Health risk early warning method and system based on body fat scale and sports bracelet data

Abstract

The application relates to a health risk early warning method and system based on body fat scale and sports bracelet data. The method comprises the following steps: synchronously collecting body composition data of the body fat scale and physiological activity data of the sports bracelet, performing timestamp alignment and abnormal value elimination on the data to generate a multi-source fusion data set, constructing an individualized health state trajectory according to the data set, identifying a coordinated change mode of body composition fluctuation and physiological activity deviation, matching and judging the coordinated change mode with preset health risk triggering rules, and generating and outputting a graded early warning signal when a dynamic evolution path meeting the risk condition exists, so that more accurate and timely health risk early warning is realized.

Description

Technical Field

[0001] This application relates to the fields of smart health and information technology, and in particular to a health risk early warning method and system based on body fat scale and fitness tracker data. Background Technology

[0002] In the field of health management and monitoring, multi-source data fusion technology is gradually becoming a research hotspot. With the widespread adoption of wearable devices such as body fat scales and fitness trackers, utilizing user data collected by these devices for health analysis is of great significance. Body fat scales can measure body composition data, such as weight and body fat percentage, reflecting the body's basic physiological condition; fitness trackers can record users' physiological activity data, such as heart rate and steps, reflecting the user's daily exercise and activity levels. Through comprehensive analysis of this multi-source data, it is hoped that more accurate health monitoring and risk warning can be achieved.

[0003] To effectively utilize multi-source data, existing technologies typically employ data fusion methods, integrating data from body fat scales and fitness trackers. A common solution involves simply stitching and summarizing data collected from different devices, then building a health assessment model based on this aggregated data. While this method can assess a user's health status to some extent, it has certain limitations.

[0004] The main problem with existing solutions is the lack of in-depth analysis of the inherent correlations and synergistic change patterns among multi-source data. Simple data splicing and summarization cannot accurately capture the complex relationship between body composition fluctuations and deviations in physiological activity, resulting in insufficient accuracy and timeliness in health risk warnings. For example, in reality, an increase in body fat percentage may be synergistically related to a sustained increase in resting heart rate or a significant reduction in sleep duration, but existing methods struggle to identify such synergistic change patterns, thus failing to detect potential health risks in a timely manner. Summary of the Invention

[0005] The main purpose of this application is to provide a health risk early warning method and system based on body fat scale and fitness tracker data, which can deeply explore the intrinsic correlation and synergistic change patterns between multi-source data to achieve more accurate and timely health risk early warning.

[0006] To achieve the above objectives, embodiments of the present invention provide a health risk warning method based on data from a body fat scale and a fitness tracker, the method comprising the following steps: The system simultaneously collects body composition data measured by the user on the body fat scale and physiological activity data recorded by the fitness tracker. The body composition data includes weight, body fat percentage, skeletal muscle mass, and basal metabolic rate, while the physiological activity data includes heart rate sequence, cumulative step count, and sleep duration. The body composition data and physiological activity data are time-stamp aligned and outlier removed to generate a multi-source fusion dataset, wherein the multi-source fusion dataset contains valid data records aligned to daily granularity and verified for consistency. Based on the multi-source fusion dataset, an individualized health status trajectory is constructed to identify the synergistic change pattern of body composition fluctuations and physiological activity deviations over multiple consecutive days. The synergistic change pattern includes a combination of features such as increased body fat percentage accompanied by a sustained increase in resting heart rate or a significant reduction in sleep duration. The collaborative change pattern is matched with the preset health risk triggering rules to determine whether there is a dynamic evolution path that meets the risk conditions. The health risk triggering rules define the joint deviation threshold relationship between body composition and physiological activity indicators in the time dimension. When a dynamic evolution path that meets the risk conditions exists, a health risk warning sign is generated and output to the user terminal, wherein the health risk warning sign is a graded warning signal corresponding to the dynamic evolution path.

