Disease prevention management and monitoring system design method based on adaptive algorithm

Abstract

The invention discloses a disease prevention management and monitoring system design method based on an adaptive algorithm. The method comprises the following steps: S1, obtaining health data of a user; s2, performing feature extraction and fusion on the health data by using a hypergraph neural network; s3, constructing a disease risk prediction model based on the generative reasoning network, and introducing an anti-fact generation module to generate a health change hypothesis scene; s4, adaptively adjusting a personalized intervention strategy through a nested game model, and adjusting the intervention intensity in the first layer; s5, in the second layer, generating an optimized intervention combination strategy based on a Nash equilibrium driven multi-target adaptive game; s6, risk weighted scoring is carried out on the health data through a structured clustering algorithm, and a feedback event is generated; and S7, dynamically optimizing a disease prevention management strategy through a chained feedback mechanism, and managing node permission. According to the method, the self-adaptive algorithm is combined with the multi-layer game and the distributed anomaly detection, so that personalized disease prevention management and monitoring are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of disease prevention, and in particular to a disease prevention management and monitoring system design method based on an adaptive algorithm. Background Art

[0002] The existing disease prevention management and monitoring systems have gradually evolved with the rapid development of big data and artificial intelligence technologies, but there are still great technical challenges in dealing with complex health data, real-time analysis of multi-source information and personalized intervention. Traditional disease prevention management mainly relies on data collection and processing by medical institutions, combining multi-source data such as wearable devices and environmental monitoring devices to obtain health-related information. However, these systems can usually only simply monitor or record specific physiological parameters and risk factors, making it difficult to achieve dynamic prediction of disease development and personalized intervention guidance, thus limiting the effectiveness and pertinence of disease prevention management.

[0003] In the existing disease prevention management and monitoring systems, most of them adopt a centralized data processing method, which summarizes genetic information, physical parameters and environmental factors and then analyzes them. This centralized data processing mode makes the system unable to flexibly respond to the complex relationship between multi-source data. For example, individual genetic factors, living habits and environmental changes jointly determine health risks, but these risk factors often have highly nonlinear and dynamic correlations. Traditional data fusion technology is usually unable to deeply explore the potential correlations between multi-source data, so it is easy to cause information omissions and cannot accurately reflect the true situation of individual disease risks. The lack of deep fusion and analysis of multi-source health data makes it difficult for the current disease prevention system to capture the early signals of disease development in a timely manner, thus failing to achieve efficient early warning.

[0004] In terms of disease risk prediction, traditional disease monitoring systems often rely on static data and models to identify health risks. Although this method can identify an individual's current health status, it has significant deficiencies in simulating the dynamic process of disease development. The occurrence and development of a disease is not the result of a single time period, but a gradual evolution process under the combined effect of multiple factors. Current disease risk prediction methods lack multi-scenario simulation and dynamic adjustment functions, and cannot generate personalized health change scenarios based on individual health characteristics. With limited static data support, it is difficult for the prediction model to generate accurate disease risk assessment results, thereby reducing the effectiveness and accuracy of preventive interventions.

[0005] In the formulation and implementation of personalized intervention strategies, the existing system lacks adaptability and flexibility, and usually uniformly intervenes high-risk groups based on fixed preset strategies. This "one-size-fits-all" approach fails to take into account the differentiated needs of individuals, which limits the effectiveness of intervention. For example, the competition and synergy between multiple intervention factors such as lifestyle, nutrition and exercise have not been fully explored. Due to the differences in health status and risk factors among different individuals, the acceptance of intervention measures also varies, but the existing system often lacks a dynamic adjustment mechanism based on individual feedback and status changes, resulting in limitations in the intervention effect. This strategy mechanism that lacks multi-level adaptive games is difficult to meet personalized needs in disease prevention and management, and it is impossible to optimize intervention plans in real time to adapt to changes in health risks.

[0006] In terms of anomaly detection and feedback control, traditional disease monitoring systems usually use a single threshold or rule detection method. For the correlation and dynamic changes between multi-source data, the centralized threshold detection method fails to accurately capture disease risk signals. For example, the health status of an individual is often the result of the interaction of multiple factors, and a single threshold or rule is difficult to fully monitor the nonlinear correlation between these factors. The current disease monitoring system lacks a distributed collaborative anomaly detection mechanism, resulting in a relatively lagging detection process and difficulty in timely identifying potential health risks. Moreover, the existing technology fails to achieve real-time feedback control and fails to adaptively adjust health monitoring parameters after detecting anomalies, making the system response slow and unable to effectively guarantee the real-time nature and prevention effect of monitoring.

[0007] In terms of privacy protection and data management, since disease prevention management and monitoring systems often involve personal sensitive health information, data security and privacy protection have become key issues. Traditional centralized storage and permission management methods are prone to data leakage risks, especially the fusion of multi-source data and the application of distributed nodes increase the complexity of data management. Existing technologies usually lack effective data distribution management and permission control mechanisms, and cannot perform refined management of permissions of different nodes, resulting in difficulty in balancing the sharing of health data and privacy protection.

[0008] Therefore, how to provide a disease prevention management and monitoring system design method based on an adaptive algorithm is an urgent problem that technicians in this field need to solve. Summary of the invention

[0009] One purpose of the present invention is to propose a disease prevention management and monitoring system design method based on an adaptive algorithm. The present invention fully combines adaptive algorithms, multi-layer nested games and distributed collaborative anomaly detection technology to deeply integrate and dynamically analyze multi-source health data to achieve personalized disease risk prediction and intervention management. By extracting multi-dimensional health features through hypergraph neural networks and using generative reasoning networks to simulate multi-scenario health development paths, the system can accurately identify potential health risks. On this basis, distributed collaborative detection and chain feedback control mechanisms further improve the real-time and response sensitivity of monitoring, and have the advantages of timely response, accurate risk identification and efficient privacy protection.

