Aggregate heating prediction method and system based on AI time sequence prediction model
By constructing an AI time-series prediction model and integrating community medical records and population flow data, early warning information on clustered fever events is generated, solving the problem of accurate prediction and intervention of community-level clustered fever events and improving the efficiency of public health management.
Patent Information
- Application Number
- CN202511118027.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-11
- Publication Date
- 2025-11-21
AI Technical Summary
Existing technologies are insufficient for accurate prediction and dynamic intervention of community-level clustered fever events, and lack intelligent early warning methods that integrate multi-source data, leading to a high incidence of public health emergencies.
By constructing an AI-based time-series prediction model, integrating community medical records, population flow data, and real-time fever monitoring data, and using a multi-head attention mechanism to model multi-dimensional data, clustered fever early warning information is generated, and a review mechanism for community fever early warning institutions is introduced.
It enables early detection, accurate warning, and rapid response to community-acquired fever outbreaks, improving the efficiency of public health management and possessing strong generalization ability and model interpretability.
Smart Images
Figure CN120998537A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of artificial intelligence and data analysis, and particularly relates to an aggregation fever prediction method and system based on an AI time series prediction model. BACKGROUND
[0002] With the development of infectious disease prevention and control normalization, higher requirements are put forward for early identification and early warning of health risks in community, school and enterprise personnel intensive places. As one of the important signals before the outbreak of infectious diseases, aggregation fever has the characteristics of strong suddenness, fast spread and wide influence range. If it is not discovered and intervened in time, it is easy to cause regional public health events. The current common fever monitoring methods mainly rely on body temperature guns, manual reporting and single-point infrared temperature measurement methods, which have problems of data lag, insufficient coverage and lack of trend judgment, and it is difficult to realize accurate prediction and dynamic intervention of potential aggregation fever events. In recent years, artificial intelligence technology has made significant progress in time series modeling, especially the time series prediction model based on artificial intelligence, which has shown strong trend identification and reasoning ability in the fields of medical treatment, meteorology and transportation. However, there is still a lack of a multi-source modeling scheme that can integrate community medical record information, population flow characteristics and fever trend data for intelligent prediction and early warning of community-level aggregation fever risk. Therefore, there is an urgent need for a data-driven early warning method based on AI time series model and combining multi-dimensional health and environmental characteristics to improve the discovery efficiency and timeliness of intervention of aggregation fever events. SUMMARY
[0003] To solve the above problems in the prior art, the application provides an aggregation fever prediction method based on an AI time series prediction model,
[0004] The object of the application can be achieved by the following technical solutions:
[0005] S1: Obtain community medical records from a community medical institution database, and based on the community medical records, integrate the obtained data into multi-dimensional fever source information through a community management platform according to population flow data and aggregation fever data;
[0006] S2: Train an artificial intelligence time series prediction model based on the community medical records and population flow data; input the obtained multi-dimensional fever source information into the trained artificial intelligence time series prediction model to obtain the cause of aggregation fever events within the time series;
[0007] S3: In the community fever prediction management platform, combine the cause of aggregation fever events within the time series with the community medical records through an aggregation fever early warning model to generate aggregation fever early warning information;
[0008] S4: The community fever prediction management platform pushes the gathering fever early warning information to a community fever early warning institution; the community fever early warning institution audits the gathering fever early warning information according to the historical gathering fever data, and after the audit is passed, pushes the gathering fever early warning information to community residents.
[0009] Specifically, the artificial intelligence time series prediction model constructs a multivariate time series by inputting a historical fever event data set, and introduces a multi-head attention processing module to mine key time windows and key feature variables, and outputs fever number prediction values in a specified prediction time window and a spatial distribution risk heat map.
[0010] Specifically, the community medical record archive acquisition method is:
[0011] S101: Group and integrate medical record data according to community area and time dimension through a heat gathering flow label model, and construct a community-level health archive index containing individual health status, medical history label and personnel distribution information;
[0012] S102: Connect a community medical institution database, and acquire the community medical record archive according to the community-level health archive index.