[0007] In summary, the technical solution of this application provides a rich data source for subsequent health analysis by simultaneously collecting body composition data from a body fat scale and physiological activity data from a fitness tracker. Timestamp alignment and outlier removal of these data ensure the accuracy and effectiveness of the multi-source fusion dataset. Based on this dataset, a personalized health status trajectory can be constructed, identifying coordinated change patterns between body composition fluctuations and physiological activity deviations over multiple days, such as an increase in body fat percentage accompanied by a sustained increase in resting heart rate or a significant decrease in sleep duration. Matching these coordinated change patterns with preset health risk trigger rules accurately determines whether a dynamic evolution path meeting risk conditions exists. When a valid risk path exists, a graded early warning signal is generated and output, enabling users to understand their health risk status in a timely manner and take corresponding intervention and adjustment measures, thereby improving the effectiveness and efficiency of health management. Attached Figure Description

[0008] Figure 1a This is a schematic diagram of a health risk warning method based on body fat scale and fitness tracker data in an embodiment of this application. Figure 1b This is a visual data display of a health management application in the embodiments of this application; Figure 2 A flowchart of a health risk warning method based on body fat scale and fitness tracker data is provided for embodiments of this application; Figure 3 A schematic diagram of the data preprocessing and verification process provided for embodiments of this application; Figure 4 This is a schematic diagram of the constant value determination and removal process provided in an embodiment of this application; Figure 5 A schematic diagram illustrating the process of forming a coordinated change mode provided in the embodiments of this application; Figure 6 This is a schematic diagram of the identification and marking process provided in the embodiments of this application; Figure 7 A flowchart illustrating the dynamic evolution path determination provided in this application embodiment; Figure 8 This is a schematic diagram illustrating the process of generating the activation rule list provided in the embodiments of this application; Figure 9 A schematic diagram of the structure of a dynamic health risk early warning system based on multi-source data fusion of body fat scale and fitness tracker provided in an embodiment of this application; Figure 10 A schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0009] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0010] This application provides a health risk warning method and system based on body fat scale and fitness tracker data, which will be described in detail below.

[0011] like Figure 1a As shown, a health risk warning method scenario based on body fat scale and fitness tracker data is provided. In this scenario, a body fat scale, a fitness tracker, and a user terminal are included; wherein the body fat scale, fitness tracker, and user terminal are connected through a wireless network.

[0012] Taking a home health management scenario as an example, in this scenario, users can use a body fat scale to measure their body composition data at home, while simultaneously wearing a fitness tracker to record daily physiological activity data. The body fat scale is typically placed in a convenient location such as the living room or bedroom. Users stand on the scale, and body composition indicators such as weight, body fat percentage, skeletal muscle mass, and basal metabolic rate are measured using technologies such as bioelectrical impedance analysis. The fitness tracker is worn on the wrist to monitor the user's heart rate, steps, sleep, and other physiological activity data in real time.

[0013] Data collected by the body fat scale and fitness tracker is transmitted wirelessly to the user's terminal, such as a mobile phone or tablet. A dedicated health management application is installed on the user's terminal, which receives, stores, and processes the data from the body fat scale and fitness tracker. The application first preprocesses the collected multi-source data. For example, it establishes an association index between the fitness tracker data and body composition data within a set time window before and after the body fat scale measurement, ensuring the temporal correspondence of the data. Then, it standardizes the units and format of the body composition data, converting the raw measurements into structured fields for easier subsequent analysis. Physiological activity data is segmented and aggregated according to sampling frequency to generate daily statistical summaries, such as average daily heart rate, total steps, and effective sleep duration. Figure 1b As shown, the health management application can also visualize the data collected by the body fat scale, such as temporary weight, body fat percentage, skeletal muscle and other data.

[0014] Next, the application constructs an individualized health status trajectory based on the preprocessed multi-source fusion dataset. For example, it extracts a continuous N-day data sequence from the dataset, constructs an individual observation window, and calculates the trends and deviations of body composition and physiological activity indicators within the observation window. By cross-referencing the trends of body composition indicators with the directions of deviation of physiological activity indicators, windows that simultaneously meet specific combinations of conditions are selected, such as an increase in body fat percentage and a continuous increase in resting heart rate. These selected windows are marked as potential risk evolution segments, and their corresponding co-change patterns are extracted.

[0015] The application then matches the co-change pattern with preset health risk triggering rules. The rule base defines the joint deviation threshold relationship between body composition and physiological activity indicators over time, such as an increase in body fat percentage exceeding a certain proportion and a continuous increase in resting heart rate exceeding a preset number of days. By comparing the indicator combinations in the co-change pattern with the rule entries one by one, the application finds matching rule identifiers and verifies whether their corresponding time-series continuity conditions are met. If met, the risk assessment flag of that rule entry is activated, generating a risk path set.

