A health monitoring and early warning method and system based on multi-source data fusion

Abstract

The application discloses a health monitoring and early warning method and system based on multi-source data fusion, relates to the field of medical information technology, and comprises the following steps: performing health semantic mapping and medical correlation marking on multi-source monitoring health data to form a correlation health data set; identifying health data missing items with missing data in the correlation health data set and outputting user health completion data; extracting the correlation relationship of the user health completion data and constructing a health status correlation graph; identifying the state change result of the health data node in the health status correlation graph in the time dimension and the correlation influence transmission path in the data correlation relationship, generating health warning data; converting the health warning data into early warning prompt information and outputting health intervention suggestions. Through the cooperation of medical correlation marking processing and cross-source inference supplement, the application realizes the correlation consistency construction and data integrity improvement of multi-source health data.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, and in particular to a health monitoring and early warning method and system based on multi-source data fusion. Background Technology

[0002] In recent years, with the development of medical information and the popularization of wearable health monitors, smart homes, and medical testing equipment, continuous monitoring and risk warning of individual health status have gradually become research hotspots. In the field of medical and health information processing, users' physiological indicators, behavioral data, and environmental data are typically acquired through multi-source data collection methods. These data are then combined with data fusion and analysis methods to comprehensively assess users' health status. Simultaneously, machine learning and data-driven methods are introduced to perform correlation analysis and trend identification on health data, enabling early detection and warning of health risks. This has gradually formed a methodological system centered on multi-source data fusion, health status analysis, and early warning intervention, which has been widely applied in scenarios such as chronic disease management, home health monitoring, and telemedicine.

[0003] However, existing methods have shortcomings in the correlation processing of multi-source health data and the time-series analysis of health status. Current methods often focus on simple integration or independent analysis of multi-source monitoring health data, lacking mechanisms for health semantic mapping and correlation labeling. This results in unclear relationships between different data points, making it difficult to form a unified set of correlated health data, thus affecting the data fusion effect. In the health status analysis process, the lack of organization and analysis methods based on health data nodes and data correlations makes it difficult to accurately identify anomalies in health status changes and generate effective health early warning information, limiting the accuracy of health monitoring and early warning and the effectiveness of intervention decisions. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a health monitoring and early warning method based on multi-source data fusion to address the shortcomings in multi-source health data correlation processing and health status time-series analysis.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: Firstly, this invention provides a health monitoring and early warning method based on multi-source data fusion, comprising: collecting multi-source monitoring health data; performing health semantic mapping and medical association labeling on the multi-source monitoring health data to form an associated health data set; identifying missing health data items in the associated health data set; extracting candidate associated data from the associated health data set based on the association labeling of the missing health data items; performing cross-source inference to supplement the candidate associated data with the missing health data items; outputting user health completion data; extracting association relationships from the user health completion data; outputting health data nodes and data association relationships; based on the state continuity of health data nodes in the time dimension, performing consistency constraint screening and association strength adjustment on the data association relationships; outputting optimized data association relationships; and using... Based on health data nodes and optimized data relationships, a health status correlation graph is constructed. The graph identifies the temporal changes in the health data nodes and their impact transmission paths within the data relationships. A bidirectional propagation control mechanism with trigger and inhibition constraints is applied to these paths to construct a state impact propagation sequence. The sequence is then evaluated for path convergence and differentiation, and the results are output. Based on these results, the evolutionary information of the correlation relationships within the propagation sequence is extracted. This evolutionary information is then categorized into correlation patterns and health anomalies are extracted to generate health early warning data. This data is then converted into warning alerts, and intervention strategies are matched to user health based on these alerts, resulting in health intervention recommendations.

[0007] As a preferred embodiment of the health monitoring and early warning method based on multi-source data fusion described in this invention, the specific steps for forming the associated health data set are as follows: Extract status and time information from multi-source health monitoring data, annotate the health meaning, and output health semantic mapping data; Based on health semantic mapping data, the correspondence between various monitoring health data in multi-source monitoring health data is identified, and associated labels are generated to produce associated labeled data; Based on the associated labeled data, the multi-source monitoring health data is structured and organized to form an associated health data set.

[0008] The multi-source health monitoring data includes user physiological indicator data, user behavior data, and user environment data.

[0009] As a preferred embodiment of the health monitoring and early warning method based on multi-source data fusion described in this invention, the specific steps for extracting candidate correlation data are as follows: Scan and locate missing health data items within the associated health data set; Using missing health data items as the tag index objects, retrieve the associated tags corresponding to the tag index objects from the missing health data items; By using association tags, the association paths of the associated health data set are parsed, and a multi-source association path set is output. Based on each multi-source association path in the multi-source association path set, health data is aggregated from the associated health data set to generate candidate association data.

[0010] As a preferred embodiment of the health monitoring and early warning method based on multi-source data fusion described in this invention, the specific steps for outputting user health completion data are as follows: The candidate related data is associated with the missing health data items, and the associated content data that matches the missing health data items in the candidate related data is identified. Cross-source relationship deduction and content supplementation are performed on related content data, and the supplemented content results are added to the missing items of health data to form complete health data; The missing health data is filled into the corresponding missing positions in the associated health data set, and the association is organized to complete the data, and the user's health data is output.

[0011] As a preferred embodiment of the health monitoring and early warning method based on multi-source data fusion described in this invention, the specific steps for outputting health data nodes and data association relationships are as follows: Identify monitoring targets from user health completion data and output a set of monitoring targets; Based on the set of monitored objects, identify the changing relationships and influence transmission relationships in the user health data completion data, and integrate them to form a set of related relationships; The monitored objects in the set of monitored objects are taken as health data nodes, and the corresponding change relationships and impact transmission relationships in the set of relationships are taken as data relationships.

[0012] As a preferred embodiment of the health monitoring and early warning method based on multi-source data fusion described in this invention, the specific steps for constructing the health status correlation map are as follows: The status information of health data nodes at different time locations is expanded in chronological order, and the status information at adjacent time locations is checked for continuity, outputting status continuity data; Based on the state continuity data, the node connections in the data association are verified item by item. Connections consistent with the state continuity data are retained and inconsistent connections are eliminated to form a constraint-filtered association relationship. Based on the constraint-based screening of association relationships, the state differences of the corresponding health data nodes at different time locations are extracted for each data association relationship, and the association strength of each data association relationship is determined based on the state differences. Based on the strength of the association, the association relationships of the constraint screening are adjusted to form the adjusted data association relationships; The data association relationships were adjusted and uniformly organized according to the connection direction and connection level between health data nodes to obtain optimized data association relationships. Using health data nodes as graph nodes and optimizing data relationships as graph node connections, the connection structure between health data nodes is organized and constructed to form a health status association graph.

[0013] As a preferred embodiment of the health monitoring and early warning method based on multi-source data fusion described in this invention, the specific steps of constructing the state influence propagation sequence are as follows: The health data nodes in the health status association graph are scanned sequentially according to the time dimension to extract status change information, which is then organized sequentially and the status change results are output. Based on the state change results, the data association relationships in the health state association map are analyzed by change mapping, the influence transmission direction and influence transmission range of the state change results along the data association relationships are identified, and the association influence transmission path is integrated. Trigger constraint control is applied to the positive influence transmission direction in the associated influence transmission path, and suppression constraint control is applied to the negative feedback influence direction to form a controlled influence transmission path; The state change results are organized sequentially based on the controlled influence propagation path to form a state influence propagation sequence.