[0010] A disease prevention management and monitoring system design method based on an adaptive algorithm according to an embodiment of the present invention comprises the following steps:

[0011] S1. Obtain the user's health data through a multi-source data acquisition module, wherein the health data includes genetic information, real-time physical parameters and environmental factors, and input the health data into a hypergraph neural network;

[0012] S2. Use the hypergraph convolution module of the hypergraph neural network to extract and fuse features of health data, map the health data into a hypergraph structure, identify potential associations between different data sources, and output feature fusion results;

[0013] S3. Construct a disease risk prediction model based on a generative reasoning network, generate personalized future health status predictions through feature fusion results and simulation of health development paths, and introduce a counterfactual generation module to simulate different health scenarios through changes in a small number of key variables to generate health change hypothetical scenarios;

[0014] S4, adaptively adjust the personalized intervention strategy through the nested game model. In the first layer of the nested game model, the intervention intensity is adjusted according to the changes in the user's health status and the adoption of suggestions in the hypothetical scenario;

[0015] S5. In the second layer of the nested game model, the competition and cooperation of lifestyle, nutrition and exercise are simulated, and the optimized intervention combination strategy is generated based on the data in the health change hypothetical scenario and the multi-objective adaptive game driven by Nash equilibrium;

[0016] S6. Apply the distributed collaborative anomaly detection module to monitor and detect abnormal changes in health data in real time, perform risk-weighted scoring on health data through a structured clustering algorithm, and trigger a chain feedback control mechanism to generate feedback events when health risks are detected;

[0017] S7. Through the closed-loop management function of the chain feedback mechanism, the disease prevention management strategy is dynamically optimized according to the user's health data, feedback response results and health change hypothesis scenarios, and the authority of each node in the feedback chain is managed based on smart contracts.

[0018] Optionally, the S2 specifically includes:

[0019] S21, normalizing the acquired user health data, including genetic information, real-time vital sign parameters and environmental factors, to form a multi-dimensional feature vector, and establishing a health data node set with the feature vector of each health data type as a node;

[0020] S22. Based on the similarity between health data sources, different types of hyperedges are defined to connect multiple nodes with potential associations. The connection weight of each hyperedge is determined by feature similarity and data source type to increase the association of similar features. A weight matrix is ​​assigned to each hyperedge to establish a hyperedge set.

[0021] S23, mapping the health data node set and the hyperedge set to a hypergraph structure G=(V,E), where V represents the health data node set and E represents the hyperedge set;

[0022] S24. Apply the improved hypergraph convolution operation to the hypergraph structure G, realize complex multi-source data association aggregation by combining convolution aggregation functions, fuse the feature information of each node and iteratively update:

[0023]

[0024] Among them, X ′ represents the feature matrix after the hypergraph convolution operation, X represents the input feature matrix, σ represents the nonlinear activation function, G represents the hypergraph association matrix, W represents the weight matrix, and D v represents the degree matrix of node v, D e represents the degree matrix of the hyperedge e, β and α represent the edge connection weight adjustment parameters, N represents the convolution depth, represents the convolution kernel;

[0025] S25. Perform multi-layer hypergraph convolution aggregation operations, iteratively extract multi-dimensional features of health data, construct an aggregate feature matrix containing multi-layer health data relationships, and retain inter-layer feature information through residual connections:

[0026]

[0027] Among them, Z (l+1) represents the feature matrix of the l+1th layer, Z (l) represents the feature matrix of the lth layer, G (l) represents the hypergraph incidence matrix at level l, W (l)represents the master weight matrix of the lth layer, represents the additional weight matrix of the lth layer, λ i represents the additional weight scaling factor, δ represents the association adjustment factor, γ represents the residual weight factor, and I represents the number of additional weight matrices;

[0028] S26. The final feature fusion matrix obtained after multi-layer hypergraph convolution aggregation is used as the high-dimensional representation of health data, and the output is the feature fusion result of disease risk prediction.

[0029] Optionally, the S3 specifically includes:

[0030] S31, using the feature fusion result generated by the hypergraph convolution as the input data of the generated inference network, and establishing a feature vector set F = {f1, f2, ..., f n}, where each eigenvector f i The feature fusion representation of a specific health data node is represented by inputting the feature vector set into the generator module of the generative reasoning network to construct the baseline health state feature vector S;

[0031] S32. In the generation module of the generation reasoning network, the reference health state feature vector S is combined with the random noise vector z, and a generator G is used to output multiple simulated health states:

[0032]

[0033] in, represents the simulated health state output by the generator G, tanh represents the hyperbolic tangent function, A represents the number of feature vectors, B represents the dimension of the random noise vector, and W ij represents the health status weight matrix of the generator, b ij represents the health state bias term, exp represents the exponential function, ψ k and χ l represents the healthy development kernel function in the generator, K represents the kernel depth, L represents the number of frequency kernels, and ζ represents the healthy regulation factor of the generation process;

[0034] S33. Introduce a counterfactual generation module in the generative reasoning network, define a set of key health variables, set potential risk factors related to specific diseases in the user's health data, and construct a counterfactual generation framework under different scenarios by applying different amplitudes of changes to each variable in the set of key health variables;

[0035] S34. In the counterfactual generation module, by controlling the change value of each variable in the key health variable set, the baseline health state is adjusted to generate future health state vectors with different health scenarios:

[0036]

[0037] in, represents the future health state vector, ΔR represents the change value of the variable, C represents the total number of key health variables, and W i represents the adaptive weight of the i-th key health variable, μ and ν represent the parameters that control the sensitivity of health status changes, and λ p Represents the response coefficient of different key variables, R p represents the pth variable in the key health variable set, z p represents the pth noise component in the random noise vector, P represents the depth of the key variable response, and Q represents the number of frequency kernels of the key variable;

[0038] S35, integrating the health status vectors generated by multiple health scenarios into a prediction matrix Each of these Represents a hypothetical scenario of health change, reflecting the future health status under different health scenarios and changes in key variables

[0039] S36. Use the prediction matrix of the health change hypothesis scenario as the final output of the personalized health status prediction to generate a personalized future health status prediction.

[0040] Optionally, the S4 specifically includes:

[0041] S41, establishing a user health status monitoring module of the first-layer game model, by collecting the user's health data in real time, generating a current health status feature vector for dynamically analyzing the changing trend of the health status;

[0042] S42, based on the monitored health status characteristics, compare the current health status with the historical health data, identify the key change nodes of the user's health status, and provide adjustment basis for the intervention strategy;

[0043] S43. Introduce hypothetical health change scenarios based on the user's health status changes, generate multiple health scenarios by simulating the user's possible health development path, evaluate the possible health risks in each scenario, and design different adjustment plans for the intervention strategy;

[0044] S44. Analyze the adoption of intervention suggestions in the health change hypothetical scenario, and evaluate the effectiveness of each intervention plan based on the user's acceptance and feedback of different intervention suggestions;

[0045] S45, generating an adaptive intervention intensity adjustment plan based on the real-time changes in the user's health status and the adoption of intervention suggestions;

[0046] S46. Apply the adjusted intervention intensity to the personalized intervention strategy, update the strategy output of the first-level game model in real time, and regularly evaluate the intervention effect.