[0013] Specifically, the cause of the gathering fever event in the time sequence is constructed by identifying the time period with a sudden change inflection point or a significantly increased growth slope of the number of fever people in the gathering fever time sequence, and constructing an abnormal window sequence; according to the abnormal time window sequence, calling the event log, video monitoring record, access control data and environmental sensor information in the community management platform, extracting the personnel gathering, space closure, environmental anomaly and flow concentration label in the corresponding time period, and generating the gathering fever time reason in the time sequence according to the label, which is specifically defined as:
[0014] K(t)=log[e tX ],
[0015] Where X is a risk variable, and the point where the first derivative is zero is approximated by a saddle point:
[0016] K′(t)=0,
[0017] The tail probability distribution is approximated by a saddle point formula:
[0018]
[0019] Where t is a saddle point, and K''(t) is the second derivative.
[0020] Specifically, the community fever prediction management platform runs and includes:
[0021] S201: Call the deployment of artificial intelligence time series prediction model, process the multi-dimensional heat source information, output the reason of the time series of the gathering heat event;
[0022] S202: Joint coding of the heat event reason and community medical records archive, building high-dimensional input features for risk analysis, generating gathering heat warning information containing time, space, risk level, and possible inducement;
[0023] S203: Interface with community heat warning agencies, push the generated gathering heat warning information, and receive feedback results.
[0024] Specifically, the multi-head attention processing module receives high-dimensional vector input from time series event features and community medical record semantic embedding, and models the interaction and dependence between various time series factors and health risk factors according to multiple sets of attention heads in parallel. The calculation formula of attention weight is as follows:
[0025]
[0026] Where Q i is the query vector, corresponding to the heat time feature code, K j is the key vector, corresponding to the community medical record risk factor embedding, d k is the dimension of the key vector, used for scaling and normalization processing, a ij is the attention weight of the i-th event to the j-th medical record feature.
[0027] Specifically, the gathering heat warning model receives the gathering heat event reason output by the artificial intelligence time series prediction model, structures the community medical record archive, extracts multi-dimensional health risk features including chronic disease markers, historical heat records, and personnel labels, generates a high-dimensional gathering risk representation vector, and based on the risk representation vector, combines the event reason and personnel features to generate gathering heat warning information.
[0028] Specifically, the gathering heat warning information includes the time period and geographic area of the gathering heat event, the main event reason label triggering the warning, and the system-generated intervention suggestions, including health management measures, environmental ventilation reminders, and personnel detection prompts.
[0029] Specifically, the community warning agency receives the gathering heat warning information pushed by the community heat prediction management platform, performs structural analysis and archiving on the information content, and calls the heat gathering rule model to judge the accuracy of the warning information. If the audit conditions are met, it is marked as passed, and the gathering heat warning information that passes the audit is pushed to the corresponding community resident terminal, community grid worker platform, and health and epidemic prevention agencies.
[0030] Specifically, a clustered fever prediction system based on an AI time-series prediction model, used to perform any one of the methods described in claims 1-9, is characterized by comprising:
[0031] Data acquisition and integration module: used to obtain community medical records from the community medical institution database and integrate data from individual body temperature detection devices, infrared temperature measurement systems, population movement tracking platforms, environmental sensors and health check-in systems to construct a standardized, multi-source time-series input dataset;
[0032] Time series modeling and prediction module: used to deploy and run artificial intelligence-based time series prediction models, model and infer the input multidimensional time series, and output the predicted value of the number of people with fever in the region within a specified future time window;
[0033] Risk assessment and early warning generation module: It is used to comprehensively calculate the regional fever risk level based on the prediction results and event fusion characteristics, and automatically generate structured clustered fever early warning information, including risk level, time period, triggering event label and intervention suggestions;
[0034] Information push and visualization module: It is used to push the generated early warning information to responsible personnel and users through the community management platform, mobile app and public health interface, and at the same time display fever trend map, risk heat map and spatial transmission prediction map on the visualization screen.
[0035] The beneficial effects of this invention are as follows:
[0036] By integrating community medical records, population flow data, and real-time fever monitoring data, and constructing spatiotemporal fever source information based on multi-dimensional data fusion, this invention effectively improves the structured utilization efficiency of community-level health data. Based on an artificial intelligence time-series prediction model combined with a multi-head attention mechanism, it can dynamically identify key time periods and risk factors, accurately predicting future fever trends under multi-variable input conditions and outputting regional risk heat maps. It possesses strong generalization ability and model interpretability. Through a clustered fever early warning model, the AI prediction results are jointly encoded with community medical records to uncover potential causal relationships and automatically generate early warning information covering the time period, spatial distribution, triggering factors, and intervention suggestions for fever events. A community fever early warning agency review mechanism is introduced, combining AI intelligent judgment with human supervision to ensure the accuracy and enforceability of early warning information. Finally, the early warning information can be accurately pushed and visualized through community platforms and multi-terminal interfaces. This invention enables early detection, accurate early warning, and rapid response to community clustered fever events, effectively improving the efficiency of grassroots public health management and possessing broad application and promotion value and social benefits. Attached Figure Description
[0037] For the convenience of those skilled in the art to understand, the present application is further described below in conjunction with the drawings.
