Personalized health management service system based on multi-source data fusion

Abstract

The invention belongs to the technical field of health information processing and management, and discloses a personalized health management service system based on multi-source data fusion, which comprises a data acquisition device access wearable device, a health application, a medical information system and a user terminal entry data, the fusion processing platform carries out missing completion, abnormity elimination, time alignment and weighted fusion on the data to form a weighted fusion sequence, the cross-domain correlation builder generates a correlation map of lifestyles and clinical indexes, and the risk modeling engine outputs individual health risks under the constraint of the map. The management scheme generator generates personalized diet, exercise, sleep and disease prevention schemes under the constraint of the rule base. The compliance evaluation system combines subjective feedback and objective monitoring to form scores, the scheme adjustment mechanism executes minimum disturbance update based on scores and atlas trends, and the auditing unit realizes whole-process data tracing and compliance guarantee.

Description

Technical Field

[0001] This invention belongs to the field of health information processing and management technology, specifically a personalized health management service system based on multi-source data fusion. Background Technology

[0002] With the widespread adoption of smart wearable devices and mobile health applications, users can easily access daily monitoring data such as heart rate, steps, and sleep. They can also record their diet and exercise within these applications. Medical institutions' information systems store physical examination reports and medical records, including clinical indicators such as blood pressure, blood sugar, and blood lipids. Users also proactively input their daily routines and dietary preferences. This multi-source data can reflect an individual's health status from different dimensions, providing a wealth of information for personalized health management.

[0003] However, existing health management systems suffer from significant technical problems when processing multi-source data. First, data from different sources differ in format, sampling frequency, and update cycle, often accompanied by missing values, outliers, and inconsistencies in time. Without a unified processing mechanism, this can easily lead to data distortion. Second, traditional methods often employ static averaging or fixed-weighted fusion, failing to consider the differences in authority and timeliness between medical data, wearable device data, and user-entered data, resulting in a lack of dynamic credibility in the fusion results. Third, existing health assessments are mostly based on single-dimensional modeling, lacking in-depth exploration of cross-domain correlations between lifestyle data and clinical indicators, making it difficult to reveal the comprehensive effects of diet, exercise, and sleep habits on chronic disease risk and physical fitness changes. Furthermore, personalized solutions often only output general dietary or exercise recommendations, failing to combine risk assessment results with cross-domain mapping for overall optimization, thus lacking specificity.

[0004] Regarding interaction and feedback, existing systems generally rely on subjective user evaluations or single device monitoring, failing to form compliance indicators that combine subjective and objective factors, resulting in a lack of scientific basis for solution adjustments. Meanwhile, with increasing compliance and regulatory requirements, relying solely on data storage encryption is no longer sufficient. Existing technologies lack log recording and traceability mechanisms for the entire process of data processing, modeling, solution generation, and adjustment, making it difficult to guarantee transparency and verifiability. Summary of the Invention

[0005] The purpose of this invention is to provide a personalized health management service system based on multi-source data fusion to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a personalized health management service system based on multi-source data fusion, the system comprising: The data acquisition device is configured to acquire heart rate, step count and sleep monitoring data from smart wearable devices, dietary records and exercise plan data from mobile health applications, physical examination reports and medical records from medical institution information systems, and daily routines and dietary preferences from user terminals. The fusion processing platform is connected to the data acquisition device, performs missing value imputation, outlier removal and timestamp alignment, and calculates the source credibility weight by source type and acquisition time, and converts various types of data into weighted fusion sequences and stores them in the database; A cross-domain correlation builder generates a correlation graph containing lifestyle indicators and clinical indicators based on the weighted fusion sequence. The edge weights of the correlation graph are jointly determined by the source credibility weight and the time decay function, and its output serves as a constraint for downstream modeling and scheme generation. The risk modeling engine takes the weighted fusion sequence and the association graph as input, uses a clustering algorithm to divide the user group, and uses a classification prediction model to generate individual health risk assessment results. The assessment results represent the risk of chronic diseases, the trend of physical decline, and the probability of sleep disorders. The management plan generator, under the constraints of the risk assessment results and the correlation map, calls a rule base containing upper limits for nutrient intake, thresholds for exercise intensity, and ranges for sleep duration, and outputs dietary recommendations, exercise plans, sleep schedules, and disease prevention measures. The compliance assessment system is used to obtain users' subjective feedback on the management plan and collect the execution deviations monitored by wearable devices. The two types of data are fused to form a compliance score, which is then fed back to the management plan generator. The scheme adjustment mechanism controls the management scheme generator to perform the scheme adjustment with the smallest disturbance under the rule base constraint when the compliance score exceeds the set threshold and is consistent with the edge weight trend of the association graph, so as to correct the personalized management scheme. The audit and compliance control unit performs data encryption and access control during the collection, transmission and storage process, and writes the source weight, graph version, modeling parameters, rule base version and scheme adjustment content into the audit log to achieve full traceability.