[0014] As a preferred embodiment of the health monitoring and early warning method based on multi-source data fusion described in this invention, the following steps are taken: the path convergence determination and path differentiation identification refer to distinguishing and determining the convergence characteristics and diffusion characteristics of the propagation path based on the degree of node convergence and branch expansion of each propagation path in the state influence propagation sequence, and outputting the determination and identification results. The extraction of the correlation evolution information in the state influence propagation sequence refers to determining the corresponding propagation path in the state influence propagation sequence based on the judgment and identification results, and sequentially reading the changes in the data correlation between each health data node along the propagation path as the correlation evolution information; The aforementioned correlation pattern classification and health anomaly extraction refer to classifying the correlation evolution information according to the judgment and identification results, and identifying the correlation patterns reflecting abnormal health changes from the pattern classification as health early warning data.

[0015] As a preferred embodiment of the health monitoring and early warning method based on multi-source data fusion described in this invention, the specific steps for outputting health intervention suggestions are as follows: Transform health warning data into information representation and generate warning alerts; To match early warning information with corresponding intervention strategies, strategy matching data is generated. Based on the strategy matching data, the user's health is analyzed to generate health intervention suggestions.

[0016] Secondly, this invention provides a health monitoring and early warning system based on multi-source data fusion, comprising: The data acquisition module collects health data from multiple sources, performs health semantic mapping and medical association labeling on the multi-source health data, and forms a set of associated health data. The data completion module is used to identify missing health data items in the associated health data set, extract candidate related data from the associated health data set based on the association tags of the missing health data items, perform cross-source inference to supplement the missing health data items with the candidate related data, and output the user's health completion data. The graph construction module is used to extract the relationships from the user's health completion data, output the health data nodes and data relationships, and construct the health status relationship graph based on the health data nodes and data relationships. The health analysis module identifies the state changes of health data nodes in the health status correlation graph over time and the transmission paths of their correlation effects in the data relationships. It then implements bidirectional propagation control of the correlation effect transmission paths using trigger constraints and inhibition constraints, constructs a state influence propagation sequence, performs path convergence determination and path differentiation identification on the state influence propagation sequence, outputs the determination and identification results, extracts the correlation evolution information in the state influence propagation sequence based on the determination and identification results, classifies the correlation patterns and extracts health anomalies from the correlation evolution information, and generates health early warning data. The intervention suggestion module is used to convert health warning data into warning prompts, match intervention strategies for users' health based on the warning prompts, and output health intervention suggestions.

[0017] The beneficial effects of this invention are as follows: By synergistically combining medical association labeling processing and cross-source inference supplementation, it achieves the construction of consistent associations and improves data integrity of multi-source health data. Through health semantic mapping and medical association labeling of multi-source monitoring health data, a unified correspondence is established between health data from different sources, providing a standardized foundation for data processing. Based on the association labels, candidate related data are extracted for missing health data items, and cross-source inference supplementation is performed, achieving association completion of missing data positions, improving data usability and analytical support capabilities. A health status association graph is constructed, and combined with temporal association analysis methods to identify the evolution of association relationships, realizing the structured expression and dynamic analysis of health status changes. This enables the generation of accurate health early warning data and supports intervention strategy matching, improving the accuracy of health monitoring and early warning and its decision-making application value. Attached Figure Description

[0018] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a health monitoring and early warning method based on multi-source data fusion.

[0020] Figure 2 This is a schematic diagram of a health monitoring and early warning system based on multi-source data fusion.

[0021] Figure 3 A flowchart for generating candidate association data.

[0022] Figure 4 A flowchart for constructing a health status association map. Detailed Implementation

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0024] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0025] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0026] Reference Figures 1-4 This is one embodiment of the present invention, which provides a health monitoring and early warning method based on multi-source data fusion, comprising the following steps: S1. Collect multi-source monitoring health data, perform health semantic mapping and medical association labeling on the multi-source monitoring health data, and form a set of associated health data.

[0027] Collect health data from multiple sources, extract status and time information from the health data, annotate the health meaning, and output health semantic mapping data.

[0028] Specifically, wearable health monitors, smart home devices, and medical testing continuously acquire user physiological, behavioral, and environmental data. This data is then aggregated according to its source to form multi-source health monitoring data. After this data is generated, measurements such as heart rate, blood pressure, blood oxygen saturation, and body temperature are retrieved from the user's physiological data as status information. Activity frequency, sleep duration, and step count changes are also retrieved from the user's behavioral data as status information. Temperature, humidity, and air quality are retrieved from the user's environmental data as status information. The acquisition time for each status information is also retrieved as time information. After extracting the status and time information, the status information from different sources is sequentially matched according to the time information, forming a one-to-one data record. Following the data record formation, the status information in each record is matched according to preset health meaning rules, establishing a clear correspondence between status information and health meaning. This results in the output of health semantic mapping data, which includes status information, time information, and health meaning annotations.

[0029] Furthermore, the pre-defined health meaning rules are a set of mapping rules based on physiological indicator threshold ranges, behavioral pattern characteristics, and environmental influence relationships. The physiological indicator ranges are set according to medical standard indicator intervals; behavioral pattern characteristics are defined based on behavioral frequency, duration, and trends; and environmental influence relationships are established based on the interaction of environmental parameters with physiological states. By classifying and summarizing the state change characteristics of various types of data in multi-source health monitoring, a one-to-one correspondence is established between state information and corresponding health state labels, forming health meaning mapping rules to describe normal, abnormal, and trend-changing states. The rules for health semantic mapping are set based on multiple dimensions, including physiological standards, individualized baselines, behavioral monitoring patterns, and the influence of environmental factors on physiological states. Specifically, physiological indicator threshold ranges are set by referring to medical literature, health management guidelines, and individualized medical test results, with different standard ranges used for different age groups, genders, and health conditions. Behavioral pattern characteristics are based on the frequency, duration, and regularity of users' historical behavioral data. Environmental influence relationships are constructed using multi-source environmental data and their real-time impact on physiological data, and deduced based on rules from correlation analysis and a medical knowledge base.

[0030] When performing health meaning labeling on the status information in each data record, the values ​​in the status information are compared with the medical standard index range of the corresponding physiological indicator object. When the status information value is within the range of the medical standard index range, it is marked as a normal state; when the status information value exceeds the range of the medical standard index range, it is marked as an abnormal state. The status information corresponding to the behavioral status object is arranged in chronological order according to the time information. The status information of adjacent time positions is statistically analyzed to obtain the changes of the behavioral status object in the time dimension. When the change result deviates from the stable change range of the behavioral status object in the historical time series, it is marked as an abnormal state.

[0031] The status information corresponding to environmental impact objects is matched with the status information corresponding to physiological indicator objects at the same time location. When the change in the status information corresponding to environmental impact objects and the change in the status information corresponding to physiological indicator objects are consistent in time sequence, the corresponding status information is marked as an environmentally affected state. When there is status information from multiple data sources at the same time location, the status information from multiple data sources is matched according to time information. When the health meaning labeling results corresponding to the status information from multiple data sources are consistent, the consistent result is used as the health meaning labeling result, thus completing the generation of health semantic mapping data.

[0032] Based on health semantic mapping data, the correspondence between various monitoring health data in multi-source monitoring health data is identified, and associated labels are generated to produce associated labeled data.