[0038] Figure 1 The flowchart of the AI time series prediction model-based clustering fever prediction method and system of the present application. DETAILED DESCRIPTION
[0039] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined application purpose, the specific embodiments, structures, features and effects according to the present application are described in detail below in conjunction with the drawings and preferred embodiments.
[0040] Please refer to Figure 1 An AI time series prediction model-based clustering fever prediction method and system:
[0041] S1: Obtain community medical records from a community medical institution database, and based on the community medical records, integrate the obtained data into multi-dimensional fever source information through a community management platform according to population flow data and clustering fever data;
[0042] S2: Train an artificial intelligence time series prediction model based on the community medical records and population flow data; input the obtained multi-dimensional fever source information into the trained artificial intelligence time series prediction model to obtain the reason for clustering fever events within a time sequence;
[0043] S3: In the community fever prediction management platform, combine the clustering fever time reason within the time sequence with the community medical records through a clustering fever early warning model to generate clustering fever early warning information;
[0044] S4: The community fever prediction management platform pushes the clustering fever early warning information to a community fever early warning institution; the community fever early warning institution audits the clustering fever early warning information according to the historical clustering fever data, and after the audit is passed, pushes the clustering fever early warning information to community residents.
[0045] Specifically, the artificial intelligence time series prediction model constructs a multivariate time series by inputting a historical fever event data set, and introduces a multi-head attention processing module to mine key time windows and key feature variables, and outputs fever number prediction values and spatial distribution risk heat maps within a specified prediction time window.
[0046] Specifically, the method for obtaining the community medical records is:
[0047] S101: Group and integrate medical record data according to community area and time dimension through a heat flow marking model, and construct a community-level health record index containing individual health status, medical history label and personnel distribution information;
[0048] S102: Connect the community medical institution database, and obtain the community medical record according to the community health record index.
[0049] Specifically, the cause of the gathering fever event in the time sequence is obtained by identifying the time period with a sudden inflection point or a significantly increased growth slope of the number of fever patients in the gathering fever time sequence, constructing an abnormal window sequence; according to the abnormal time window sequence, calling the event log, video monitoring record, access control data and environmental sensor information in the community management platform, extracting the personnel gathering, space closure, environmental anomaly and flow concentration labels in the corresponding time period, and generating the gathering fever time reason in the time sequence according to the labels, which is specifically defined as:
[0050] K(t)=log[e tX ],
[0051] Where X is a risk variable, and the saddle point is used to approximate the point where the first derivative is zero:
[0052] K′(t)=0,
[0053] The tail probability distribution is approximated by the saddle point formula:
[0054]
[0055] Where t is the saddle point, and K”(t) is the second derivative.
[0056] Specifically, the community fever prediction management platform runs includes:
[0057] S201: Call the deployed artificial intelligence time sequence prediction model, process the multi-dimensional fever source information, and output the cause of the gathering fever event in the time sequence;
[0058] S202: Jointly encode the fever event reason and the community medical record to construct high-dimensional input features for risk analysis, and generate gathering fever warning information containing time, space, risk level and possible causes;
[0059] S203: Interface with the community fever warning mechanism, push the generated gathering fever warning information, and receive the audit feedback result.
[0060] In this rural community healthcare implementation, fever case data is obtained from county-level hospitals via API, environmental data from local weather stations, and population flow data from mobile operators. User health status feedback is collected via a mobile application. The data processing module employs a distributed computing framework to clean and normalize the data, extracting features such as case time series and environmental factor fluctuations. The time series prediction module uses an AI-based time series prediction model, with training data covering January 2024 to June 2025. After inputting preprocessed data, it predicts the risk of clustered fever outbreaks over the next 14 days. The early warning module generates a visual risk map, which is pushed to public health management nodes. After review by management personnel, early warnings are pushed to residents via the mobile application, providing reminders of protective measures. The system supports online incremental learning, updating the model every 100 new data points received and adjusting parameters within 30 seconds. This implementation successfully achieves real-time early warning in a rural healthcare resource-constrained environment, effectively reducing the risk of clustered fever outbreaks spreading.