[0016] When the risk path set is not empty, the application generates a health risk warning indicator and outputs it to the user's terminal. The warning indicator is a graded warning signal corresponding to the dynamically evolving path, such as low risk, medium risk, and high risk. Users can view these warning messages in the application and take corresponding measures based on the warning level. For example, at low risk, users can adjust their diet and exercise habits appropriately; at medium risk, they may need to increase their exercise or consult a professional doctor; and at high risk, they should seek medical attention promptly for further examination and diagnosis.

[0017] Throughout the health management process, users can also view their historical health data and trends in health status through the application to understand their own health condition. Simultaneously, the application can provide personalized health advice and guidance to help users improve their lifestyle and reduce health risks. For example, refer to... Figure 1b The application can provide health analysis results and suggestions. In this way, the health risk warning method based on body fat scale and fitness tracker data can provide users with comprehensive, accurate, and timely health management services, improving users' health levels.

[0018] refer to Figure 2 , Figure 2 This is a flowchart illustrating a health risk warning method based on body fat scale and fitness tracker data provided in this application embodiment. The execution subject of this method can be a computer device (such as a mobile phone), etc. The health risk warning method based on body fat scale and fitness tracker data provided in this application embodiment specifically includes: S10: Synchronously collect body composition data measured by the user on the body fat scale and physiological activity data recorded by the fitness tracker. The body composition data includes weight, body fat percentage, skeletal muscle mass and basal metabolic rate, and the physiological activity data includes heart rate sequence, cumulative step count and sleep duration.

[0019] In this embodiment, body composition data are indicators reflecting the basic physiological condition of the human body. Body weight is the total weight of the human body and is one of the basic indicators for measuring human health and nutritional status; body fat percentage refers to the proportion of fat weight in the total body weight, reflecting the body's fat content and degree of obesity; skeletal muscle mass refers to the weight of skeletal muscle, which plays an important role in maintaining the body's athletic ability and metabolic level; basal metabolic rate refers to the energy metabolism rate of the human body in a state of wakefulness and extreme rest, unaffected by muscle activity, ambient temperature, food, or mental stress, reflecting the energy required to maintain basic life activities.

[0020] Among them, physiological activity data reflects the user's daily exercise and activity level. Heart rate sequence records the user's heart rate changes at different time periods. Heart rate is an important indicator reflecting cardiac function and physical condition; cumulative steps is the total number of steps the user takes in a day, reflecting the user's exercise volume; sleep duration is the user's sleep time at night.

[0021] In this embodiment, the simultaneous collection of body composition data and physiological activity data is to ensure the temporal correspondence of the data, enabling accurate subsequent analysis and fusion. In practice, a synchronous triggering mechanism can be set up so that when the user completes the measurement operation on the body fat scale, the fitness tracker simultaneously uploads the data for the corresponding time period to the user terminal, allowing the user terminal to correlate the body composition data and physiological activity data.

[0022] In one embodiment, Bluetooth communication technology can be used to connect the body fat scale and fitness tracker to the user terminal. When the user stands on the body fat scale for measurement, the scale automatically collects body composition data and transmits the data to the user terminal via Bluetooth. Simultaneously, the fitness tracker records the user's physiological activity data in real time and transmits the physiological activity data within a set time window before and after the body fat scale measurement to the user terminal via Bluetooth. The user terminal can perform preliminary processing and storage of this data to prepare for subsequent analysis. This synchronous data collection method ensures the temporal consistency of body composition data and physiological activity data, improving data accuracy and usability.

[0023] In one embodiment, step S10 can be implemented as follows: W1: After the user completes the body fat scale measurement, extract the fitness tracker data within a set time window before and after the measurement time, and establish a correlation index between the body composition data and the physiological activity data of the corresponding time window, wherein the correlation index uses the measurement time as the reference anchor point.

[0024] In this embodiment, the association index is a mechanism for establishing a correspondence between body composition data and physiological activity data. Using the measurement time as a baseline anchor point ensures accurate temporal correspondence between body composition data and physiological activity data. Setting a time window allows for the acquisition of data from the fitness tracker within a certain range before and after the body fat scale measurement time, ensuring data relevance and completeness.