[0033] Specifically, the correspondence of health meanings between different data items in the health semantic mapping data is analyzed. By jointly comparing the status information, time information, and health meaning annotation results, the correspondence between data from different sources under the same time dimension and the same health meaning conditions is identified. That is, the health semantic mapping data is sorted by time information, and data items with the same time information are selected to form a data set at the same time position. In the data set at the same time position, they are grouped according to the health meaning annotation results. Data items with consistent health meaning annotation results form a correspondence, and the data source information in the correspondence is recorded. Based on the identification of the correspondence, the association path is established for the data items with corresponding relationships. That is, the data items in the correspondence are arranged in the order of time information, and the correspondence with consistent health meaning annotation results in adjacent time positions is connected by a marker, so that the same health meaning forms a continuous connection sequence in the time dimension. The connection order and connection relationship are recorded to form an association path. A unified association marker is assigned to each group of data items with corresponding relationships so that the association marker can reflect the correspondence and association path between the data, forming association marker data for characterizing the association between multi-source monitoring health data.

[0034] Furthermore, in this embodiment, the associated marker data refers to the identification data used to characterize the corresponding attribution relationship and continuous association relationship between multi-source monitoring health data; the formation of the associated marker data includes time information, health meaning annotation results, and the correspondence between different data sources; specifically, when data items from different sources are at the same time position and the health meaning annotation results have consistency or continuous succession relationship, the relevant data items are identified as data items with corresponding relationships and assigned a unified associated marker; the associated marker is used to indicate that the corresponding data items belong to the same continuous health association structure, so as to perform structured organization, missing item location, and association path resolution in the future.

[0035] For example, during continuous monitoring of the same user, if the first time location collects data showing an increase in heart rate, activity level, and ambient temperature, these are respectively classified as an abnormal heart rate state, an increased activity state, and an ambient temperature rise state after health semantic mapping. If the second time location collects data showing an still elevated heart rate, continued increase in activity level, and continued rise in ambient temperature, this is also classified as a continuously connected state after health semantic mapping. Since multiple data points at the first and second time locations are continuous in time and correspond in terms of health meaning, the same association label can be assigned to each data item. The resulting association label data is not a single label, but rather a data expression used to characterize the multi-source monitoring health data belonging to the same association structure in both the continuous time dimension and the health semantic dimension.

[0036] Based on the associated labeled data, the multi-source monitoring health data is structured and organized to form an associated health data set.

[0037] Specifically, based on the association marker data, data items with the same association marker (identical association markers are determined by time information, health meaning annotation results, and data source correspondence; data items with consistent time information, consistent health meaning annotation results, and belonging to the same correspondence are judged as data items with the same association marker) in multi-source monitoring health data are grouped to form a unified data structure. On the basis of grouping, each data structure is arranged in a structured manner according to time order and health meaning annotation results, so that each data structure maintains temporal continuity and semantic consistency. At the same time, the association paths between different data structures are preserved and organized. That is, the established association paths in each data structure are arranged in the order of time information, the connection relationship between adjacent time positions is continuously recorded, the connection relationship with time interruption or inconsistent health meaning annotation results is removed, and the preserved continuous connection relationship is summarized according to the data structure correspondence. This makes the multi-source monitoring health data as a whole form a data organization form with clear association relationships and structural hierarchy, thereby outputting a set of associated health data with clear association relationships and a unified structure.

[0038] S2. Identify missing health data items in the associated health data set, extract candidate related data from the associated health data set based on the association markers of the missing health data items, perform cross-source inference to supplement the missing health data items with the candidate related data, and output the user's health data completion data.

[0039] The scan locates missing health data items within the associated health data set.

[0040] Specifically, the process involves scanning and locating missing health data items in the associated health data set. This is achieved by iterating through the time information, status information, and health meaning annotations of each data item in the associated health data set, verifying the continuity of each data item in the time series, and checking the consistency of the status information and health meaning annotations. Specifically, locations with abnormal time intervals (where the time difference between two adjacent time positions in the data sequence corresponding to the same associated marker is not continuously recorded), missing status information, or missing health meaning annotations are marked and recorded. The corresponding data item is then located in the associated health data set by combining the associated markers, thus establishing a clear correspondence between the missing location and the corresponding associated marker. Finally, the missing health data items in the associated health data set are output.

[0041] Using missing health data items as the tag index object, retrieve the associated tags corresponding to the tag index object from the missing health data items.

[0042] Specifically, using missing health data items as the tag index objects, time information, status information, and health meaning annotation results are extracted from the missing health data items. Based on the distribution of time information and health meaning annotation results in the associated health data set, indexing and positioning are performed. The associated tags in the data structure to which the missing health data items belong are searched and matched. By comparing the consistency of time information, the consistency of health meaning annotation, and the association path attribution relationship, the associated tags corresponding to the missing health data items are determined, so that a one-to-one correspondence is formed between the tag index objects and the associated tags. The associated tags corresponding to the tag index objects are obtained from the missing health data items.

[0043] By using association tags, association paths are parsed on the associated health data set, and a multi-source association path set is output.

[0044] Specifically, by using association tags, data items with the same association tags in the associated health data set are retrieved and aggregated. A temporal connection is established based on the distribution order of data items with the same association tags in terms of time information. The association between data items is confirmed based on the consistency between the status information and the health meaning annotation results. That is, data items at adjacent time positions are paired. When the health meaning annotation results of two data items are consistent and the direction of status information change is consistent, it is determined that there is an association between the two data items, and the temporal connection between the two data items is recorded as an association relationship. Furthermore, a path is constructed for the connection order between data items, linking data items with the same association tag that span different data sources and have continuous temporal relationships in chronological order to form a path structure covering multi-source monitoring health data. The path structures corresponding to all association tags are summarized and organized, and a multi-source association path set is output.

[0045] Furthermore, in the process of constructing the path for the connection order between data items, hierarchical connection constraints are applied to the data items based on the association tags. The data items are arranged according to the time information order based on the data items corresponding to the association tags. Data items at the same time position are considered to be in the same layer, and data items at adjacent time positions are considered to be in adjacent layers. Only data items in adjacent layers are allowed to establish connection relationships. At the same time, the connected data items are required to meet the requirement that the health meaning labeling results are consistent. This forms a data structure constraint that connects layer by layer according to the time hierarchy, so that data items under the same association tag can only establish connection relationships under the condition that the time is continuous and the health meaning labeling is consistent.

[0046] The path extension process is defined with termination conditions. When a data item experiences a discontinuity in the time dimension (i.e., in the data item sequence corresponding to the same associated marker, there is no continuous temporal order between adjacent time positions, or there are intermediate time positions where no corresponding data item appears, resulting in time positions not covered by data in the time series, this is considered a discontinuity in the time dimension) or when the health meaning label changes inconsistently, the current path extension is terminated. The formed connection sequence is then retained as an independent multi-source association path, thus ensuring that each multi-source association path in the multi-source association path set has clear connection boundaries and structural consistency.

[0047] In this embodiment, an association path refers to a continuous sequence of data items from multiple data sources arranged in chronological order and according to their health meaning under the same association marker constraint. The association path is used to characterize the evolution of multi-source health data during continuous monitoring. The multi-source association path set refers to a data set composed of multiple association paths formed by multiple association markers.

[0048] For example, in a data structure corresponding to the same association marker, the first time position includes the environmental temperature rise state, the activity enhancement state, and the heart rate abnormality state; the second time position includes the environmental temperature rise state, the activity enhancement state, and the heart rate abnormality state; and the third time position includes the activity enhancement state and the heart rate abnormality state but lacks environmental state records. In this case, an association path can be formed between the first time position and the second time position based on the same association marker and continuous time relationship. The association path reflects the common evolution process of environmental state, activity state, and heart rate state over continuous time. When the continuous connection sequences corresponding to all association markers are summarized, a multi-source association path set is formed. The association path reflects the continuous evolution relationship between multi-source health data, rather than the static correspondence between isolated data items.