[0047] Optionally, the S5 specifically includes:

[0048] S51. Construct the utility functions of the lifestyle intervention module, the nutrition intervention module and the exercise intervention module. Based on the data in the health change hypothesis scenario, set the objective function of each intervention module, quantify the degree of impact on the user's health status, and generate the expected health benefits of each module according to the specific intervention strategy.

[0049] S52. Define the strategy space between intervention modules to form a set of lifestyle intervention modules, nutrition intervention modules and exercise intervention strategies X i , where i = 1, 2, and 3 represent lifestyle intervention module, nutrition intervention module, and exercise intervention module respectively. The strategy set X of each intervention module i Affected by resource constraints, user acceptance and implementation costs, the set of feasible solutions is maximized by optimizing the strategy space;

[0050] S53. Construct a multi-objective adaptive game model, introduce the Nash equilibrium mechanism, and define the benefit function U of each intervention module. i (x i ,x -i ), where x i represents the strategy of intervention module i, x -i Represents a set of strategies for other intervention modules:

[0051]

[0052] Among them, U i (x i ,x -i ) represents the profit function of intervention module i, exp represents the exponential function, ω i represents the intervention effect weight, E i (x i ) represents the health benefit function of intervention module i, λ ij represents the synergistic effect coefficient, S ij (x i ,x j ) represents the cooperation benefit function between intervention module i and intervention module j, C i (x i ) represents the intervention cost of intervention module i, τ represents the cost adjustment factor, D represents the number of features used in the benefit function in the strategy set, M represents the strategy dimension of the intervention module, ψ ij , and θ lrepresents the strategy adjustment parameter, x ik represents the k-th strategy feature for the profit function in intervention module i, x il represents the lth strategy feature in intervention module i;

[0053] S54. Use the evolutionary game algorithm to solve the multi-objective adaptive game model, and calculate the benefit gradient optimization strategy of the intervention module through each iteration:

[0054]

[0055] in, represents the t+1th iteration strategy of intervention module i, represents the t-th iteration strategy of intervention module i, η represents the learning rate, Y represents the number of frequency kernels used to respond to different health status characteristics during the strategy update process, r represents the number of features used for adjustment in the strategy update, t represents the cumulative effect depth in the strategy update, and x iq represents the qth strategy feature in intervention module i, x is represents the sth strategy feature in intervention module i, x iu represents the u-th strategy feature in intervention module i, represents the adjustment factor in the strategy update;

[0056] S55. During the iteration process, the synergy coefficient is adaptively adjusted according to user health feedback and intervention effects; when the synergy effect is enhanced, the synergy coefficient is increased to promote cooperation; when the synergy effect is weakened, the synergy coefficient is reduced to strengthen competition;

[0057] S56. The iteration process continues until the strategies of each intervention module converge to the Nash equilibrium point, generating the final optimized intervention combination strategy.

[0058] Optionally, the S6 specifically includes:

[0059] S61. Construct a distributed anomaly detection model for health data, distribute multi-source health data to different distributed nodes according to data source, feature type and time series, and form a multi-level distributed feature vector set D = {d1, d2, ..., d n}, where d i is the i-th feature data node, which represents the node representation of a specific health parameter in the distributed structure;

[0060] S62: Apply the collaborative anomaly detection algorithm in the distributed nodes and set the anomaly detection threshold τ of each feature data node. i , detect abnormal data points according to the real-time changes of health data. When the feature value of a node is detected to exceed the set abnormal detection threshold τ iWhen , the node is marked as abnormal, the abnormal data is recorded and sent to the central control node;

[0061] S63, perform structured cluster analysis on the health data marked as abnormal, cluster the abnormal data according to the associated feature similarity, and form a risk cluster set C = {c1, c2, ..., c m}, where each risk cluster c j Represents a group of health data points with similar abnormal characteristics. Each risk cluster is risk-weighted and scored according to the characteristic weight to quantify the risk level of abnormal health data.

[0062] S64. Assess risk scores for each risk cluster and adjust intervention strategies according to different risk levels:

[0063]

[0064] Among them, R represents the health risk score, R low represents the low score threshold, R medium represents the scoring threshold, R high represents the high score threshold, exp represents the exponential function, α1, β1, γ1, δ1, λ1, α2, β2, θ2, γ2, δ2, α3, β3, θ3, and δ3 represent adjustment parameters that control the response intensity at different risk levels;

[0065] S65. When the risk score exceeds the set safety threshold, the chain feedback control mechanism is triggered to generate a feedback event. The health monitoring strategy is adjusted accordingly according to the risk level of the feedback event. The feedback event is transmitted to each distributed node through the chain feedback mechanism, and the anomaly detection threshold τ is dynamically updated in each node. i ;

[0066] S66. Under the closed-loop management of the chain feedback control mechanism, the health monitoring strategy is continuously optimized according to the feedback events, and the feedback results of the risk cluster are used to adaptively adjust the anomaly detection parameters of each node by dynamically adjusting the detection threshold and weighted scoring parameters of each distributed node.

[0067] The beneficial effects of the present invention are:

[0068] First, the present invention uses a hypergraph neural network to perform multi-dimensional feature extraction and deep fusion of health data, mapping genetic information, physical parameters, living habits and environmental factors to a hypergraph structure, and fully exploring the potential correlations between data. In this way, the system can generate high-dimensional and detailed health features, accurately reflect the dynamic health status of individuals, and provide richer and more accurate basic data for disease risk assessment. Compared with traditional data analysis methods, the present invention greatly improves the ability to analyze health features at the data fusion level, and can identify potential risk signals in the early stages of the disease, laying a solid foundation for personalized early prevention.

[0069] In terms of disease risk prediction, the present invention uses a generative reasoning network to dynamically model and simulate multiple scenarios of individual health status, simulates future health status under different health scenarios through a counterfactual generation module, and generates personalized health change hypothesis scenarios for each individual. Compared with traditional static health prediction models, the present invention can dynamically adjust the prediction path according to the individual's current health status and historical data through a generative reasoning network, and flexibly respond to changing trends in health status. The present invention can not only simulate potential health development paths, but also identify the potential impact of changes in key variables on health status, thereby providing an accurate basis for intervention decisions. This dynamic prediction mechanism greatly improves the accuracy of disease risk prediction and provides an important reference for the formulation of personalized disease prevention strategies.