[0061] Specifically, the multi-head attention processing module receives high-dimensional vector inputs from temporal event features and semantic embeddings of community medical records, and models the interaction dependencies between various temporal factors and health risk factors in parallel using multiple attention heads. The formula for calculating the attention weight is as follows:
[0062]
[0063] Q i For the query vector, corresponding to the heating time feature encoding, K j Let d be the key vector, corresponding to the community medical record risk factor embedding. k Let a be the dimension of the key vector, used for scaling and normalization. ij Let be the attention weight of the i-th event on the j-th medical record feature.
[0064] Specifically, the clustered fever early warning model receives the causes of clustered fever events output by the artificial intelligence time series prediction model, performs structured encoding on the community medical records, extracts multidimensional health risk features including chronic disease markers, historical fever records, and personnel tags, generates a high-dimensional clustered risk representation vector, and generates clustered fever early warning information based on the risk representation vector and the event causes and personnel characteristics.
[0065] In this embodiment, the community fever early warning mechanism serves as a public health management node, receives the aggregated fever early warning information pushed by the prediction platform, and contains the risk probability, confidence interval and high-risk area distribution map. The main function is to verify the accuracy and applicability of the early warning information. The management personnel evaluate the prediction results in combination with the local actual situation, confirm whether to trigger the early warning, and the mechanism verifies the consistency of the information through professional judgment. After approval, the early warning is pushed to the residents and health departments through the Web interface and SMS, coordinates the follow-up prevention and control measures to ensure that the early warning is converted into actual action. Through the auditing and distribution of the community fever early warning mechanism, AI prediction and community response are bridged, and the efficiency and pertinence of urban public health decision-making are improved.
[0066] Specifically, the aggregated fever early warning information includes the time period and geographical area of the aggregated fever event, the main event reason label triggering the early warning, and the system-generated intervention suggestion, including health management measures, environmental ventilation reminders, and personnel detection prompts.
[0067] Specifically, the community early warning mechanism receives the aggregated fever early warning information pushed by the community fever prediction management platform, performs structural analysis and archive processing on the information content, and calls the heat aggregation rule model to audit and judge the accuracy of the early warning information. If the audit conditions are met, it is marked as passed, and the passed aggregated fever early warning information is pushed to the corresponding community resident terminal, community grid employee platform and health and epidemic prevention mechanism.
[0068] Specifically, an aggregated fever prediction system based on an AI time series prediction model is used to perform any method as claimed in claims 1-9, characterized in that it comprises:
[0069] Data collection and integration module: used to obtain community medical records from community medical institution databases, and fuse data from individual body temperature detection equipment, infrared temperature measurement system, population flow tracking platform, environmental sensor and health clock-in system to construct a standardized, multi-source time series input dataset;
[0070] Time series modeling and prediction module: used to deploy and run an artificial intelligence-based time series prediction model to model and reason the input multi-dimensional time series, and output the regional fever number prediction value in the specified future time window;
[0071] Risk assessment and early warning generation module: used to calculate the regional fever risk level based on the prediction results and event fusion features, automatically generate structured aggregated fever early warning information, including risk level, time period, inducing event label and intervention suggestion;
[0072] Information push and visualization module: used for pushing the generated early warning information to the responsible personnel and users through the community management platform, mobile terminal App and public health interface, and displaying fever trend graph, risk heat map and space propagation prediction graph in the visualization large screen.
[0073] The above is only the preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above with the preferred embodiment, it is not intended to limit the present application. Any person skilled in the art can make slight changes or modifications to the above disclosed technical content to obtain equivalent embodiments with equivalent changes without departing from the technical solution of the present application. Any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application still belongs to the scope of the technical solution of the present application.
Claims
1. A method for predicting clustered heat generation based on an AI time-series prediction model, characterized in that, include: S1: Obtain community medical records from the community medical institution database, and based on the community medical records, integrate the obtained data into multi-dimensional fever source information through the community management platform according to population flow data and clustered fever data; S2: Based on the community medical records and population flow data, train an artificial intelligence time series prediction model; input the acquired multidimensional fever source information into the trained artificial intelligence time series prediction model to obtain the causes of clustered fever events within the time series; S3: Within the community fever prediction and management platform, the cause of the clustered fever within the specified time series is jointly encoded with the community medical records through the clustered fever early warning model to generate clustered fever early warning information; S4: The community fever prediction and management platform pushes the cluster fever early warning information to the community fever early warning agency; The community fever early warning agency reviews the cluster fever early warning information based on the historical cluster fever data, and pushes the cluster fever early warning information to community residents after the review is approved.
