on the temporal change relationship of the risk status data, a transfer connection between nodes is established. The generation of a multi-level causal reasoning decision tree includes: By standardizing and reducing the dimension of medical service risk status data, using the autoencoder network to extract potential feature vectors and calculate the risk status similarity, clustering the risk status nodes based on similarity, establishing transfer connections between nodes based on the time series change relationship, and using the maximum information coefficient to optimize the connection structure of the decision tree, specifically including: Acquire risk status data in the medical service process and extract risk status features, perform standardization on the risk status features to obtain standardized features, and perform dimensionality reduction on the standardized features using a principal component analysis method to obtain reduced-dimensionality risk status features; Inputting the reduced-dimensionality risk state feature into an autoencoder network, wherein the encoder part of the autoencoder network maps the reduced-dimensionality risk state feature into a latent feature vector, and the decoder part of the autoencoder network reconstructs the reduced-dimensionality risk state feature based on the latent feature vector, and calculates the Euclidean distance between risk states according to the latent feature vector to obtain the risk state similarity; A clustering operation is performed based on the risk state similarity, and the corresponding risk states are classified into a risk state cluster according to a preset similarity threshold, so as to obtain a plurality of risk state clusters at different levels, and each of the risk state clusters is mapped to a risk state node of a causal reasoning decision tree; based on the temporal changes of the risk states in the risk state clusters, the temporal correlation strength between the risk state nodes is calculated, and the temporal correlation strength is determined according to the jump frequency of the risk state in the time series; Establishing a directed connection relationship between the risk status nodes according to the temporal association strength, establishing a transfer connection between risk status node pairs whose temporal association strength is greater than a preset strength threshold, and constructing an initial structure of a multi-level causal reasoning decision tree based on the risk status nodes and the transfer connections; The maximum information coefficient method is used to calculate the causal relationship strength between the risk status nodes, and the initial structure of the multi-level causal reasoning decision tree is optimized based on the causal relationship strength. The transfer connections with causal relationship strength lower than the preset causal threshold are deleted, and the transfer connections with strength higher than the preset causal threshold are retained to obtain an optimized multi-level causal reasoning decision tree.

8. A medical service item compliance dynamic monitoring system based on artificial intelligence, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to use the non-equilibrium entropy theory to dynamically map the medical service project data, establish a data association relationship matrix, and obtain project knowledge association data; Based on the data association matrix, chaotic neural network is used to extract the risk feature distribution in project knowledge association data; Lyapunov stability analyzer is used to predict the long-term trend of risk feature distribution and generate a risk benchmark model including risk assessment weights and dynamic monitoring thresholds; The second unit is used to extract diagnosis and treatment feature sequences from project operation data using a deep generative adversarial diagnosis and treatment modeler to build a knowledge representation model; Based on the knowledge representation model, a heterogeneous medical network neural evolution device is used to analyze the correlation strength and time series characteristics between various diagnosis and treatment activities, and a risk prediction spectrum is generated after comparison with historical compliance records. Combined with the risk benchmark model, the risk prediction spectrum is dynamically evolved and analyzed through the medical value dynamics framework to output real-time compliance analysis results; The third unit is used to build a time-series evolution model of real-time compliance assessment results using a causal reasoning decision tree, calculate the risk evolution gradient, dynamically decompose the risk state matrix based on the risk evolution gradient, locate abnormal risk points, and generate risk warning information and risk intervention strategies based on the spatiotemporal distribution characteristics of abnormal risk points; The spatiotemporal distribution characteristics, risk warning information and risk intervention strategies are fed back to the data association matrix to update the medical service item knowledge association data.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images