[0007] Preferably, the data acquisition device includes: Smart wearable device interface: Establishes communication with smart bracelets or smartwatches via Bluetooth or WiFi to obtain heart rate data (expressed in bpm), step data (expressed in steps / day), and sleep monitoring data (including total sleep duration, deep sleep duration, light sleep ratio, and number of nighttime awakenings). Mobile application interface: Periodically call diet and exercise data from health applications via API. Diet information is structured with daily energy intake and the ratio of the three macronutrients (protein g, fat g, carbohydrate g). Exercise information includes exercise type (running, swimming, etc.), exercise intensity level (light / moderate / high), and target heart rate zone. Medical information access interface: It uses HL7 or FHIR protocol to connect with the medical institution's information system and import physical examination reports (blood pressure, blood sugar, blood lipids, BMI, etc.) and medical records (diagnosis code, prescription information, medical history summary). User terminal input interface: Provides input forms to collect users' daily routines (such as average bedtime and wake-up time) and dietary preferences (such as low-salt, vegetarian, and low-sugar), and converts them into standardized fields for storage; Unlike existing technologies, it achieves full data access from wearable devices, applications, medical institutions, and user input, forming a complete health profile, unlike systems that rely on only a single source.

[0008] Preferably, the fusion processing platform performs data cleaning and standardization processing: Missing value handling: When heart rate or sleep data is missing at a certain time point, linear interpolation or the most recent time slice substitution method is used to fill in the missing data to ensure the continuity of the sequence; Outlier detection: If a certain indicator deviates from the user's historical mean by more than 3 times the standard deviation, or does not conform to the physiologically reasonable range (such as heart rate > 220 bpm), it is marked as abnormal and removed; Time alignment: Map low-frequency data such as physical examination reports to high-frequency sampled wearable data onto a unified time axis, and align them with hours as the basic unit; Weight Calculation: The credibility weight is calculated based on the source category and data collection time. ; In the formula, The source data is assigned a basic weight (0.9-1.0 for medical institution data, 0.6-0.8 for wearable devices, and 0.4-0.6 for user data). For data delay days, The attenuation coefficient is recommended to be between 0.01 and 0.1. Fusion sequence generation: The cleaned and weighted data are used to generate a weighted fusion sequence and stored in the database; Difference: Unlike the simple averaging of existing technologies, this case achieves dynamic fusion by combining source credibility with time decay, which better reflects the authenticity of the data.

[0009] Preferably, the cross-domain correlation builder generates a lifestyle-clinical indicator association map based on the fused sequences: Feature extraction: Lifestyle indicators include steps, exercise duration, dietary energy to nutrient ratio, sleep duration and sleep efficiency; Clinical indicators include BMI, blood glucose, blood pressure, and blood lipids; Graph generation: Construct a network of connections with the above indicators as nodes, where edges represent significant correlations between indicators; Edge weight update: Edge weights not only reflect statistical correlation (such as Pearson correlation and mutual information), but are also dynamically updated in combination with source credibility weights and time decay factors; Output constraints: The generated cross-domain graph is passed to the risk modeling engine and management solution generator as model input and rule constraints; Unlike existing systems that only model in a single dimension, this invention reveals the coupling relationship between exercise, diet, sleep and clinical health through cross-domain mapping, and shows how this relationship evolves dynamically over time.

[0010] Preferably, the risk modeling engine takes the fused sequence and cross-domain graph as input to generate a health risk assessment: Group clustering: Based on users' lifestyles and clinical parameters, user groups are divided into different categories, such as high-exercise, low-risk groups and low-sleep, high-BMI groups, in order to discover common characteristics.