[0049] Based on each multi-source association path in the multi-source association path set, health data is aggregated from the associated health data set to generate candidate association data.

[0050] Specifically, based on each multi-source association path in the multi-source association path set, the data items corresponding to each multi-source association path are retrieved from the associated health data set. The retrieved data items are arranged in chronological order according to the multi-source association paths, and the data items are filtered and integrated based on the consistency between the status information and the health meaning labeling results, so that the data content corresponding to the same multi-source association path forms a continuous correspondence in the time dimension and the health meaning dimension. The integrated data content is subjected to integrity and consistency checks, so that the data content corresponding to each multi-source association path forms a data set with consistent structure and continuous content. The data sets corresponding to all multi-source association paths are summarized and organized to generate candidate association data.

[0051] The candidate related data is associated with missing health data items, and the associated content data that matches the missing health data items in the candidate related data is identified.

[0052] Specifically, candidate related data is associated with missing health data items. Based on the time information, status information, and health meaning annotation results in the missing health data items, position and content matching are performed in the candidate related data. The time information of each data item in the candidate related data is aligned and compared, and the consistency comparison between the status information and the health meaning annotation results is performed. Data content that is consistent with the missing health data items in terms of time dimension, status change characteristics, and health meaning is selected. The association relationship of the selected results is verified to ensure that the positional relationship between the matched data content and the missing health data items in the association path is consistent, and the associated content data is output.

[0053] Cross-source relationship deduction and content supplementation are performed on related content data, and the supplemented content results are added to the missing items of health data to form complete health data.

[0054] Specifically, based on the distribution relationship of associated content data in the multi-source association path set, the time information, status information, and health meaning annotation results in the associated content data are correlated and deduced. That is, in the same multi-source association path, data items from different data sources are arranged in chronological order, and the existing status information at each time position is aligned to extract the status correspondence between different data sources at the same time position. At the time position corresponding to the missing health data item, the known status information of adjacent time positions in the same multi-source association path is selected, and combined with the status information correspondence of other data sources at the current time position, the status information of the missing position is matched and filled. When multiple data sources form a consistent status correspondence at the same time position, the consistent result is used as the status information of the missing health data item, and the corresponding health meaning annotation result is determined simultaneously, thereby completing the cross-source relationship deduction. The consistency verification of the status information obtained from the association deduction and the health meaning annotation result is performed, that is, the deduction result is consistent with the association relationship in the multi-source association path, and the verified content supplementation result is filled according to the time position and association mark of the missing health data item, so that the missing health data item is filled in and complete the health data is formed.

[0055] The missing health data is filled into the corresponding missing positions in the associated health data set, and the association is organized to complete the data, and the user's health data is output.

[0056] Specifically, based on the time information, status information, and health meaning annotation results in the supplemented health data, the corresponding missing positions are located in the associated health data set, and the supplemented health data is embedded into the corresponding positions according to the association tags and time order; the structural consistency verification and association path continuity verification are performed on the backfilled associated health data set, and the connection relationship between health data nodes is sorted out to ensure that each data item maintains a continuous correspondence in the time dimension and the association tag dimension, and the user's health supplemented data is output.

[0057] By collaboratively locating missing health data items and resolving related paths, the system achieves accurate identification of missing locations and traceability of related relationships in multi-source health data. This enables missing health data items to establish clear data source paths under the constraints of related markers. Through the aggregation of candidate related data driven by multi-source related path sets and cross-source inference supplementation, the system ensures that missing health data items are continuously and consistently supplemented in both the time and health meaning dimensions. This improves the completeness and consistency of user health data completion, providing a stable data foundation for subsequent health status analysis and early warning determination.

[0058] S3. Extract the correlation of user health data, output health data nodes and data correlation relationships. Based on the continuity of health data nodes in the time dimension, perform consistency constraint screening and correlation strength adjustment on the data correlation relationships, output optimized data correlation relationships, and construct a health status correlation graph based on health data nodes and optimized data correlation relationships.

[0059] The system identifies monitoring targets based on user health data and outputs a set of monitoring targets.

[0060] Specifically, by analyzing the time information, status information, and health meaning annotation results in the user health completion data item by item, the physiological indicators, behavioral states, and environmental factors corresponding to each data item are categorized. Based on the continuous distribution relationship of the same category at different time locations, the corresponding data items are aggregated and organized. Data content with the same category identifier and a continuous change relationship in the time dimension is merged, so that the same monitoring object forms a unified correspondence between different data sources and different time locations in the user health completion data, outputting a monitoring object set. The monitoring object set is a data set composed of various monitoring objects identified in the user health completion data, including physiological indicator objects, behavioral state objects, and environmental influence objects. Each monitoring object corresponds to a unified category identifier and is associated with corresponding time information, status information, and health meaning annotation results.

[0061] Based on the set of monitored objects, identify the changing relationships and influence transmission relationships in the user health data, and integrate them to form a set of related relationships.

[0062] Specifically, the time information corresponding to each monitoring object in the monitoring object set is arranged in chronological order, and the corresponding status information is arranged in the same order to form a time series. Two status information at adjacent time positions are taken in the time series, and the difference between the status information value at the later time position and the status information value at the previous time position is calculated. When the difference is greater than zero, it is marked as a state increase; when the difference is less than zero, it is marked as a state decrease; when the difference is equal to zero, it is marked as a state that remains unchanged. The absolute value of the difference is recorded as the change amplitude, forming the change relationship of the same monitoring object in the time dimension.

[0063] Alignment is achieved by selecting state information values ​​at the same time location among different monitoring objects. The changes in state information values ​​of two monitoring objects at that time location and adjacent time locations are then arranged accordingly. When two monitoring objects show the same change marker at the same time location, it is recorded as a synchronous change. When one monitoring object changes at a previous time location and another monitoring object shows the same change marker at a subsequent time location, the time locations corresponding to the previous and subsequent monitoring objects are connected and recorded. The connection order is used as a transmission direction identifier, so that state changes form an ordered connection relationship among different monitoring objects, resulting in an influence transmission relationship. After all monitoring objects have completed the connection recording, the change relationship and influence transmission relationship are summarized and organized according to the time information order and connection order, so that the connection relationship at the same time location and adjacent time locations forms a continuous expression, resulting in a set of association relationships.

[0064] The monitored objects in the set of monitored objects are taken as health data nodes, and the corresponding change relationships and impact transmission relationships in the set of relationships are taken as data relationships.

[0065] Specifically, by mapping each monitoring object in the monitoring object set item by item, each monitoring object is transformed into a corresponding health data node according to a unified identifier, and the time information, status information and health meaning annotation results of the monitoring object are retained as the attribute information of the health data node. The structure of the change relationship and the influence transmission relationship in the relationship set is analyzed, that is, the time change path corresponding to the change relationship and the transmission direction corresponding to the influence transmission relationship are associated and identified, so that the change relationship and the influence transmission relationship can establish a connection relationship between the health data nodes, and the health data nodes and their corresponding connection relationships are uniformly expressed, and the health data nodes and data association relationships are output.

[0066] Furthermore, in the change relationship, the state changes of the same monitored object at continuous time positions are arranged in chronological order to form a time change path, which consists of multiple time positions arranged in chronological order. In the influence transmission relationship, the connection order of state changes between different monitored objects is used as the transmission direction, that is, the monitored object that changes at a previous time position points to the monitored object that undergoes the same change at a later time position. When establishing the association identifier, adjacent time positions in the time change path are matched one-to-one with the connection order in the transmission direction, so that the change order in the time change path is consistent with the transmission order between different monitored objects, thereby completing the association identifier of the time change path and the transmission direction.