2. The method according to claim 1, characterized in that, The AI-based time series prediction model constructs a multivariate time series by inputting a historical fever event dataset, and introduces a multi-head attention processing module to mine key time windows and key feature variables, outputting the predicted number of fever cases and a spatial distribution risk heat map within the specified prediction time window.
3. The method according to claim 1, characterized in that, The method for obtaining the community medical records is as follows: S101: Using the heat flow labeling model, medical record data is grouped and integrated according to community area and time dimensions to construct a community-level health record index containing individual health status, medical history tags and personnel distribution information; S102: Connect to the community medical institution database and obtain the community medical records based on the community-level health record index.
4. The method according to claim 1, characterized in that, The causes of clustered fever events within the time series are identified by constructing an abnormal window sequence in which the number of fever cases suddenly changes or the growth rate increases significantly. Based on the abnormal time window sequence, event logs, video surveillance records, access control data, and environmental sensor information from the community management platform are called to extract tags for personnel gathering, spatial closure, environmental anomalies, and dense mobility within the corresponding time period, and the causes of clustered fever events within the time series are generated based on the tags.
5. The method according to claim 1, characterized in that, The operation of the community fever prediction and management platform includes: S201: Call the deployed artificial intelligence time series prediction model to process the multidimensional heat source information and output the cause of the clustered heat events in the time series; S202: Jointly encode the causes of the fever events with community medical records to construct high-dimensional input features for risk analysis and generate clustered fever early warning information that includes time, space, risk level, and possible causes; S203: Connect with community fever warning agencies, push out generated cluster fever warning information, and receive their review feedback results.
6. The method according to claim 4, characterized in that, The multi-head attention processing module receives high-dimensional vector inputs from temporal event features and semantic embeddings of community medical records, and models the interaction dependencies between various temporal factors and health risk factors in parallel based on multiple sets of attention heads.
7. The method according to claim 4, characterized in that, The clustered fever early warning model receives the causes of clustered fever events output by an artificial intelligence time-series prediction model, performs structured encoding on the community medical records, extracts multidimensional health risk features including chronic disease markers, historical fever records, and personnel tags, generates a high-dimensional clustered risk representation vector, and generates clustered fever early warning information based on the risk representation vector and the event causes and personnel characteristics.
8. The method according to claim 7, characterized in that, The cluster fever warning information includes: the time period and geographical area where the cluster fever event occurred, the main event cause label that triggered the warning, and the intervention suggestions generated by the system, including health management measures, environmental ventilation reminders, and personnel testing prompts.
9. The method according to claim 1, characterized in that, The community early warning agency receives clustered fever warning information pushed from the community fever prediction and management platform, performs structural analysis and archiving of the information content, and calls the heat clustering rule model to review and judge the accuracy of the warning information. If the review conditions are met, it is marked as approved and the approved clustered fever warning information is pushed to the corresponding community resident terminal, community grid worker platform and health and epidemic prevention agency.
10. A clustered heat generation prediction system based on an AI time-series prediction model, used to perform the method as described in any one of claims 1-9, characterized in that, include: Data acquisition and integration module: used to obtain community medical records from the community medical institution database and integrate data from individual body temperature detection devices, infrared temperature measurement systems, population movement tracking platforms, environmental sensors and health check-in systems to construct a standardized, multi-source time-series input dataset; Time series modeling and prediction module: used to deploy and run artificial intelligence-based time series prediction models, model and infer the input multidimensional time series, and output the predicted value of the number of people with fever in the region within a specified future time window; Risk assessment and early warning generation module: It is used to comprehensively calculate the regional fever risk level based on the prediction results and event fusion characteristics, and automatically generate structured clustered fever early warning information, including risk level, time period, triggering event label and intervention suggestions; Information push and visualization module: It is used to push the generated early warning information to responsible personnel and users through the community management platform, mobile app and public health interface, and at the same time display fever trend map, risk heat map and spatial transmission prediction map on the visualization screen.