[0011] Individual prediction: Under the constraint of group segmentation, health risk prediction is performed for individual users. The results include: probability of chronic disease occurrence (such as hypertension and diabetes), physical fitness decline trend value (reflecting the risk of declining exercise endurance), and probability of sleep disorder. Results delivery: The prediction results are delivered to the management plan generator as the basic input for generating personalized plans; Unlike existing single classification models, this invention adopts a two-layer structure of group clustering + individual prediction to avoid overfitting caused by small sample noise.

[0012] Preferably, the management solution generator generates personalized management solutions based on the risk assessment results and the constraints of the cross-domain graph: Rule invocation: Invokes the rule base. Rule entries include daily energy intake limit (e.g., ≤2500kcal), exercise intensity threshold (220 - age), and sleep duration range (7–9h). Solution generation: Combining risk assessment and graph constraints, the rule base is solved. For example, when the graph shows a significant association between sleep deprivation and the risk of hypertension, the system generates a solution that simultaneously adjusts diet (low salt) and sleep (prolongs sleep onset time). Results output: The final plan includes dietary recommendations (nutrient ratios), exercise plans (frequency and duration), sleep schedules, and disease prevention measures (such as follow-up reminders and early warnings). Differentiation: Unlike the one-dimensional recommendations of existing health applications, this invention generates a holistic solution under the dual constraints of a rule base and a cross-domain graph, ensuring personalization and scientific rigor.

[0013] Preferably, the compliance assessment system comprehensively evaluates the user's implementation of the plan: Subjective feedback: User satisfaction with the diet, exercise and sleep plans was obtained through terminal questionnaires and recorded on a 1-5 scale. Objective execution: Collect data on the user's actual execution through wearable devices, such as the number of steps taken, the difference between the exercise duration and the target value, and the difference between the actual sleep duration and the recommended plan. Overall rating: ; In the formula, For subjective feedback scores, Objective completion level (between 0 and 1), The value range is 0.4-0.6; Results feedback: Compliance scores are transmitted to the management plan generator to control plan adjustments; Difference: This case combines subjective and objective performance, which differs from traditional systems that rely on a single dimension, resulting in a more comprehensive evaluation.

[0014] Preferably, the scheme adjustment mechanism is dynamically adjusted based on compliance scores and graph trends: Conditional judgment: Adjustment is triggered when the compliance score falls below a set threshold (e.g., 70 / 100). Consistency check: Check whether the compliance trend is consistent with the direction of change of the graph edge weights to avoid unreasonable adjustments due to short-term fluctuations; Adjustment control: Under rule base constraints, allow the management scheme generator to perform updates; Disturbance optimization: Optimize and update with the principle of minimal adjustment, making minor adjustments only to some diet, exercise or sleep items, without disrupting the overall plan; Unlike traditional overlay replacement, this invention ensures the continuity of the solution and user acceptance through "minimum disturbance".

[0015] Preferably, the audit and compliance control unit ensures compliance and traceability throughout the entire process: Encryption processing: AES-256 encryption is performed on the data content during the acquisition, transmission, and storage stages; Access control: Set access permissions based on user roles (doctor, patient, system administrator); Log recording: Records source weights, graph version number, modeling parameters, rule base version, and scheme adjustment content; Traceability Output: When a compliance check or user application occurs, the system outputs a log file to verify the compliance and traceability of the entire process.

[0016] The beneficial effects of this invention are as follows: 1. This invention sets up a unified data acquisition device and fusion processing platform, integrating multidimensional health data entered from wearable devices, health applications, medical information systems, and user terminals into the same platform. During the fusion process, the system not only performs missing data completion, anomaly removal, and time alignment, but also introduces source credibility and time decay mechanisms to ensure that medical institution data and real-time monitoring data are reasonably weighted. This avoids the bias caused by the simple superposition of data from different sources in existing systems. The weighted fusion sequence formed after processing can more realistically reflect the user's dynamic health status, providing stable and reliable basic data for cross-domain modeling and risk prediction. This mechanism ensures the scientific nature and timeliness of data integration, effectively improving the accuracy of data utilization in health management.