[0025] In one embodiment, after a user completes the measurement on the body fat scale, the scale records the measurement time and sends this information to the user's terminal. The user terminal then searches for physiological activity data before and after the measurement time in the data recorded by the fitness tracker, based on a preset time window. Then, using the measurement time as a reference, the body composition data and the corresponding physiological activity data within the time window are correlated to establish a correlation index. For example, a database approach can be used, storing body composition data and physiological activity data in different tables, with the measurement time used as the correlation field to establish an index relationship between the two tables. This allows for quick location and retrieval of relevant data during subsequent data queries and analysis. For instance, assuming the user completes the body fat scale measurement at 9:00 AM, with a set time window of 30 minutes before and after, the physiological activity data recorded by the fitness tracker between 8:30 AM and 9:30 AM will be extracted and correlated with the body composition data obtained from the body fat scale measurement. Establishing a correlation index improves data organization and management efficiency, facilitates subsequent data analysis, and more accurately uncovers correlations between multi-source data, thereby enhancing the accuracy of health risk assessment.

[0026] W2: Unit unification and format standardization are performed on each indicator in the body composition data, and the original measurement values ​​are converted into structured fields, wherein the structured fields are stored in data containers according to the indicator type.

[0027] In this embodiment, unit unification refers to converting the units of different indicators in body composition data into a unified standard unit, such as unifying the unit of weight to kilograms and the unit of body fat percentage to percentage. Format standardization refers to standardizing the format of the original measurement values ​​to meet specific format requirements, such as retaining a certain number of decimal places. Structured fields refer to organizing body composition data according to a certain structure, with each indicator corresponding to a specific field. Data containers are the carriers used to store structured fields, such as database tables and arrays.

[0028] In one embodiment, for the body composition data measured by a body fat scale, first check the units and formats of each indicator. If it is found that the unit of weight is jin, convert it to kilograms; if the format of the body fat percentage is in decimal form, convert it to percentage form. Then, convert the original measurement values into structured fields according to the indicator types. For example, store the weight, body fat percentage, skeletal muscle mass, and basal metabolic rate in different fields respectively. Finally, store these structured fields in a database table, where each record corresponds to one measurement data. For example, in the database table, there are fields such as "weight (kg)", "body fat percentage (%)", "skeletal muscle mass (kg)", "basal metabolic rate (kcal / day)", etc., and store the body composition data of each measurement according to the corresponding fields. By unifying the units and standardizing the formats, converting the body composition data into structured fields and storing them in a data container can improve the data quality and management efficiency, providing a more accurate and standardized data basis for subsequent health data analysis.

[0029] W3: Segment and aggregate the physiological activity data according to the sampling frequency to generate a daily statistical summary, where the statistical summary includes the average daily heart rate, total number of steps, and effective sleep duration.

[0030] In the embodiments of the present application, the sampling frequency refers to the time interval at which the sports bracelet collects physiological activity data. Segment aggregation means dividing the physiological activity data according to a certain time period and performing aggregation processing on the data within each time period. The daily statistical summary means summarizing and statistically analyzing the physiological activity data within a day to obtain statistical information such as the average daily heart rate, total number of steps, and effective sleep duration. The average daily heart rate refers to the average value of the heart rate within a day, the total number of steps refers to the total number of steps walked within a day, and the effective sleep duration refers to the time actually in the effective sleep state within a day.

[0031] In one embodiment, the fitness tracker records heart rate data at a sampling frequency of one minute, step count data at a sampling frequency of one step, and sleep status data at a sampling frequency of every 15 minutes. For heart rate data, all heart rate values ​​within a day are summed and then divided by the total number of samples to obtain the daily average heart rate. For step count data, all steps recorded within a day are summed to obtain the total number of steps. For sleep status data, effective sleep time periods are identified according to preset sleep status judgment criteria, and the durations of these time periods are summed to obtain the effective sleep duration. For example, assuming the fitness tracker records 1440 heart rate values ​​in a day, these values ​​are summed and divided by 1440 to obtain the daily average heart rate; the recorded step count is 8000 steps, then the total step count is 8000; based on sleep status judgment, the effective sleep time period is identified as 7 hours and 30 minutes, then the effective sleep duration is 7.5 hours. This statistical information is stored and used as a daily statistical summary. By segmenting and aggregating physiological activity data to generate statistical summaries, the physiological activity data can be made more concise and valuable, facilitating subsequent health risk analysis and comparison.

[0032] W4: The structured volume component data is bound to the daily statistical summary through user identity to form an initial multi-source data pair, wherein the initial multi-source data pair retains the original timestamp information.