[0067] In this embodiment, a health data node refers to a node-based representation formed by extracting monitoring objects from user health completion data according to a unified object identifier. Each health data node corresponds to a monitoring object that can independently represent changes in health status. Data association refers to the change relationship and influence transmission relationship established between health data nodes, which is used to represent the status changes of the same monitoring object in the time dimension, as well as the status influence between different monitoring objects.

[0068] For example, heart rate monitoring objects in user health completion data can be extracted as heart rate health data nodes, activity status monitoring objects as activity health data nodes, and environmental status monitoring objects as environmental health data nodes. If the heart rate health data node shows a continuous increase at adjacent time positions, a time change relationship can be formed within the same node. If the environmental health data node first experiences a temperature increase, the activity health data node then experiences an increase in activity, and the heart rate health data node subsequently experiences an abnormal heart rate change, an influence transmission relationship can be established between the environmental health data node and the activity health data node, and an influence transmission relationship can also be established between the activity health data node and the heart rate health node. The resulting data association relationship can serve as the basis for the subsequent construction of a health status association map.

[0069] The status information of health data nodes at different time locations is expanded in chronological order, and the status information at adjacent time locations is checked for continuity, outputting status continuity data.

[0070] Specifically, during continuous health monitoring of users, each status information corresponding to a health data node contains time information. The time information in the health data node is extracted and arranged in chronological order to form a continuously increasing time sequence. Simultaneously, the status information corresponding to the time information is arranged synchronously to form a sequence set with a one-to-one correspondence between time information and status information. In the sequence set, two status information at adjacent time positions are selected sequentially, and the time difference between the corresponding time information of the two status information is compared. The time difference is also compared with the time interval relationship between adjacent time positions, and adjacent time positions with continuous time order are retained. The corresponding status information values ​​are extracted from the retained adjacent time positions, and the changes in status information values ​​between adjacent time positions are compared. The consistency between the changes in status information values ​​and the change direction of the corresponding status information at adjacent time positions is determined, and adjacent time positions with continuous change direction are retained. Adjacent time positions that simultaneously satisfy continuous time order and continuous state change direction are connected by time order markers, and the time information and status information corresponding to the connection markers are organized and combined to form a continuous state expression sequence, i.e., state continuity data.

[0071] Based on the state continuity data, the node connections in the data association are verified item by item. Connections consistent with the state continuity data are retained and inconsistent connections are eliminated, thus forming a constraint-filtered association relationship.

[0072] Specifically, after obtaining the state continuity data, the continuous state expression sequence corresponding to each healthy data node in the state continuity data is extracted, and the node connection relationship corresponding to each healthy data node is located in the data association relationship. The healthy data nodes at both ends of the node connection relationship are matched with the continuous state expression sequence in the state continuity data. For each node connection relationship, the state information of the healthy data nodes at the same time position at both ends of the node connection relationship is selected, and the continuous identifiers at the corresponding time positions in the state continuity data are compared to determine whether the healthy data nodes at both ends of the node connection relationship have continuous state expression at the same time position. After the comparison is completed for each time position, the continuous state expression of the healthy data nodes at both ends of the node connection relationship at all time positions is summarized. When the node connection relationship meets the continuous state expression condition at all corresponding time positions, the current node connection relationship is retained. When the continuous state expression is missing at any time position, the current node connection relationship is removed. After verifying all data association relationships one by one, the retained node connection relationships are organized to form a constraint-filtered association relationship.

[0073] Based on the constraint-based screening of correlation relationships, the state differences of the corresponding health data nodes at different time locations are extracted for each data correlation relationship, and the correlation strength of each data correlation relationship is determined based on the state differences.

[0074] Specifically, based on the constraint-based filtering relationships, the status information of the health data nodes at both ends of each constraint-based filtering relationship is extracted at the same time position. The status information values ​​of the health data nodes at both ends at the same time position are arranged one-to-one. After the arrangement, the difference in the status information values ​​of the health data nodes at both ends at each time position is statistically analyzed to obtain the status difference value corresponding to each time position. The status difference values ​​of each time position are then organized in chronological order to form a status difference sequence. After the status difference sequence is formed, the status differences at each time position in the status difference sequence are accumulated and averaged to make each constraint-based filtering relationship form a unique overall status difference value. The degree of association between each constraint-based filtering relationship is characterized based on the overall status difference value. The smaller the status difference value, the higher the association strength, and the larger the status difference value, the lower the association strength, thereby determining the association strength of each data relationship.

[0075] The formula for calculating the correlation strength is as follows: ; in, This represents the index identifier of a healthy data node that participates in the association strength calculation. Indicates the relationship with health data nodes The index identifier of another healthy data node used for association strength calculation. Represents health data nodes With health data nodes The strength of the data association between them. Indicates the time location index. This indicates the total number of time locations involved in the calculation. Represents health data nodes In time location The corresponding status information value, Represents health data nodes In time location The corresponding status information value.

[0076] The correlation strength of the constraint screening correlation is adjusted based on the correlation strength to form the adjusted data correlation.

[0077] Specifically, after the association strength is determined, each constraint-filtered association relationship corresponds to an association strength value. First, the node connection direction, node connection level, and health data node identifiers at both ends of the connection are extracted from the constraint-filtered association relationships. The corresponding association strength values ​​are written into the connection description of the same node connection relationship, so that the node connection relationship not only retains the connection structure between health data nodes, but also records the connection tightness. Multiple node connection relationships issued by the same health data node are checked, and the corresponding association strength values ​​are compared. Node connection relationships with different association strength values ​​are rearranged according to the connection tightness, so that node connection relationships with higher connection tightness are placed at the beginning and node connection relationships with lower connection tightness are placed at the end. After the arrangement is completed, all node connection relationships with association strength values ​​are summarized according to the connection direction and connection level between health data nodes to obtain adjusted data association relationships with clear connection structure and distinguishable connection tightness.

[0078] The data association relationships are adjusted and organized in a unified manner according to the connection direction and connection level between health data nodes to obtain optimized data association relationships.

[0079] Specifically, the process involves reading and adjusting the node connections in the data association relationships one by one, extracting the health data node identifier, connection direction, and association strength value for each node connection relationship. Based on the correspondence between the health data nodes and their adjacent health data nodes in the node connection relationships, all node connections are rearranged according to a unified connection direction, ensuring that node connections with the same connection direction form a continuous arrangement. After unifying the connection direction, multiple node connections corresponding to each health data node are categorized. Node connections with continuous differences in association strength values ​​and consistent connection paths are grouped into the same connection level, while node connections with different association strength values ​​are divided into different connection levels, creating a hierarchical structure for node connections under the same health data node. After the connection direction and connection level are both organized, all node connections are summarized uniformly, ensuring that each node connection relationship maintains a consistent correspondence between the health data node identifier, connection direction, connection level, and association strength value, thereby obtaining optimized data association relationships.

[0080] Using health data nodes as graph nodes and optimizing data relationships as graph node connections, the connection structure between health data nodes is organized and constructed to form a health status association graph.