[0017] 2. With the support of weighted fusion sequences, this invention utilizes a cross-domain correlation builder to generate a correlation map between lifestyle and clinical test indicators, establishing a systematic link between diet, exercise, sleep, and blood pressure, blood sugar, body mass index, etc. Under the constraints of this map, the risk modeling engine completes group segmentation and individual prediction, outputting results such as chronic disease risk, physical fitness decline trend, and sleep disorder probability. The management plan generator further combines the risk assessment results with the rule base content to generate personalized plans covering diet, exercise, sleep, and disease prevention. Compared with existing technologies, this system no longer relies on a single indicator or fixed template, but performs overall optimization based on cross-domain constraints, making the plans more in line with individual health characteristics and potential risks, significantly enhancing the scientific nature and pertinence of health management.

[0018] 3. This invention uses a compliance assessment system to integrate user subjective feedback and execution data monitored by wearable devices to generate a compliance score reflecting the actual execution status. The plan adjustment mechanism modifies the health management plan based on the trend relationship between this score and the correlation graph, using the principle of minimum disturbance. This ensures the continuity and user acceptability of the update process, avoiding execution obstacles caused by frequent and large-scale adjustments. At the same time, the audit and compliance control unit implements encryption protection and log recording for the entire process of data collection, fusion, modeling, generation, and adjustment, ensuring that data processing and plan changes are traceable. This design not only improves the adaptability and reliability of the system in dynamic operation, but also meets the requirements of health data security and compliance, ensuring the verifiability and long-term stability of health management services. Attached Figure Description

[0019] Figure 1 This is a flowchart of the personalized health management service system based on multi-source data fusion according to the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] like Figure 1 As shown, this embodiment of the invention provides a personalized health management service system based on multi-source data fusion, the system comprising: The data acquisition device is configured to acquire heart rate, step count and sleep monitoring data from smart wearable devices, dietary records and exercise plan data from mobile health applications, physical examination reports and medical records from medical institution information systems, and daily routines and dietary preferences from user terminals. The fusion processing platform is connected to the data acquisition device, performs missing value imputation, outlier removal and timestamp alignment, and calculates the source credibility weight by source type and acquisition time, and transforms various types of data into weighted fusion sequences and stores them in the database; The cross-domain correlation builder generates a correlation graph containing lifestyle indicators and clinical indicators based on weighted fusion sequences. The edge weights of the correlation graph are jointly determined by the source credibility weight and the time decay function, and its output serves as a constraint for downstream modeling and scheme generation. The risk modeling engine takes weighted fusion sequences and association graphs as input, uses clustering algorithms to segment user groups, and uses classification prediction models to generate individual health risk assessment results. The assessment results characterize the risk of chronic diseases, the trend of declining physical fitness, and the probability of sleep disorders. The management plan generator, under the constraints of risk assessment results and correlation maps, calls a rule base containing upper limits for nutrient intake, thresholds for exercise intensity, and ranges for sleep duration, and outputs dietary recommendations, exercise plans, sleep schedules, and disease prevention measures. The compliance assessment system is used to obtain users' subjective feedback on the management plan and collect execution deviations monitored by wearable devices. The two types of data are merged to form a compliance score, which is then fed back to the management plan generator. The scheme adjustment mechanism, when the compliance score exceeds the set threshold and is consistent with the edge weight trend of the association graph, controls the management scheme generator to perform the scheme adjustment with the smallest disturbance under the constraints of the rule base, so as to correct the personalized management scheme. The audit and compliance control unit performs data encryption and access control during the collection, transmission and storage process, and writes the source weight, graph version, modeling parameters, rule base version and scheme adjustment content into the audit log to achieve full traceability.

[0022] The data acquisition device includes: Smart wearable device interface: Establishes communication with smart bracelets or smartwatches via Bluetooth or WiFi to obtain heart rate data (expressed in bpm), step data (expressed in steps / day), and sleep monitoring data (including total sleep duration, deep sleep duration, light sleep ratio, and number of nighttime awakenings). Mobile application interface: Periodically call diet and exercise data from health applications via API. Diet information is structured with daily energy intake and the ratio of the three macronutrients (protein g, fat g, carbohydrate g). Exercise information includes exercise type (running, swimming, etc.), exercise intensity level (light / moderate / high), and target heart rate zone. Medical information access interface: It uses HL7 or FHIR protocol to connect with the medical institution's information system and import physical examination reports (blood pressure, blood sugar, blood lipids, BMI, etc.) and medical records (diagnosis code, prescription information, medical history summary). User terminal input interface: Provides input forms to collect users' daily routines (such as average bedtime and wake-up time) and dietary preferences (such as low-salt, vegetarian, and low-sugar), and converts them into standardized fields for storage; Unlike existing technologies, it achieves full data access from wearable devices, applications, medical institutions, and user input, forming a complete health profile, unlike systems that rely on only a single source.