[0033] In this embodiment, the user identity identifier is information used to uniquely identify each user, such as a user ID. Binding the structured body composition data to the daily statistical summary using the user identity identifier ensures that each user's data is independent and corresponding. The initial multi-source data pair refers to a data pair that combines the statistical summaries of body composition data and physiological activity data; retaining the original timestamp information facilitates subsequent time-dimensional analysis of the data.

[0034] In one embodiment, assuming the user's identity is "User001", the user's structured body composition data (e.g., weight 65 kg, body fat percentage 20%) is bound to a daily statistical summary (e.g., average daily heart rate 75 bpm, total steps 8000, effective sleep duration 7.5 hours). By adding a user identity field and a timestamp field to the data, these data are combined into an initial multi-source data pair. For example, a record in the database table can be represented as (User001, 2024-10-10, 65 kg, 20%, 75 bpm, 8000 steps, 7.5 hours). In this way, the multi-source data of each user at different times can be easily queried and analyzed using the user identity and timestamp. By binding the structured body composition data with the statistical summary to form an initial multi-source data pair and retaining the timestamp information, effective integration and management of multi-source data can be achieved, providing a more comprehensive and accurate data foundation for subsequent health risk analysis.

[0035] W5: Perform integrity checks on the initial multi-source data pairs, remove records missing from either data source, and generate a data set to be processed, wherein the data set to be processed contains only entries where both source data are complete.

[0036] In this embodiment, integrity verification refers to checking whether statistical summaries of body composition data and physiological activity data exist in the initial multi-source data pair, i.e., whether both body composition data and daily statistical summaries are included. If a record is missing either data source, it indicates that the data in that record is incomplete and may affect subsequent analysis results; therefore, it needs to be removed. The dataset to be processed refers to the set of entries that remain after integrity verification, where both source data are complete.

[0037] In one embodiment, the initial multi-source data pairs are iterated and checked. For each record, it is checked whether it contains body composition data and a daily statistical summary. If a record is found to be missing body composition data (e.g., fields such as weight and body fat percentage are empty) or missing a statistical summary of physiological activity data (e.g., fields such as average daily heart rate, total steps, and effective sleep duration are empty), the record is marked as invalid and removed. For example, in a set containing 100 initial multi-source data pairs, after integrity checks, 10 records are found to be missing body composition data, and 5 records are found to be missing statistical summaries of physiological activity data. After removing these 15 records, a dataset of 85 records to be processed is generated. By performing integrity checks and removing incomplete records to generate a dataset to be processed, the quality and usability of the data can be improved.

[0038] S20: The body composition data and physiological activity data are time-stamp aligned and outlier removed to generate a multi-source fusion dataset, wherein the multi-source fusion dataset contains valid data records aligned to daily granularity and verified for consistency.

[0039] In this embodiment, timestamp alignment refers to matching and aligning body composition data and physiological activity data in chronological order to ensure that data from the same point in time or time period can correspond. Since the sampling frequency and time recording of body fat scales and fitness trackers may differ, timestamp alignment is necessary to ensure data consistency and comparability. Outlier removal involves removing values ​​that significantly deviate from the normal range. These outliers may be due to equipment malfunction, measurement errors, or specific user behaviors. The presence of outliers affects the accuracy of subsequent data analysis, therefore they need to be removed.

[0040] In this embodiment, the multi-source fusion dataset is a dataset obtained by integrating body composition data and physiological activity data after timestamp alignment and outlier removal. This dataset is aligned at the daily granularity, meaning that data from each day is integrated together, facilitating daily-level analysis. Simultaneously, consistency checks ensure that the data in the dataset is logically self-consistent and contains no contradictory or unreasonable combinations.

[0041] S30: Construct an individualized health status trajectory based on the multi-source fusion dataset, and identify the synergistic change pattern of body composition fluctuation and physiological activity deviation over multiple consecutive days, wherein the synergistic change pattern includes a combination of features such as increased body fat percentage accompanied by a sustained increase in resting heart rate or a significant reduction in sleep duration.

[0042] In this embodiment, the individualized health status trajectory is constructed based on the user's multi-source fusion dataset, reflecting changes in the user's health status over a period of time. By analyzing the coordinated change patterns of body composition fluctuations and deviations in physiological activity, potential trends in the user's health status can be identified. An increase in body fat percentage may indicate increased fat accumulation in the user, while a sustained increase in resting heart rate or a significant reduction in sleep duration may be related to factors such as stress response and metabolic disorders. These coordinated change patterns may be early signals of potential health risks.