[0081] Specifically, the process involves reading the node connections in the optimized data association, extracting the health data node identifier, connection direction, connection level, and association strength value corresponding to each node connection, mapping the health data node identifier to a graph node identifier, and retaining the time and status information corresponding to the health data node as the attribute information of the graph node. Based on the correspondence between the health data node and its adjacent health data nodes in the node connection relationship, a connection relationship is established between the corresponding graph nodes, so that each node connection relationship forms a unique corresponding connection expression between graph nodes. At the same time, the connection direction, connection level, and association strength value are attached to the corresponding connection relationship as connection attributes. After all node connection relationships are mapped, all graph nodes are arranged according to time information, and the graph nodes form an ordered distribution in the time dimension. The connection paths between graph nodes are then organized according to the connection direction and connection level to form a health status association graph.

[0082] Furthermore, in this embodiment, optimizing data association refers to the set of connection relationships obtained after the original data association relationships have undergone state continuity verification, consistency constraint screening, and association strength adjustment. Optimizing data association relationships retains node connection relationships consistent with the continuous evolution of health status, and uniformly expresses the connection direction, connection level, and connection tightness of node connection relationships. The health status association graph refers to a structured graph formed by using health data nodes as graph nodes and optimized data association relationships as the connection relationships between graph nodes, used to characterize the association organization results of user health status under multiple monitoring objects and multiple time locations.

[0083] S4. Identify the state changes of health data nodes in the health status association graph over time and the transmission paths of their association effects in the data association relationships. Implement bidirectional propagation control of the association effect transmission paths using trigger constraints and inhibition constraints, construct a state effect propagation sequence, determine the path convergence and path differentiation of the state effect propagation sequence, output the determination and identification results, extract the association relationship evolution information in the state effect propagation sequence based on the determination and identification results, classify the association pattern and extract health anomalies from the association relationship evolution information, and generate health early warning data.

[0084] Triggering constraints are used to maintain path connectivity for adjacent healthy data nodes that satisfy continuous change relationships in the positive influence propagation direction, while suppression constraints are used to block the propagation of adjacent healthy data nodes that do not satisfy continuous change relationships and remove path connectivity in the negative feedback influence direction.

[0085] The health data nodes in the health status association graph are scanned sequentially according to the time dimension to extract status change information, which is then organized sequentially and the status change results are output.

[0086] Specifically, based on the time markers corresponding to the health data nodes in the health status association graph, the health data nodes are read time-by-time, and the status records corresponding to each time position are compared item by item during the reading process to identify the status changes between adjacent time positions. Based on the identification of the status changes between adjacent time positions, the status change information is extracted and arranged continuously according to the chronological order corresponding to the time markers, so that the status change information forms an ordered connection relationship in the time dimension, and the status change results are output.

[0087] Based on the state change results, the data association relationships in the health state association map are analyzed by change mapping, the influence transmission direction and range of the state change results along the data association relationships are identified, and the association influence transmission path is integrated.

[0088] Specifically, the data association relationships between health data nodes in the health status association graph are read, and each status change information in the status change results is linked to the health data node positions at both ends of the data association relationship according to the corresponding relationship of health data nodes, so as to obtain the corresponding distribution results of status change information and data association relationship; the status change information corresponding to the health data nodes at both ends of each data association relationship is compared before and after to determine whether the status change information has a continuous extension relationship along the data association relationship, and the direction of influence transmission of status change information in the data association relationship is determined based on the continuous extension relationship. At the same time, the scope of influence transmission is determined based on the number of health data nodes covered by the state change information along the continuous connected data association relationship; the data association relationships with consistent influence transmission direction and continuous influence transmission scope are sequentially connected and organized to form the association influence transmission path.

[0089] Trigger constraint control is applied to the positive influence transmission direction in the associated influence transmission path, and suppression constraint control is applied to the negative feedback influence direction, thus forming a controlled influence transmission path.

[0090] Specifically, consistency comparison is performed on the state change information between adjacent healthy data nodes in the positive influence transmission direction. When the state change information between adjacent healthy data nodes satisfies the continuous change relationship, the path connectivity in the positive influence transmission direction is maintained. Suppression and constraint control are applied to the negative feedback influence direction. In the associated influence transmission path, conflict relationship is determined on the state change information between adjacent healthy data nodes in the negative feedback influence direction. When the state change information between adjacent healthy data nodes does not satisfy the continuous change relationship, the transmission of the corresponding negative feedback influence direction is blocked and the path connectivity is removed. The path segments that remain connected are reconnected and reorganized to form a controlled influence transmission path.

[0091] The state change results are organized sequentially based on the controlled influence propagation path to form a state influence propagation sequence.

[0092] Specifically, the state change information is matched one-to-one with the location of the healthy data node in the controlled influence transmission path. The state change information corresponding to adjacent healthy data nodes in the same controlled influence transmission path is arranged sequentially according to the transmission direction of the path. The state change information in different controlled influence transmission paths is uniformly sorted according to the connection hierarchy between the paths. This makes the state change information form a combination structure of continuous and hierarchical arrangement in the controlled influence transmission path structure, and outputs the state influence propagation sequence.

[0093] Perform path convergence determination and path differentiation identification on the state influence propagation sequence (performing path convergence determination and path differentiation identification means distinguishing and determining the convergence characteristics and diffusion characteristics of the propagation path based on the degree of node convergence and branch expansion of each propagation path in the state influence propagation sequence, and outputting the determination and identification results).

[0094] Specifically, the arrangement of healthy data nodes is read path by path along the propagation sequence of state influence. The number of times the healthy data node corresponding to the termination position and intermediate connection position of each propagation path is pointed to by multiple propagation paths is counted. The degree of node convergence is determined based on the number of times it is pointed to by multiple paths. The path structure in which multiple propagation paths converge towards the same healthy data node is identified as a convergence feature based on the degree of node convergence. The number of branches formed by the extension of a single path during the transmission process of each propagation path is counted. The degree of branch expansion is determined based on the number of branches. The path structure in which a single propagation path extends and disperses in multiple directions is identified as a diffusion feature based on the degree of branch expansion. The convergence feature and diffusion feature of each propagation path are sorted out and the judgment and identification results are output.

[0095] Furthermore, the criteria for determining path convergence and path differentiation are as follows: In each propagation path, the number of times the same healthy data node is pointed to by multiple propagation paths is counted. When the same healthy data node is pointed to by two or more propagation paths, the corresponding path structure is determined to be path convergence. The number of branches formed by a single propagation path during transmission is counted. When a single propagation path extends to form two or more branch paths, the corresponding path structure is determined to be path differentiation. The determination of path convergence and path differentiation are completed based on the number of times a node is pointed to by multiple paths and the number of path branches.

[0096] Based on the judgment and identification results, information on the evolution of relationships in the propagation sequence of state influence is extracted; Specifically, based on the judgment and identification results, the corresponding propagation path is located path by path in the state influence propagation sequence. The arrangement order of the health data nodes in each propagation path is read, and the data association relationship between adjacent health data nodes is extracted according to the connection order in the propagation path. The changes in connection direction, connection level, and association strength of each data association relationship at different time positions are recorded item by item. The changes in the data association relationship between adjacent health data nodes in the same propagation path are continuously organized according to the propagation path order, so that the change trajectory of the data association relationship in each propagation path forms an expression result that is continuously arranged according to the time dimension and path structure, and the association relationship evolution information is output.

[0097] The association evolution information is classified into association patterns and health anomalies are extracted to generate health early warning data.

[0098] Specifically, the process of classifying association patterns and extracting health anomalies in the association evolution information involves matching the convergence and diffusion features corresponding to each propagation path in the identification results with the association change trajectories in the association evolution information item by item. Association change trajectories with the same combination of convergence and diffusion features are merged and organized to form different categories of association pattern sets. The change trends and magnitude differences of the association change trajectories in each association pattern set are then compared to identify those deviating from the normal range of change. These deviating association patterns are marked as abnormal association patterns, and the corresponding association change trajectories are extracted from the abnormal association patterns as health warning data.