[0023] The fusion processing platform performs data cleaning and standardization: Missing value handling: When heart rate or sleep data is missing at a certain time point, linear interpolation or the most recent time slice substitution method is used to fill in the missing data to ensure the continuity of the sequence; Outlier detection: If a certain indicator deviates from the user's historical mean by more than 3 times the standard deviation, or does not conform to the physiologically reasonable range (such as heart rate > 220 bpm), it is marked as abnormal and removed; Time alignment: Map low-frequency data such as physical examination reports to high-frequency sampled wearable data onto a unified time axis, and align them with hours as the basic unit; Weight Calculation: The credibility weight is calculated based on the source category and data collection time. ; In the formula, The source data is assigned a basic weight (0.9-1.0 for medical institution data, 0.6-0.8 for wearable devices, and 0.4-0.6 for user data). For data delay days, The attenuation coefficient is recommended to be between 0.01 and 0.1. Fusion sequence generation: The cleaned and weighted data are used to generate a weighted fusion sequence and stored in the database; Difference: Unlike the simple averaging of existing technologies, this case achieves dynamic fusion by combining source credibility with time decay, which better reflects the authenticity of the data.

[0024] Among them, the cross-domain correlation builder generates a lifestyle-clinical indicator association map based on the fusion sequence: Feature extraction: Lifestyle indicators include steps, exercise duration, dietary energy to nutrient ratio, sleep duration and sleep efficiency; Clinical indicators include BMI, blood glucose, blood pressure, and blood lipids; Graph generation: Construct a network of connections with the above indicators as nodes, where edges represent significant correlations between indicators; Edge weight update: Edge weights not only reflect statistical correlation (such as Pearson correlation and mutual information), but are also dynamically updated in combination with source credibility weights and time decay factors; Output constraints: The generated cross-domain graph is passed to the risk modeling engine and management solution generator as model input and rule constraints; Unlike existing systems that only model in a single dimension, this invention reveals the coupling relationship between exercise, diet, sleep and clinical health through cross-domain mapping, and shows how this relationship evolves dynamically over time.

[0025] The risk modeling engine will take both the fusion sequence and cross-domain graph as input to generate a health risk assessment. Group clustering: Based on users' lifestyles and clinical parameters, user groups are divided into different categories, such as high-exercise, low-risk groups and low-sleep, high-BMI groups, in order to discover common characteristics.

[0026] Individual prediction: Under the constraint of group segmentation, health risk prediction is performed for individual users. The results include: probability of chronic disease occurrence (such as hypertension and diabetes), physical fitness decline trend value (reflecting the risk of declining exercise endurance), and probability of sleep disorder. Results delivery: The prediction results are delivered to the management plan generator as the basic input for generating personalized plans; Unlike existing single classification models, this invention adopts a two-layer structure of group clustering + individual prediction to avoid overfitting caused by small sample noise.

[0027] The management solution generator generates personalized management solutions based on the risk assessment results and cross-domain graph constraints. Rule invocation: Invokes the rule base. Rule entries include daily energy intake limit (e.g., ≤2500kcal), exercise intensity threshold (220 - age), and sleep duration range (7–9h). Solution generation: Combining risk assessment and graph constraints, the rule base is solved. For example, when the graph shows a significant association between sleep deprivation and the risk of hypertension, the system generates a solution that simultaneously adjusts diet (low salt) and sleep (prolongs sleep onset time). Results output: The final plan includes dietary recommendations (nutrient ratios), exercise plans (frequency and duration), sleep schedules, and disease prevention measures (such as follow-up reminders and early warnings). Differentiation: Unlike the one-dimensional recommendations of existing health applications, this invention generates a holistic solution under the dual constraints of a rule base and a cross-domain graph, ensuring personalization and scientific rigor.

[0028] The compliance assessment system comprehensively evaluates users' implementation of the plan: Subjective feedback: User satisfaction with the diet, exercise and sleep plans was obtained through terminal questionnaires and recorded on a 1-5 scale. Objective execution: Collect data on the user's actual execution through wearable devices, such as the number of steps taken, the difference between the exercise duration and the target value, and the difference between the actual sleep duration and the recommended plan. Overall rating: ; In the formula, For subjective feedback scores, Objective completion level (between 0 and 1), The value range is 0.4-0.6; Results feedback: Compliance scores are transmitted to the management plan generator to control plan adjustments; Difference: This case combines subjective and objective performance, which differs from traditional systems that rely on a single dimension, resulting in a more comprehensive evaluation.