[0043] In this embodiment, identifying collaborative change patterns requires a comprehensive analysis of body composition indicators and physiological activity indicators in the multi-source fusion dataset. By calculating the changing trends and deviations of these indicators, time periods that simultaneously meet specific combination conditions are selected. These time periods are marked as potential risk evolution segments, and their corresponding collaborative change patterns are extracted.

[0044] S40: Match the collaborative change pattern with the preset health risk triggering rules to determine whether there is a dynamic evolution path that meets the risk conditions, wherein the health risk triggering rules define the joint deviation threshold relationship between body composition and physiological activity indicators in the time dimension.

[0045] In this embodiment, the preset health risk triggering rules are formulated based on medical research and clinical experience to determine whether synergistic changes in body composition and physiological activity indicators constitute a health risk. These rules define the joint deviation threshold relationship between body composition and physiological activity indicators over a time dimension, such as an increase in body fat percentage exceeding a certain proportion and a continuous increase in resting heart rate exceeding a preset number of days. By matching the synergistic change pattern with these rules, it can be determined whether there is a dynamic evolution path that meets the risk conditions.

[0046] In this embodiment, the dynamic evolution path refers to the change process of body composition and physiological activity indicators over a period of time. When this change process meets the health risk triggering rules, it means that there is a potential health risk. The matching judgment process involves comparing the combination of indicators in the collaborative change pattern with the rule entries in the rule base one by one, finding the matching rule identifier, and verifying whether the corresponding temporal duration condition is met.

[0047] S50: When a dynamic evolution path that meets the risk conditions exists, a health risk warning sign is generated and output to the user terminal, wherein the health risk warning sign is a graded warning signal corresponding to the dynamic evolution path.

[0048] In this embodiment, the health risk warning indicator is a signal generated to alert users to potential health risks. The tiered warning signals are classified according to the severity of the risk, for example, into low risk, medium risk, and high risk. Different levels of warning signals can use different colors, icons, or text prompts to help users intuitively understand the degree of risk.

[0049] In this embodiment, a health risk warning indicator is output to a user terminal, such as a mobile phone or tablet computer. Users can view the warning information on the terminal and take corresponding measures according to the warning level. For example, at low risk, diet and exercise habits can be adjusted appropriately; at medium risk, it may be necessary to increase exercise or consult a professional doctor; at high risk, medical attention should be sought promptly for further examination and diagnosis.

[0050] In one embodiment, when a dynamic evolution path that meets risk conditions is identified, the health management application on the user's terminal generates corresponding graded warning signals based on the severity of the risk. For example, a low-risk warning can be indicated by a green icon and mild text, a medium-risk warning by a yellow icon and more serious text, and a high-risk warning by a red icon and strong text. These warning signals are then displayed on the application's main interface and can also be pushed to the user for notification. By generating and outputting graded warning signals, potential health risk information can be promptly communicated to the user, helping them take appropriate intervention and adjustment measures, thereby protecting their health.

[0051] In one embodiment, reference Figure 3 Step S20 may include steps S21-S25, which will be described in detail below: S21: Using the measurement date of the body composition data in the dataset to be processed as the primary key, match the statistical summary of the physiological activity data on the same day to establish a day-level alignment mapping table, wherein the day-level alignment mapping table ensures that there is only one fusion record per day.

[0052] In this embodiment, the primary key is a field in the database used to uniquely identify each record. Using the measurement date of body composition data as the primary key allows for matching body composition data with statistical summaries of physiological activity data from the same day. A daily alignment mapping table is a data structure used to store the correspondence between body composition data and physiological activity data at the daily level, ensuring that there is only one fused record per day and avoiding data duplication and confusion.

[0053] In one embodiment, it is assumed that the dataset to be processed contains statistical summaries of body composition data and physiological activity data. The body composition data records information such as measurement date, weight, and body fat percentage, while the physiological activity data statistical summaries record information such as date, average daily heart rate, and total steps. The measurement date of the body composition data is used as the primary key, and records of the same date are searched in the physiological activity data statistical summaries. Then, the body composition data and the corresponding physiological activity data statistical summaries are combined into a single fused record and stored in a day-level alignment mapping table. For example, if the measurement date of the body composition data is 2024-10-10, the record corresponding to 2024-10-10 is found in the physiological activity data statistical summaries, and the two information are merged and stored in the day-level alignment mapping table. In this way, the day-level alignment mapping table contains only one fused record per day, facilitating subsequent querying and analysis. By establishing a day-level alignment mapping table, day-level alignment of body composition data and physiological activity data can be achieved, improving data organization and usability.