[0070] Secondly, the present invention adopts a nested game model in the personalized intervention strategy. The first-layer game model adaptively adjusts the intervention intensity according to the changes in the user's health status and the adoption of suggestions in the health change hypothesis scenario; in the second-layer game model, the competition and cooperation between the lifestyle, nutrition, and exercise intervention modules are simulated, and the optimal intervention combination strategy is generated based on multi-objective adaptive games. This hierarchical game mechanism not only effectively coordinates the resource allocation and strategy selection among the intervention modules, but also ensures the dynamic matching of the intervention intensity with the user's health status. Compared with the traditional fixed intervention plan, the present invention can adjust the intensity and direction of the intervention strategy in real time, improve the accuracy of personalized intervention and user acceptance, avoid the problem of insufficient or excessive intervention, and make the intervention plan more in line with the actual needs of the individual.

[0071] In terms of anomaly detection and real-time feedback, the present invention uses a distributed collaborative anomaly detection module to monitor abnormal changes in health data in real time, and uses a structured clustering algorithm to perform cluster analysis and risk-weighted scoring on abnormal data. At the same time, the system introduces a chain feedback control mechanism to achieve closed-loop management of anomaly detection and feedback. When a health risk is detected, it can immediately generate a feedback event and dynamically adjust the monitoring parameters to adapt to real-time changes in health status. Compared with the existing centralized anomaly detection method, the distributed collaborative anomaly detection mechanism of the present invention can identify abnormal changes in health data more quickly, improve the system's response speed to health risks, and ensure the timeliness of disease risk warnings. The chain feedback mechanism further optimizes the sensitivity of anomaly detection, and can adaptively adjust the monitoring parameters of each distributed node according to the feedback results to ensure the continuous stability and efficiency of monitoring.

[0072] In addition, the present invention also realizes the security management and privacy protection of health data through distributed rights management and smart contracts. User health data is stored independently in distributed nodes, and the rights of each node are managed by smart contracts, ensuring the security of data access and the privacy protection of sensitive information. Compared with the traditional centralized data storage method, the distributed rights management mechanism of the present invention can effectively prevent the risk of data leakage, and supports data sharing and collaboration among multiple nodes, providing security for personalized health management and cross-institutional data sharing. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0074] Figure 1 A flowchart of a disease prevention management and monitoring system design method based on an adaptive algorithm proposed by the present invention;

[0075] Figure 2 This is a structural schematic diagram of a disease prevention management and monitoring system design method based on an adaptive algorithm proposed by the present invention. DETAILED DESCRIPTION

[0076] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0077] refer to Figure 1 and Figure 2 , a disease prevention management and monitoring system design method based on adaptive algorithm, comprising the following steps:

[0078] S1. Obtain the user's health data through a multi-source data acquisition module, wherein the health data includes genetic information, real-time physical parameters and environmental factors, and input the health data into a hypergraph neural network;

[0079] S2. Use the hypergraph convolution module of the hypergraph neural network to extract and fuse features of health data, map the health data into a hypergraph structure, identify potential associations between different data sources, and output feature fusion results;

[0080] S3. Construct a disease risk prediction model based on a generative reasoning network, generate personalized future health status predictions through feature fusion results and simulation of health development paths, and introduce a counterfactual generation module to simulate different health scenarios through changes in a small number of key variables to generate health change hypothetical scenarios;

[0081] S4, adaptively adjust the personalized intervention strategy through the nested game model. In the first layer of the nested game model, the intervention intensity is adjusted according to the changes in the user's health status and the adoption of suggestions in the hypothetical scenario;

[0082] S5. In the second layer of the nested game model, the competition and cooperation of lifestyle, nutrition and exercise are simulated, and the optimized intervention combination strategy is generated based on the data in the health change hypothetical scenario and the multi-objective adaptive game driven by Nash equilibrium;

[0083] S6. Apply the distributed collaborative anomaly detection module to monitor and detect abnormal changes in health data in real time, perform risk-weighted scoring on health data through a structured clustering algorithm, and trigger a chain feedback control mechanism to generate feedback events when health risks are detected;

[0084] S7. Through the closed-loop management function of the chain feedback mechanism, the disease prevention management strategy is dynamically optimized according to the user's health data, feedback response results and health change hypothesis scenarios, and the authority of each node in the feedback chain is managed based on smart contracts.

[0085] In this implementation, S2 specifically includes:

[0086] S21, normalizing the acquired user health data, including genetic information, real-time vital sign parameters and environmental factors, to form a multi-dimensional feature vector, and establishing a health data node set with the feature vector of each health data type as a node;

[0087] S22. Based on the similarity between health data sources, different types of hyperedges are defined to connect multiple nodes with potential associations. The connection weight of each hyperedge is determined by feature similarity and data source type to increase the association of similar features. A weight matrix is ​​assigned to each hyperedge to establish a hyperedge set.

[0088] S23, mapping the health data node set and the hyperedge set to a hypergraph structure G=(V,E), where V represents the health data node set and E represents the hyperedge set;

[0089] S24. Apply the improved hypergraph convolution operation to the hypergraph structure G, realize complex multi-source data association aggregation by combining convolution aggregation functions, fuse the feature information of each node and iteratively update:

[0090]

[0091] Among them, X ′ represents the feature matrix after the hypergraph convolution operation, X represents the input feature matrix, σ represents the nonlinear activation function, G represents the hypergraph association matrix, W represents the weight matrix, and D v represents the degree matrix of node v, D e represents the degree matrix of the hyperedge e, β and α represent the edge connection weight adjustment parameters, N represents the convolution depth, represents the convolution kernel;

[0092] S25. Perform multi-layer hypergraph convolution aggregation operations, iteratively extract multi-dimensional features of health data, construct an aggregate feature matrix containing multi-layer health data relationships, and retain inter-layer feature information through residual connections:

[0093]

[0094] Among them, Z (l+1) represents the feature matrix of the l+1th layer, Z (l) represents the feature matrix of the lth layer, G (l) represents the hypergraph incidence matrix at level l, W (l) represents the master weight matrix of the lth layer, represents the additional weight matrix of the lth layer, λ i represents the additional weight scaling factor, δ represents the association adjustment factor, γ represents the residual weight factor, and I represents the number of additional weight matrices;

[0095] S26. The final feature fusion matrix obtained after multi-layer hypergraph convolution aggregation is used as the high-dimensional representation of health data, and the output is the feature fusion result of disease risk prediction.