[0029] The program adjustment mechanism dynamically adjusts the program based on compliance scores and graph trends. Conditional judgment: Adjustment is triggered when the compliance score falls below a set threshold (e.g., 70 / 100). Consistency check: Check whether the compliance trend is consistent with the direction of change of the graph edge weights to avoid unreasonable adjustments due to short-term fluctuations; Adjustment control: Under rule base constraints, allow the management scheme generator to perform updates; Disturbance optimization: Optimize and update with the principle of minimal adjustment, making minor adjustments only to some diet, exercise or sleep items, without disrupting the overall plan; Unlike traditional overlay replacement, this invention ensures the continuity of the solution and user acceptance by minimizing disturbances.

[0030] Among them, the audit and compliance control unit ensures compliance and traceability throughout the entire process: Encryption processing: AES-256 encryption is performed on the data content during the acquisition, transmission, and storage stages; Access control: Set access permissions based on user roles (doctor, patient, system administrator); Log recording: Records source weights, graph version number, modeling parameters, rule base version, and scheme adjustment content; Traceability Output: When a compliance check or user application occurs, the system outputs a log file to verify the compliance and traceability of the entire process.

[0031] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0032] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A personalized health management service system based on multi-source data fusion, characterized in that: The system includes: The data acquisition device is configured to acquire heart rate, step count and sleep monitoring data from smart wearable devices, dietary records and exercise plan data from mobile health applications, physical examination reports and medical records from medical institution information systems, and daily routines and dietary preferences from user terminals. The fusion processing platform is connected to the data acquisition device, performs missing value imputation, outlier removal and timestamp alignment, and calculates the source credibility weight by source type and acquisition time, and converts various types of data into weighted fusion sequences and stores them in the database; A cross-domain correlation builder generates a correlation graph containing lifestyle indicators and clinical indicators based on the weighted fusion sequence. The edge weights of the correlation graph are jointly determined by the source credibility weight and the time decay function, and its output serves as a constraint for downstream modeling and scheme generation. The risk modeling engine takes the weighted fusion sequence and the association graph as input, uses a clustering algorithm to divide the user group, and uses a classification prediction model to generate individual health risk assessment results. The assessment results represent the risk of chronic diseases, the trend of physical decline, and the probability of sleep disorders. The management plan generator, under the constraints of the risk assessment results and the correlation map, calls a rule base containing upper limits for nutrient intake, thresholds for exercise intensity, and ranges for sleep duration, and outputs dietary recommendations, exercise plans, sleep schedules, and disease prevention measures. The compliance assessment system is used to obtain users' subjective feedback on the management plan and collect the execution deviations monitored by wearable devices. The two types of data are fused to form a compliance score, which is then fed back to the management plan generator. The scheme adjustment mechanism controls the management scheme generator to perform the scheme adjustment with the smallest disturbance under the rule base constraint when the compliance score exceeds the set threshold and is consistent with the edge weight trend of the association graph, so as to correct the personalized management scheme. The audit and compliance control unit performs data encryption and access control during the collection, transmission and storage process, and writes the source weight, graph version, modeling parameters, rule base version and scheme adjustment content into the audit log to achieve full traceability.

2. The personalized health management service system based on multi-source data fusion according to claim 1, characterized in that: The data acquisition device includes: The smart wearable device interface is configured to receive heart rate data, step count data, and sleep monitoring data uploaded by smart bracelets or smartwatches. The mobile application interface is configured to receive dietary information and exercise plans recorded by health management applications and synchronize them periodically via interface calls. The medical information access interface is configured to establish a communication connection with the medical institution's information system, obtain the user's physical examination report and medical record, and import them into the system in the form of electronic medical records or examination report files. The user terminal input interface is configured to receive the user's daily routine and dietary preferences entered on the terminal interface and convert them into structured data storage.