[0054] S22: Perform sliding window median filtering on each index in the body composition data, identify and mark abrupt change points, wherein the abrupt change points are values ​​that deviate from the median of the window by more than a preset proportion.

[0055] In this embodiment, sliding window median filtering is a data processing method that calculates the median of the data within a fixed-size window by sliding it across the data sequence, and then replaces the data value at the center of the window with the median. This method can effectively remove noise and outliers from the data and smooth the data curve. A mutation point refers to a value in the data sequence that suddenly deviates from the normal range. It may be caused by measurement errors, special user behavior, or sudden changes in physical condition. By setting a preset proportion, when a data value deviates from the median of the window by more than that proportion, it is marked as a mutation point.

[0056] In one embodiment, it is assumed that the weight index in the body composition data exists as a data sequence, and the sliding window size is set to 5 days. Starting from the first data point in the data sequence, weight data for 5 consecutive days is selected, and the median of these 5 days is calculated. The median is used as the replacement value for the center data of the window, and then the window is slid one position to the right, continuing to calculate the median of the data in the new window and replacing the center data, and so on, until the entire data sequence is processed. Next, a preset ratio of 10% is set, and for the filtered data sequence, the window is slid again, and the median of the data within the window is calculated. When a data value deviates from the median of the window by more than 10%, the data point is marked as a mutation point. For example, in a 30-day weight data sequence, after sliding window median filtering, if the weight value on day 15 deviates from the median of its window by 12%, then the weight data on day 15 is marked as a mutation point. By performing sliding window median filtering and mutation point marking, body composition data can be effectively smoothed, abnormal changes in the data can be identified, and interference from outliers on health risk analysis results can be avoided.

[0057] S23: Perform continuity detection on the heart rate sequence in the physiological activity data and remove constant value segments caused by device detachment, wherein the constant value segment is a heart rate subsequence with a duration exceeding a threshold and a constant value.

[0058] In this embodiment, continuity detection refers to checking the continuity and changes in heart rate data. Device detachment may cause the fitness tracker to fail to collect heart rate data properly, resulting in constant value segments—that is, heart rate values ​​remaining unchanged for a period of time. By setting a threshold, when the duration of a constant value segment exceeds this threshold, the segment is determined to be invalid data caused by device detachment and needs to be discarded.

[0059] In one embodiment, reference Figure 4Step S23 may include sub-steps from steps S231 to S235, which will be described in detail below.

[0060] S231: Divide the heart rate sequence into continuous segments in chronological order, and calculate the absolute value of the difference between adjacent sampling points within each segment, wherein the absolute value of the difference reflects the amplitude of heart rate fluctuation.

[0061] In this embodiment, the heart rate sequence is divided into continuous segments according to time order, which facilitates local analysis of the heart rate data. The absolute value of the difference refers to the absolute value of the difference between the heart rate values ​​of adjacent sampling points, which can intuitively reflect the degree of fluctuation of the heart rate between adjacent time points. By calculating the absolute value of the difference, the fluctuation amplitude of the heart rate can be quantified, providing a basis for subsequent identification of constant value segments.

[0062] In one embodiment, assuming the heart rate sequence records 120 heart rate values ​​at a sampling frequency of one minute, it is divided into consecutive 10-minute segments in chronological order, with each segment containing 10 sampling points. For the heart rate values ​​within the first segment, the absolute value of the difference between adjacent sampling points is calculated sequentially. For example, if the heart rate values ​​within the segment are 70, 72, 71, 73, 72, 74, 73, 75, 74, and 76, the absolute value of the difference between the first adjacent sampling point is |72-70|=2, the second is |71-72|=1, and so on, resulting in a sequence of absolute values ​​of the difference between adjacent sampling points within the segment: [2, 1, 2, 1, 2, 1, 2, 1, 2]. By calculating the absolute values ​​of the differences, the fluctuation of the heart rate within the segment can be clearly observed. If the absolute values ​​of the differences are generally small, there may be a possibility of a constant value segment. By segmenting the heart rate sequence and calculating the absolute value of the difference, we can delve deeper into the fluctuation characteristics of heart rate data, laying the foundation for accurately identifying constant value segments caused by equipment detachment.

[0063] S232: For each segment, the percentage of sampling points whose absolute difference is less than a first preset value is calculated, wherein the first preset value represents the threshold value without physiological fluctuations.