[0096] In this implementation, S3 specifically includes:

[0097] S31, using the feature fusion result generated by the hypergraph convolution as the input data of the generated inference network, and establishing a feature vector set F = {f1, f2, ..., f n}, where each eigenvector f iThe feature fusion representation of a specific health data node is represented by inputting the feature vector set into the generator module of the generative reasoning network to construct the baseline health state feature vector S;

[0098] S32. In the generation module of the generation reasoning network, the reference health state feature vector S is combined with the random noise vector z, and a generator G is used to output multiple simulated health states:

[0099]

[0100] in, shows the simulated health status output by the generator G, tanh represents the hyperbolic tangent function, A represents the number of feature vectors, B represents the dimension of the random noise vector, and W ij represents the health status weight matrix of the generator, b ij represents the health state bias term, exp represents the exponential function, ψ k and χ l represents the healthy development kernel function in the generator, K represents the kernel depth, L represents the number of frequency kernels, and ζ represents the healthy regulation factor of the generation process;

[0101] S33. Introduce a counterfactual generation module in the generative reasoning network, define a set of key health variables, set potential risk factors related to specific diseases in the user's health data, and construct a counterfactual generation framework under different scenarios by applying different amplitudes of changes to each variable in the set of key health variables;

[0102] S34. In the counterfactual generation module, by controlling the change value of each variable in the key health variable set, the baseline health state is adjusted to generate future health state vectors with different health scenarios:

[0103]

[0104] in, represents the future health state vector, ΔR represents the change value of the variable, C represents the total number of key health variables, and W i represents the adaptive weight of the i-th key health variable, μ and ν represent the parameters that control the sensitivity of health status changes, and λ p Represents the response coefficient of different key variables, R p represents the pth variable in the key health variable set, z p represents the pth noise component in the random noise vector, P represents the depth of the key variable response, and Q represents the number of frequency kernels of the key variable;

[0105] S35, integrating the health status vectors generated by multiple health scenarios into a prediction matrix Each of these Represents a hypothetical scenario of health change, reflecting the future health status under different health scenarios and changes in key variables

[0106] S36. Use the prediction matrix of the health change hypothesis scenario as the final output of the personalized health status prediction to generate a personalized future health status prediction.

[0107] In this implementation, S4 specifically includes:

[0108] S41, establishing a user health status monitoring module of the first-layer game model, by collecting the user's health data in real time, generating a current health status feature vector for dynamically analyzing the changing trend of the health status;

[0109] S42, based on the monitored health status characteristics, compare the current health status with the historical health data, identify the key change nodes of the user's health status, and provide adjustment basis for the intervention strategy;

[0110] S43. Introduce hypothetical health change scenarios based on the user's health status changes, generate multiple health scenarios by simulating the user's possible health development path, evaluate the possible health risks in each scenario, and design different adjustment plans for the intervention strategy;

[0111] S44. Analyze the adoption of intervention suggestions in the health change hypothetical scenario, and evaluate the effectiveness of each intervention plan based on the user's acceptance and feedback of different intervention suggestions;

[0112] S45, generating an adaptive intervention intensity adjustment plan based on the real-time changes in the user's health status and the adoption of intervention suggestions;

[0113] S46. Apply the adjusted intervention intensity to the personalized intervention strategy, update the strategy output of the first-level game model in real time, and regularly evaluate the intervention effect.

[0114] In this implementation manner, S5 specifically includes:

[0115] S51. Construct the utility functions of the lifestyle intervention module, the nutrition intervention module and the exercise intervention module. Based on the data in the health change hypothesis scenario, set the objective function of each intervention module, quantify the degree of impact on the user's health status, and generate the expected health benefits of each module according to the specific intervention strategy.

[0116] S52. Define the strategy space between intervention modules to form a set of lifestyle intervention modules, nutrition intervention modules and exercise intervention strategies X i , where i = 1, 2, and 3 represent lifestyle intervention module, nutrition intervention module, and exercise intervention module respectively. The strategy set X of each intervention modulei Affected by resource constraints, user acceptance and implementation costs, the set of feasible solutions is maximized by optimizing the strategy space;

[0117] S53. Construct a multi-objective adaptive game model, introduce the Nash equilibrium mechanism, and define the benefit function U of each intervention module. i (x i ,x -i ), where x i represents the strategy of intervention module i, x -i Represents a set of strategies for other intervention modules:

[0118]

[0119] Among them, U i (x i ,x -i ) represents the profit function of intervention module i, exp represents the exponential function, ω i represents the intervention effect weight, E i (x i ) represents the health benefit function of intervention module i, λ ij represents the synergistic effect coefficient, S ij (x i ,x j ) represents the cooperation benefit function between intervention module i and intervention module j, C i (x i ) represents the intervention cost of intervention module i, τ represents the cost adjustment factor, D represents the number of features used in the benefit function in the strategy set, M represents the strategy dimension of the intervention module, ψ ij , and θ l represents the strategy adjustment parameter, x ik represents the k-th strategy feature for the profit function in intervention module i, x il represents the lth strategy feature in intervention module i;

[0120] S54. Use the evolutionary game algorithm to solve the multi-objective adaptive game model, and calculate the benefit gradient optimization strategy of the intervention module through each iteration:

[0121]

[0122] in, represents the t+1th iteration strategy of intervention module i, represents the t-th iteration strategy of intervention module i, η represents the learning rate, Y represents the number of frequency kernels used to respond to different health status characteristics during the strategy update process, r represents the number of features used for adjustment in the strategy update, t represents the cumulative effect depth in the strategy update, and x iqrepresents the qth strategy feature in intervention module i, x is represents the sth strategy feature in intervention module i, x iu represents the u-th strategy feature in intervention module i, represents the adjustment factor in the strategy update;

[0123] S55. During the iteration process, the synergy coefficient is adaptively adjusted according to user health feedback and intervention effects; when the synergy effect is enhanced, the synergy coefficient is increased to promote cooperation; when the synergy effect is weakened, the synergy coefficient is reduced to strengthen competition;

[0124] S56. The iteration process continues until the strategies of each intervention module converge to the Nash equilibrium point, generating the final optimized intervention combination strategy.

[0125] In this implementation manner, S6 specifically includes:

[0126] S61. Construct a distributed anomaly detection model for health data, distribute multi-source health data to different distributed nodes according to data source, feature type and time series, and form a multi-level distributed feature vector set D = {d1, d2, ..., d n}, where d i is the i-th feature data node, which represents the node representation of a specific health parameter in the distributed structure;

[0127] S62: Apply the collaborative anomaly detection algorithm in the distributed nodes and set the anomaly detection threshold τ of each feature data node. i , detect abnormal data points according to the real-time changes of health data. When the feature value of a node is detected to exceed the set abnormal detection threshold τ i When , the node is marked as abnormal, the abnormal data is recorded and sent to the central control node;

[0128] S63, perform structured cluster analysis on the health data marked as abnormal, cluster the abnormal data according to the associated feature similarity, and form a risk cluster set C = {c1, c2, ..., c m}, where each risk cluster c j Represents a group of health data points with similar abnormal characteristics. Each risk cluster is risk-weighted and scored according to the characteristic weight to quantify the risk level of abnormal health data.