3. The personalized health management service system based on multi-source data fusion according to claim 2, characterized in that: The fusion processing platform includes: The missing value processing unit is configured to supplement missing items in the collected data by interpolation or replacement with adjacent time slices. An outlier detector is configured to identify data that exceeds a set threshold or deviates from historical trends, and remove it after confirming an anomaly. The time alignment component is configured to align data from different sources according to a unified time base to ensure sequence consistency. The weight calculator can be configured to calculate credibility weights based on the data source category and collection time, and assign corresponding weights to data from different sources. The fusion sequence generator is configured to transform data that has been completed, removed, aligned, and weighted into a weighted fusion sequence and write it to the database.

4. The personalized health management service system based on multi-source data fusion according to claim 3, characterized in that: The cross-domain relevance builder includes: The feature extraction unit is configured to extract lifestyle indicators related to exercise, diet and sleep, as well as clinical parameters from physical examination indicators and medical records from the weighted fusion sequence. The atlas generator is configured to build a cross-domain association atlas based on the statistical correlation between the lifestyle indicators and clinical parameters; The edge weight calculator is configured to calculate and update the weights of each associated edge by combining the source credibility weight and the time decay function when generating the graph. The constraint output component is configured to use the updated correlation graph as input conditions for the downstream risk modeling engine and management solution generator to limit the modeling and generation process.

5. The personalized health management service system based on multi-source data fusion according to claim 4, characterized in that: The risk modeling engine includes: The input receiving unit is configured to use the weighted fusion sequence and the cross-domain association graph as modeling input, wherein the weighted fusion sequence includes data such as the user's heart rate, steps, sleep quality, diet records, exercise plans, physical examination reports and medical records; A population clustering processor is configured to perform clustering operations based on the input data, dividing users into several healthy groups according to a combination of lifestyle characteristics and clinical indicators. An individual classification predictor is configured to predict the risk of chronic diseases, the trend of declining physical fitness, and the probability of sleep disorders of individual users using a classification model under the constraints of the group segmentation results. The result delivery component is configured to use the prediction results as input conditions for generating personalized health management plans.

6. The personalized health management service system based on multi-source data fusion according to claim 5, characterized in that: The management scheme generator includes: The rule invocation unit is configured to read rule base entries containing daily nutrient intake limits, exercise intensity thresholds, and sleep duration ranges; The solution generation processor is configured to perform constraint solving on the rule base entries under the constraints of the risk assessment results and the cross-domain correlation graph, so as to obtain an individualized health management solution; The output component is configured to decompose the health management plan into dietary recommendations, exercise plans, sleep schedules, and disease prevention measures, and provide them to downstream interactive systems.

7. The personalized health management service system based on multi-source data fusion according to claim 6, characterized in that: The compliance assessment system includes: The feedback collection unit acquires users' subjective evaluation data of the health management plan through the user terminal interface; The monitoring unit executes the data interface of the wearable device to collect the actual deviation values ​​of the user during the execution of the diet, exercise and sleep plans; The rating calculator integrates the subjective evaluation data with the actual deviation value according to a preset weight to generate a compliance rating. The result feedback component transmits the compliance score to the management scheme generator to control subsequent scheme updates.

8. The personalized health management service system based on multi-source data fusion according to claim 7, characterized in that: The adjustment mechanism of the scheme includes: The condition determination unit is configured to detect the compliance score generated by the compliance assessment system and determine whether the score exceeds a set threshold. A consistency checker is configured to compare the execution trend corresponding to the compliance score with the edge weight change direction of the cross-domain association graph, and confirm whether the two are consistent. Adjust the controller to configure it such that when both the condition determination unit and the consistency checker meet the triggering conditions, the control management scheme generator performs a scheme update under the constraints of the rule base. The perturbation optimizer is configured to fine-tune some recommendations in diet, exercise, sleep, or disease prevention when performing a plan update, with the optimization goal of reducing the magnitude of plan adjustments, and output a revised personalized management plan.

9. The personalized health management service system based on multi-source data fusion according to claim 8, characterized in that: The audit and compliance control unit includes: An encryption processor is configured to perform encryption operations on data content during health data collection, transmission, and storage to prevent unauthorized access; The permission manager is configured to set access permissions based on user roles, controlling the reading and modification of health data. The logger is configured to write source credibility weight, cross-domain association graph version number, risk modeling parameters, rule base version number, and scheme adjustment content to the audit log. The traceability output component is configured to provide the audit logs during compliance checks or user authorization verification to achieve traceability and compliance verification throughout the data processing process.

Images