[0099] Furthermore, the normal range of change is the stable range of change formed by health data during continuous monitoring. It is the range determined by the continuous change pattern shown by the combined effects of physiological indicators, behavioral state, and environmental factors when there are no abnormalities in the health status. The normal range of change reflects the fluctuation boundary of health data under normal conditions and is used as a reference benchmark to determine whether the evolution information of correlation deviates from the normal state. When the evolution information of correlation exceeds the normal range of change, the corresponding change is identified as an abnormal change and used to generate health early warning data.

[0100] In this embodiment, the state influence propagation sequence refers to the sequence expression formed by organizing the state change information propagating along the connection relationship in the health state association map according to the direction of influence transmission and the time order, which is used to characterize the propagation process of health state changes among multiple health data nodes; path convergence determination refers to identifying the structural features of multiple propagation paths converging towards the same health data node; path differentiation identification refers to identifying the structural features of a single propagation path extending to multiple health data nodes; association relationship evolution information refers to the expression results of the changes in connection direction, connection level, and association strength of the data association relationship between each health data node over time in the state influence propagation sequence.

[0101] For example, in a health status correlation graph, the temperature change of the environmental health data node is first transmitted to the activity health data node, and then to the heart rate health data node, thus forming a state influence propagation sequence. If another propagation path originating from the sleep health data node also ultimately points to the same heart rate health data node, it indicates that the heart rate health data node is pointed to by multiple paths, and the corresponding path structure can be determined to have path convergence. If the activity health data node subsequently influences both the heart rate health data node and the respiratory health data node, the corresponding path structure can be determined to have path differentiation.

[0102] S5. Convert health warning data into warning prompts, match intervention strategies for user health based on the warning prompts, and output health intervention suggestions.

[0103] Health warning data is transformed into warning and alert information.

[0104] Specifically, the system reads health data nodes, data relationships, abnormal evolution characteristics, and corresponding time positions from the health early warning data. It extracts the physiological indicators, behavioral states, and environmental impacts corresponding to the health data nodes according to their corresponding identifiers. The system then maps the state change direction, state change magnitude, data relationship connection change path, and relationship strength change trend contained in the abnormal evolution characteristics to the corresponding health data nodes. During this mapping process, the abnormal evolution characteristics are sorted according to their time positions. The abnormal evolution characteristics at the same time position are integrated and expressed with their corresponding health data nodes, so that each time position forms an expression containing abnormal objects and abnormal change characteristics. The expression content from different time positions is then linked and organized in chronological order, and the abnormal objects, abnormal change characteristics, and corresponding time positions are uniformly expressed, outputting early warning information.

[0105] To match early warning information with corresponding intervention strategies, strategy matching data is generated. Based on the strategy matching data, the user's health is analyzed to generate health intervention suggestions.

[0106] Specifically, the system reads the abnormal object, abnormal type, and abnormal occurrence time from the warning message. It then matches the physiological indicator object, behavioral state object, and environmental impact object corresponding to the abnormal object with the intervention strategy information in the pre-stored medical and health knowledge information. During the matching process, the abnormal type, state change direction, and state change magnitude of the abnormal object are used as matching features. These features are compared with the applicable conditions corresponding to the intervention strategy information in the pre-stored medical and health knowledge information. The consistency of the abnormal type, state change range, and change trend included in the applicable conditions is verified, so that the abnormal characteristics of the abnormal object and the applicable conditions of the intervention strategy information are in correspondence.

[0107] From the intervention strategy information that meets the applicable conditions, extract the abnormal handling methods corresponding to the abnormal objects. Retain the intervention strategy information according to the consistency between the abnormal type and the execution object, establishing a one-to-one correspondence between the abnormal objects and the corresponding intervention strategy information. Arrange the correspondence according to the abnormal object, abnormal type, and abnormal occurrence time to generate strategy matching data. After generating the strategy matching data, extract the execution object and execution content corresponding to each intervention strategy information from the strategy matching data. Map the execution object to the physiological indicator object, behavioral state object, or environmental influence object within the abnormal object, and break down the execution content into specific operation instructions, establishing a correspondence between the execution object and execution content for each intervention strategy information. After completing the correspondence organization, combine the abnormal object, abnormal type, and abnormal occurrence time in the warning prompt information with the corresponding execution object and execution content in chronological order to form a continuous expression containing the abnormal object, abnormal type, abnormal occurrence time, and corresponding intervention content, outputting health intervention recommendations.

[0108] Furthermore, the intervention strategy information consists of pre-established processing rule data for different abnormal states, including the processing methods, execution content, and execution targets corresponding to the abnormality type. The processing methods include behavioral measures to regulate physiological indicators, intervention measures to adjust behavioral states, and adjustment measures to improve environmental factors. The execution content consists of specific operational instructions, and the execution targets are the corresponding physiological indicator objects, behavioral state objects, or environmental impact objects. The applicable conditions are the judgment rules in the intervention strategy information used to limit the scope of strategy application, including abnormality type matching conditions, state change direction matching conditions, and state change magnitude matching conditions. When the abnormality type of the abnormal object is consistent with the abnormality type in the intervention strategy information, the state change direction is consistent with the corresponding change direction in the intervention strategy information, and the state change magnitude falls within the change range corresponding to the intervention strategy information, the intervention strategy information is deemed to meet the applicable conditions.

[0109] This embodiment also provides a health monitoring and early warning system based on multi-source data fusion, including: The data acquisition module collects health data from multiple sources, performs health semantic mapping and medical association labeling on the multi-source health data, and forms a set of associated health data. The data completion module is used to identify missing health data items in the associated health data set, extract candidate related data from the associated health data set based on the association tags of the missing health data items, perform cross-source inference to supplement the missing health data items with the candidate related data, and output the user's health completion data. The graph construction module is used to extract the correlations from the user's health completion data, output the health data nodes and data correlations, and construct the health status correlation graph based on the health data nodes and data correlations. The health analysis module identifies the state changes of health data nodes in the health status correlation graph over time and the transmission paths of their correlation effects in the data relationships. It then implements bidirectional propagation control of the correlation effect transmission paths using trigger constraints and inhibition constraints, constructs a state influence propagation sequence, performs path convergence determination and path differentiation identification on the state influence propagation sequence, outputs the determination and identification results, extracts the correlation evolution information in the state influence propagation sequence based on the determination and identification results, classifies the correlation patterns and extracts health anomalies from the correlation evolution information, and generates health early warning data. The intervention suggestion module is used to convert health warning data into warning prompts, match intervention strategies for users' health based on the warning prompts, and output health intervention suggestions.

[0110] In summary, this invention achieves consistent association construction and improved data integrity of multi-source health data through the synergy of medical association labeling and cross-source inference supplementation. By performing health semantic mapping and medical association labeling on multi-source monitoring health data, a unified correspondence is established between health data from different sources, providing a standardized foundation for data processing. Based on the association labels, candidate related data are extracted for missing health data items, and cross-source inference supplementation is performed, achieving association completion of missing data positions, improving data usability and analytical support capabilities. A health status association map is constructed, and the evolution of association relationships is identified by combining time-series association analysis methods, realizing a structured expression and dynamic analysis of health status changes. This enables the generation of accurate health early warning data and supports intervention strategy matching, improving the accuracy of health monitoring and early warning and its decision-making application value.