[0129] S64. Assess risk scores for each risk cluster and adjust intervention strategies according to different risk levels:

[0130]

[0131] Among them, R represents the health risk score, R lowrepresents the low score threshold, R medium represents the scoring threshold, R high represents the high score threshold, exp represents the exponential function, α1, β1, γ1, δ1, λ1, α2, β2, θ2, γ2, δ2, α3, β3, θ3, and δ3 represent adjustment parameters that control the response intensity at different risk levels;

[0132] S65. When the risk score exceeds the set safety threshold, the chain feedback control mechanism is triggered to generate a feedback event. The health monitoring strategy is adjusted accordingly according to the risk level of the feedback event. The feedback event is transmitted to each distributed node through the chain feedback mechanism, and the anomaly detection threshold τ is dynamically updated in each node. i ;

[0133] S66. Under the closed-loop management of the chain feedback control mechanism, the health monitoring strategy is continuously optimized according to the feedback events, and the feedback results of the risk cluster are used to adaptively adjust the anomaly detection parameters of each node by dynamically adjusting the detection threshold and weighted scoring parameters of each distributed node.

[0134] Embodiment 1:

[0135] In order to verify the feasibility of the present invention in implementation, the present invention is applied to a large health management organization. Kangjian Medical Center, a large health management organization, hopes to provide intelligent disease prevention management and monitoring services for its high-risk chronic disease patients. The patient population of Kangjian Medical Center includes patients with various chronic diseases, including hypertension, diabetes, coronary heart disease, etc. At present, the health management of patients is mostly achieved through regular blood pressure measurement, blood sugar testing and drug intervention. However, traditional health monitoring methods are difficult to respond to changes in the patient's health status in a timely manner. Patients often fail to provide timely warnings in risk situations such as sudden hypertension and sharp fluctuations in blood sugar, resulting in serious consequences. The organization hopes to use the method of the present invention to dynamically predict the patient's health status, conduct personalized interventions and real-time abnormal monitoring based on real-time collected data.

[0136] In this embodiment, Kangjian Medical Center uses the multi-source data acquisition module proposed in the present invention to collect patients' health data, including genetic information, real-time vital signs parameters (such as blood pressure, blood sugar, heart rate) and environmental factors (such as air quality, temperature) and other data, and uses hypergraph neural network to feature fuse the potential associations between different data sources. Subsequently, the system predicts the patient's personalized disease risk by generating an inference network, and generates the patient's future health status in different scenarios. The intervention module simulates various intervention strategies according to the changes in the patient's health status, including adjusting diet, increasing or decreasing exercise, etc., to help patients take measures before adverse health trends occur. The system also uses a distributed collaborative anomaly detection module to analyze the patient's real-time data. When an abnormality in health data is detected, the chain feedback control mechanism will respond immediately, generate feedback events and dynamically adjust monitoring parameters.

[0137] In order to verify the beneficial effects of the present invention, Kangjian Medical Center selected 100 hypertensive patients and 100 diabetic patients as the experimental group, and set up a control group with the same number and conditions, using traditional health management methods for control. The experiment was conducted at Kangjian Medical Center for 12 months, with data collection frequency once a day. Before the experiment began, the patient's basic health data was collected and preprocessed, and the feature fusion results were applied to generate the inference network to form a personalized health status prediction model. The following are the data analysis and effect verification results during the experiment.

[0138] In the management of patients with hypertension, the system of the experimental group uses real-time monitoring of blood pressure and heart rate data, combined with environmental data such as temperature and air quality, and extracts the correlation of these features through a hypergraph neural network. For example, when the temperature changes greatly, the patient's blood pressure tends to fluctuate accordingly. The experimental results show that when the temperature of the experimental group drops by more than 10°C, the system predicts two hours in advance that 25% of the patients are at risk of high blood pressure, and generates personalized intervention strategies to prompt patients to replenish water and keep warm. The data show that the system effectively controls the risk increase of patients with hypertension, and the fluctuation range of blood pressure is significantly reduced in the experimental group. Comparative data show that the average high pressure in the experimental group is 135mmHg, and that in the control group is 145mmHg, with significant differences.

[0139] For diabetic patients, the system generates health scenario predictions of future blood sugar changes by monitoring the patient's blood sugar level in real time and combining the patient's daily diet and exercise data. For example, when the system detects that the patient has consumed high-sugar foods or lacks exercise for a long time, it can adjust the intervention strategy in real time to remind the patient to increase exercise or control diet, thereby reducing the risk of sudden increase in blood sugar. During the experiment, the average blood sugar control level of patients in the experimental group was 7.1mmol / L, while that of the control group was 8.2mmol / L. In addition, the system successfully predicted 17 potential hypoglycemia events and generated an alarm 3 hours before the event, reminding patients to replenish sugar in time and avoid the risk of sudden hypoglycemia.

[0140] The chain feedback mechanism of the present invention also demonstrates superiority in abnormal monitoring. For example, in the monitoring of hypertensive patients, the system detected 57 abnormal health data (such as a sudden rise in blood pressure or a sharp drop in blood sugar) during the experiment, and generated feedback events in a timely manner, starting the chain feedback control mechanism. After detecting abnormal data, the system quickly distributes it to the corresponding doctors and caregivers through the authority management mechanism, and adjusts the detection parameters in real time. The data show that the risk response time of patients in the experimental group in the case of abnormal health was 15 minutes, while the average response time of the control group was 2 hours, which significantly shortened the risk management time.

[0141] The data show that through the system of the present invention, the effect of Kangjian Medical Center in disease prevention and management has been significantly improved. The health management compliance of patients in the experimental group increased by 22%, the response time for health abnormalities was shortened to 15 minutes, and the effectiveness of the personalized intervention strategy was also verified. Compared with the control group, the fluctuation range of blood pressure and blood sugar in the experimental group was significantly reduced, and the overall satisfaction of patients increased by 15%. The pilot data of Kangjian Medical Center clearly demonstrates the application advantages of the present invention in disease prevention and management. Through multi-source data fusion, personalized prediction and real-time feedback, more efficient disease monitoring and intervention management are achieved.