[0111] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A health monitoring and early warning method based on multi-source data fusion, characterized in that: include, Collect health data from multiple sources, perform health semantic mapping and medical association labeling on the multi-source health data, and form a set of associated health data; Identify missing health data items in the associated health data set, extract candidate related data from the associated health data set based on the association tags of the missing health data items, perform cross-source inference to supplement the missing health data items with the candidate related data, and output the user's health data completion data; Extract the correlation of user health data, output health data nodes and data correlation relationships. Based on the state continuity of health data nodes in the time dimension, perform consistency constraint screening and correlation strength adjustment on data correlation relationships, output optimized data correlation relationships, and construct a health status correlation graph based on health data nodes and optimized data correlation relationships. This study identifies the state changes of health data nodes in a health status correlation graph over time and the transmission paths of their correlation effects in data relationships. It then implements bidirectional propagation control of these transmission paths using trigger and inhibition constraints, constructs a state influence propagation sequence, determines the path convergence and path differentiation of this sequence, outputs the determination and identification results, extracts the correlation evolution information from the state influence propagation sequence based on the determination and identification results, classifies the correlation patterns and extracts health anomalies from the correlation evolution information, and generates health early warning data. Health warning data is converted into warning information, and intervention strategies are matched to users' health based on the warning information to output health intervention suggestions.

2. The health monitoring and early warning method based on multi-source data fusion as described in claim 1, characterized in that: The specific steps for forming the associated health data set are as follows: Extract status and time information from multi-source health monitoring data, annotate the health meaning, and output health semantic mapping data; Based on health semantic mapping data, the correspondence between various monitoring health data in multi-source monitoring health data is identified, and associated labels are generated to produce associated labeled data; Based on the associated label data, the multi-source monitoring health data is structured and organized to form an associated health data set; The multi-source health monitoring data includes user physiological indicator data, user behavior data, and user environment data.

3. The health monitoring and early warning method based on multi-source data fusion as described in claim 1, characterized in that: The specific steps for extracting candidate correlation data are as follows: Scan and locate missing health data items within the associated health data set; Using missing health data items as the tag index objects, retrieve the associated tags corresponding to the tag index objects from the missing health data items; By using association tags, the association paths of the associated health data set are parsed, and a multi-source association path set is output. Based on each multi-source association path in the multi-source association path set, health data is aggregated from the associated health data set to generate candidate association data.

4. The health monitoring and early warning method based on multi-source data fusion as described in claim 1, characterized in that: The specific steps for outputting user health completion data are as follows: The candidate related data is associated with the missing health data items, and the associated content data that matches the missing health data items in the candidate related data is identified. Cross-source relationship deduction and content supplementation are performed on related content data, and the supplemented content results are added to the missing items of health data to form complete health data; The missing health data is filled into the corresponding missing positions in the associated health data set, and the association is organized to complete the data, and the user's health data is output.

5. The health monitoring and early warning method based on multi-source data fusion as described in claim 1, characterized in that: The specific steps for outputting health data nodes and data association relationships are as follows: Identify monitoring targets from user health completion data and output a set of monitoring targets; Based on the set of monitored objects, identify the changing relationships and influence transmission relationships in the user health data completion data, and integrate them to form a set of related relationships; The monitored objects in the set of monitored objects are taken as health data nodes, and the corresponding change relationships and impact transmission relationships in the set of relationships are taken as data relationships.

6. The health monitoring and early warning method based on multi-source data fusion as described in claim 1, characterized in that: The specific steps for constructing the health status association map are as follows: The status information of health data nodes at different time locations is expanded in chronological order, and the status information at adjacent time locations is checked for continuity, outputting status continuity data; Based on the state continuity data, the node connections in the data association are verified item by item. Connections consistent with the state continuity data are retained and inconsistent connections are eliminated to form a constraint-filtered association relationship. Based on the constraint-based screening of association relationships, the state differences of the corresponding health data nodes at different time locations are extracted for each data association relationship, and the association strength of each data association relationship is determined based on the state differences. Based on the strength of the association, the association relationships of the constraint screening are adjusted to form the adjusted data association relationships; The data association relationships were adjusted and uniformly organized according to the connection direction and connection level between health data nodes to obtain optimized data association relationships. Using health data nodes as graph nodes and optimizing data relationships as graph node connections, the connection structure between health data nodes is organized and constructed to form a health status association graph.

7. The health monitoring and early warning method based on multi-source data fusion as described in claim 1, characterized in that: The constructed state affects the propagation sequence, and the specific steps are as follows: The health data nodes in the health status association graph are scanned sequentially according to the time dimension to extract status change information, which is then organized sequentially and the status change results are output. Based on the state change results, the data association relationships in the health state association map are analyzed by change mapping, the influence transmission direction and influence transmission range of the state change results along the data association relationships are identified, and the association influence transmission path is integrated. Trigger constraint control is applied to the positive influence transmission direction in the associated influence transmission path, and suppression constraint control is applied to the negative feedback influence direction to form a controlled influence transmission path; The state change results are organized sequentially based on the controlled influence propagation path to form a state influence propagation sequence.

8. The health monitoring and early warning method based on multi-source data fusion as described in claim 1, characterized in that: The aforementioned path convergence determination and path differentiation identification refer to distinguishing and determining the convergence and diffusion characteristics of the propagation path based on the degree of node convergence and branch expansion of each propagation path in the state influence propagation sequence, and outputting the determination and identification results. The extraction of the correlation evolution information in the state influence propagation sequence refers to determining the corresponding propagation path in the state influence propagation sequence based on the judgment and identification results, and sequentially reading the changes in the data correlation between each health data node along the propagation path as the correlation evolution information; The aforementioned correlation pattern classification and health anomaly extraction refer to classifying the correlation evolution information according to the judgment and identification results, and identifying the correlation patterns reflecting abnormal health changes from the pattern classification as health early warning data.

9. The health monitoring and early warning method based on multi-source data fusion as described in claim 1, characterized in that: The specific steps for providing health intervention recommendations are as follows: Transform health warning data into information representation and generate warning alerts; To match early warning information with corresponding intervention strategies, strategy matching data is generated. Based on the strategy matching data, the user's health is analyzed to generate health intervention suggestions.

10. A health monitoring and early warning system based on multi-source data fusion, based on the health monitoring and early warning method based on multi-source data fusion as described in any one of claims 1 to 9, characterized in that: include, The data acquisition module collects health data from multiple sources, performs health semantic mapping and medical association labeling on the multi-source health data, and forms a set of associated health data. The data completion module is used to identify missing health data items in the associated health data set, extract candidate related data from the associated health data set based on the association tags of the missing health data items, perform cross-source inference to supplement the missing health data items with the candidate related data, and output the user's health completion data. The graph construction module is used to extract the correlations from the user's health completion data, output the health data nodes and data correlations, and construct the health status correlation graph based on the health data nodes and data correlations. The health analysis module identifies the state changes of health data nodes in the health status correlation graph over time and the transmission paths of their correlation effects in the data relationships. It then implements bidirectional propagation control of the correlation effect transmission paths using trigger constraints and inhibition constraints, constructs a state influence propagation sequence, performs path convergence determination and path differentiation identification on the state influence propagation sequence, outputs the determination and identification results, extracts the correlation evolution information in the state influence propagation sequence based on the determination and identification results, classifies the correlation patterns and extracts health anomalies from the correlation evolution information, and generates health early warning data. The intervention suggestion module is used to convert health warning data into warning prompts, match intervention strategies for users' health based on the warning prompts, and output health intervention suggestions.

Images