[0142] Table 1 Comparison of blood pressure fluctuation data in patients with hypertension

[0143]

[0144] Table 2 Comparison of blood sugar control levels in diabetic patients

[0145]

[0146] Table 3 Health abnormality response time data comparison table

[0147]

[0148]

[0149] It can be seen from the data in Tables 1 to 3 above that after the experimental group applied the method of the present invention, the stability of health indicators and the efficiency of risk management were significantly improved. The fluctuation range of blood pressure in patients with hypertension was significantly reduced in the experimental group, among which the average high pressure in the 12th month was 133mmHg, which was a large gap compared with the 142mmHg in the control group. The blood sugar control of diabetic patients was also more stable. The average blood sugar in the experimental group dropped from 7.8mmol / L in the first month to 6.9mmol / L in the 12th month, while the control group hovered above 8mmol / L, indicating that the system performed well in personalized intervention and dynamic monitoring. In addition, in terms of abnormal health response, the average response time of the experimental group was significantly shortened to 15 minutes, while the control group required about 1 to 2 hours, indicating that the system effectively improved the response speed and reduced the risk of sudden health events. These data results fully demonstrate the timeliness, effectiveness and personalized advantages of the present invention in disease prevention and management.

[0150] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A method for designing a disease prevention management and monitoring system based on an adaptive algorithm, characterized in that: The steps include: S1. Acquire the user's health data through a multi-source data acquisition module, wherein the health data includes genetic information, real-time physical parameters and environmental factors, and input the health data into a hypergraph neural network; S2. Use the hypergraph convolution module of the hypergraph neural network to extract and fuse features of health data, map the health data into a hypergraph structure, identify potential associations between different data sources, and output feature fusion results; S3. Construct a disease risk prediction model based on a generative reasoning network, generate personalized future health status predictions through feature fusion results and simulation of health development paths, and introduce a counterfactual generation module to simulate different health scenarios through changes in a small number of key variables to generate health change hypothetical scenarios; S4, adaptively adjust the personalized intervention strategy through the nested game model. In the first layer of the nested game model, the intervention intensity is adjusted according to the changes in the user's health status and the adoption of suggestions in the hypothetical scenario; S5. In the second layer of the nested game model, the competition and cooperation of lifestyle, nutrition and exercise are simulated, and the optimized intervention combination strategy is generated based on the data in the health change hypothetical scenario and the multi-objective adaptive game driven by Nash equilibrium; S6. Apply the distributed collaborative anomaly detection module to monitor and detect abnormal changes in health data in real time, perform risk-weighted scoring on health data through a structured clustering algorithm, and trigger a chain feedback control mechanism to generate feedback events when health risks are detected; S7. Through the closed-loop management function of the chain feedback mechanism, the disease prevention management strategy is dynamically optimized according to the user's health data, feedback response results and health change hypothesis scenarios, and the authority of each node in the feedback chain is managed based on smart contracts.

2. The method for designing a disease prevention management and monitoring system based on an adaptive algorithm according to claim 1, characterized in that: The S2 specifically includes: S21, normalizing the acquired user health data, including genetic information, real-time vital sign parameters and environmental factors, to form a multi-dimensional feature vector, and establishing a health data node set with the feature vector of each health data type as a node; S22. Based on the similarity between health data sources, different types of hyperedges are defined to connect multiple nodes with potential associations. The connection weight of each hyperedge is determined by feature similarity and data source type to increase the association of similar features. A weight matrix is ​​assigned to each hyperedge to establish a hyperedge set. S23, mapping the health data node set and the hyperedge set to a hypergraph structure G=(V,E), where V represents the health data node set and E represents the hyperedge set; S24. Apply the improved hypergraph convolution operation to the hypergraph structure G, realize complex multi-source data association aggregation by combining convolution aggregation functions, fuse the feature information of each node and iteratively update: Among them, X ′ represents the feature matrix after the hypergraph convolution operation, X represents the input feature matrix, σ represents the nonlinear activation function, G represents the hypergraph association matrix, W represents the weight matrix, and D v represents the degree matrix of node v, D e represents the degree matrix of the hyperedge e, β and α represent the edge connection weight adjustment parameters, N represents the convolution depth, represents the convolution kernel; S25. Perform multi-layer hypergraph convolution aggregation operations, iteratively extract multi-dimensional features of health data, construct an aggregate feature matrix containing multi-layer health data relationships, and retain inter-layer feature information through residual connections: Among them, Z (l+1) represents the feature matrix of the l+1th layer, Z (l) represents the feature matrix of the lth layer, G (l) represents the hypergraph incidence matrix at level l, W (l) represents the master weight matrix of the lth layer, represents the additional weight matrix of the lth layer, λ i represents the additional weight scaling factor, δ represents the association adjustment factor, γ represents the residual weight factor, and I represents the number of additional weight matrices; S26. The final feature fusion matrix obtained after multi-layer hypergraph convolution aggregation is used as the high-dimensional representation of health data, and the output is the feature fusion result of disease risk prediction.

3. The method for designing a disease prevention management and monitoring system based on an adaptive algorithm according to claim 1, characterized in that: The S3 specifically includes: S31, using the feature fusion result generated by the hypergraph convolution as the input data of the generated inference network, and establishing a feature vector set F = {f1, f2, ..., f n }, where each eigenvector f i The feature fusion representation of a specific health data node is represented by inputting the feature vector set into the generator module of the generative reasoning network to construct the baseline health state feature vector S; S32. In the generation module of the generation reasoning network, the reference health state feature vector S is combined with the random noise vector z, and a generator G is used to output multiple simulated health states: in, represents the simulated health state output by the generator G, tanh represents the hyperbolic tangent function, A represents the number of feature vectors, B represents the dimension of the random noise vector, and W ij represents the health status weight matrix of the generator, b ij represents the health state bias term, exp represents the exponential function, ψ k and χ l represents the healthy development kernel function in the generator, K represents the kernel depth, L represents the number of frequency kernels, and ζ represents the healthy regulation factor of the generation process; S33. Introduce a counterfactual generation module in the generative reasoning network, define a set of key health variables, set potential risk factors related to specific diseases in the user's health data, and construct a counterfactual generation framework under different scenarios by applying different amplitudes of changes to each variable in the set of key health variables; S34. In the counterfactual generation module, by controlling the change value of each variable in the key health variable set, the baseline health state is adjusted to generate future health state vectors with different health scenarios: in, represents the future health state vector, ΔR represents the change value of the variable, C represents the total number of key health variables, and W i represents the adaptive weight of the i-th key health variable, μ and ν represent the parameters that control the sensitivity of health status changes, and λ p Represents the response coefficient of different key variables, R p represents the pth variable in the key health variable set, z p represents the pth noise component in the random noise vector, P represents the depth of the key variable response, and Q represents the number of frequency kernels of the key variable; S35, integrating the health status vectors generated by multiple health scenarios into a prediction matrix Each of these Represents a hypothetical scenario of health change, reflecting the future health status under different health scenarios and changes in key variables S36. Use the prediction matrix of the health change hypothesis scenario as the final output of the personalized health status prediction to generate a personalized future health status prediction.

4. The method for designing a disease prevention management and monitoring system based on an adaptive algorithm according to claim 1, characterized in that: The S4 specifically includes: S41, establishing a user health status monitoring module of the first-layer game model, by collecting the user's health data in real time, generating a current health status feature vector for dynamically analyzing the changing trend of the health status; S42, based on the monitored health status characteristics, compare the current health status with the historical health data, identify the key change nodes of the user's health status, and provide adjustment basis for the intervention strategy; S43. Introduce hypothetical health change scenarios based on the user's health status changes, generate multiple health scenarios by simulating the user's possible health development path, evaluate the possible health risks in each scenario, and design different adjustment plans for the intervention strategy; S44. Analyze the adoption of intervention suggestions in the health change hypothetical scenario, and evaluate the effectiveness of each intervention plan based on the user's acceptance and feedback of different intervention suggestions; S45, generating an adaptive intervention intensity adjustment plan based on the real-time changes in the user's health status and the adoption of intervention suggestions; S46. Apply the adjusted intervention intensity to the personalized intervention strategy, update the strategy output of the first-level game model in real time, and regularly evaluate the intervention effect.

5. The method for designing a disease prevention management and monitoring system based on an adaptive algorithm according to claim 1, characterized in that: The S5 specifically includes: S51. Construct the utility functions of the lifestyle intervention module, the nutrition intervention module and the exercise intervention module. Based on the data in the health change hypothesis scenario, set the objective function of each intervention module, quantify the degree of impact on the user's health status, and generate the expected health benefits of each module according to the specific intervention strategy. S52. Define the strategy space between intervention modules to form a set of lifestyle intervention modules, nutrition intervention modules and exercise intervention strategies X i , where i = 1, 2, and 3 represent lifestyle intervention module, nutrition intervention module, and exercise intervention module respectively. The strategy set X of each intervention module i Affected by resource constraints, user acceptance and implementation costs, the set of feasible solutions is maximized by optimizing the strategy space; S53. Construct a multi-objective adaptive game model, introduce the Nash equilibrium mechanism, and define the benefit function U of each intervention module. i (x i ,x -i ), where x i represents the strategy of intervention module i, x -i Represents a set of strategies for other intervention modules: Among them, U i (x i ,x -i ) represents the profit function of intervention module i, exp represents the exponential function, ω i represents the intervention effect weight, E i (x i ) represents the health benefit function of intervention module i, λ ij represents the synergistic effect coefficient, S ij (x i ,x j ) represents the cooperation benefit function between intervention module i and intervention module j, C i (x i ) represents the intervention cost of intervention module i, τ represents the cost adjustment factor, D represents the number of features used in the benefit function in the strategy set, M represents the strategy dimension of the intervention module, ψ ij , and θ l represents the strategy adjustment parameter, x ik represents the k-th strategy feature for the profit function in intervention module i, x il represents the lth strategy feature in intervention module i; S54. Use the evolutionary game algorithm to solve the multi-objective adaptive game model, and calculate the benefit gradient optimization strategy of the intervention module through each iteration: in, represents the t+1th iteration strategy of intervention module i, represents the t-th iteration strategy of intervention module i, η represents the learning rate, Y represents the number of frequency kernels used to respond to different health status characteristics during the strategy update process, r represents the number of features used for adjustment in the strategy update, t represents the cumulative effect depth in the strategy update, and x iq represents the qth strategy feature in intervention module i, x is represents the sth strategy feature in intervention module i, x iu represents the u-th strategy feature in intervention module i, represents the adjustment factor in the strategy update; S55. During the iteration process, the synergy coefficient is adaptively adjusted according to user health feedback and intervention effects; when the synergy effect is enhanced, the synergy coefficient is increased to promote cooperation; when the synergy effect is weakened, the synergy coefficient is reduced to strengthen competition; S56. The iteration process continues until the strategies of each intervention module converge to the Nash equilibrium point, generating the final optimized intervention combination strategy.

6. The method for designing a disease prevention management and monitoring system based on an adaptive algorithm according to claim 1, characterized in that: The S6 specifically includes: S61. Construct a distributed anomaly detection model for health data, distribute multi-source health data to different distributed nodes according to data source, feature type and time series, and form a multi-level distributed feature vector set D = {d1, d2, ..., d n }, where d i is the i-th feature data node, which represents the node representation of a specific health parameter in the distributed structure; S62: Apply the collaborative anomaly detection algorithm in the distributed nodes and set the anomaly detection threshold τ of each feature data node. i , detect abnormal data points according to the real-time changes of health data. When the characteristic value of a node is detected to exceed the set abnormal detection threshold τ i When , the node is marked as abnormal, the abnormal data is recorded and sent to the central control node; S63, perform structured cluster analysis on the health data marked as abnormal, cluster the abnormal data according to the associated feature similarity, and form a risk cluster set C = {c1, c2, ..., c m }, where each risk cluster c j Represents a group of health data points with similar abnormal characteristics. Each risk cluster is risk-weighted and scored according to the characteristic weight to quantify the risk level of abnormal health data. S64. Assess risk scores for each risk cluster and adjust intervention strategies according to different risk levels: Among them, R represents the health risk score, R low represents the low score threshold, R medium represents the scoring threshold, R high represents the high score threshold, exp represents the exponential function, α1, β1, γ1, δ1, λ1, α2, β2, θ2, γ2, δ2, α3, β3, θ3, and δ3 represent adjustment parameters that control the response intensity at different risk levels; S65. When the risk score exceeds the set safety threshold, the chain feedback control mechanism is triggered to generate a feedback event. The health monitoring strategy is adjusted accordingly according to the risk level of the feedback event. The feedback event is transmitted to each distributed node through the chain feedback mechanism, and the anomaly detection threshold τ is dynamically updated in each node. i ; S66. Under the closed-loop management of the chain feedback control mechanism, the health monitoring strategy is continuously optimized according to the feedback events, and the feedback results of the risk cluster are used to adaptively adjust the anomaly detection parameters of each node by dynamically adjusting the detection threshold and weighted scoring parameters of each distributed node.

Images