[0064] In this embodiment, the first preset value is a pre-set threshold used to determine whether the heart rate is in a state without significant physiological fluctuations. The percentage of sampling points whose absolute difference is less than the first preset value can be used to assess the stability of the heart rate in each segment. If the percentage is too high, it indicates that the heart rate in that segment fluctuates little, and there may be segments with constant values.

[0065] In one embodiment, a first preset value is set to 2. For the aforementioned divided heart rate segments, the number of sampling points with an absolute difference value less than 2 is counted. Assuming that in a segment containing 10 sampling points, there are 8 sampling points with an absolute difference value less than 2, then the proportion of sampling points with an absolute difference value less than the first preset value in that segment is 8 / 10 = 80%. By performing this statistical analysis on each segment, the stability assessment result for each segment can be obtained. The higher the proportion, the more likely the segment is to be a suspected constant segment. By statistically analyzing the proportion, the fluctuation of heart rate segments can be judged more objectively, improving the efficiency of identifying suspected constant segments.

[0066] S233: If the proportion of a certain segment exceeds the second preset value, the segment is determined to be a suspected constant segment, wherein the second preset value defines the proportion threshold of a high probability constant state.

[0067] In this embodiment, the second preset value is a proportional threshold used to determine whether a segment is a suspected constant segment. When the proportion of sampling points with an absolute difference value less than the first preset value in a certain segment exceeds the second preset value, it indicates that the internal rate fluctuation of the segment is extremely small, and there is a high probability that it is in a constant state, so it is determined to be a suspected constant segment.

[0068] In one embodiment, the second preset value is set to 70%. For a heart rate segment, if the percentage of sampling points with an absolute difference less than the first preset value is calculated to be 85%, and since 85% is greater than 70%, the segment is determined to be a suspected constant segment. For example, in a heart rate sequence containing 100 segments, after statistical analysis and judgment, 10 segments are found to have a percentage exceeding 70%, and these 10 segments are marked as suspected constant segments. By identifying suspected constant segments, the scope requiring further investigation can be narrowed down, improving the efficiency of data processing.

[0069] S234: For suspected constant segments, further check whether their duration exceeds a third preset value, where the third preset value corresponds to the lower limit of the typical duration of device detachment.

[0070] In this embodiment, the third preset value is a pre-set time threshold, representing the lower limit of the typical duration for which device detachment causes heart rate data to remain constant. Checking the duration of suspected constant segments can determine whether the segment is truly invalid data caused by device detachment. If the duration exceeds the third preset value, it is more likely due to device detachment.

[0071] In one embodiment, a third preset value is set to 30 minutes. For a heart rate segment identified as a suspected constant segment, its start and end timestamps are obtained, and the time difference between them is calculated. Assuming a suspected constant segment starts at 9:00 AM and ends at 9:40 AM, the time difference is 40 minutes. Since 40 minutes is greater than 30 minutes, this segment is likely invalid data due to device detachment. If another suspected constant segment has a time difference of 20 minutes, which is less than 30 minutes, it may be a normal physiological heart rate stabilization period, and this segment data should be retained. By further examining the duration, constant value segments caused by device detachment can be identified more accurately, improving the quality of heart rate data.

[0072] In one embodiment, step S234 can be implemented as follows: K1: Obtain the start and end timestamps of the suspected constant segment and calculate the time difference between them, where the time difference is in minutes.

[0073] In this embodiment, the start timestamp and end timestamp record the start and end times of the suspected constant segment, respectively. Calculating the time difference between the two and expressing it in minutes allows for an accurate measurement of the duration of the suspected constant segment, providing a temporal basis for determining whether it is invalid data caused by equipment detachment.

[0074] In one embodiment, assume the start timestamp of a suspected constant segment is 2024-10-15 09:10:00 and the end timestamp is 2024-10-15 09:40:00. Converting the start and end times to minutes, the start time is 9×60+10=550 minutes and the end time is 9×60+40=580 minutes. The time difference between the two is 580-550=30 minutes. By accurately calculating the time difference, the duration of the suspected constant segment can be clearly understood, providing crucial data for subsequent judgment.

[0075] K2: Compare the time difference with a third preset value, wherein the third preset value is a fixed time length.

[0076] In this embodiment, the third preset value is a pre-set fixed time length, representing the lower limit of the typical duration for which the heart rate data remains constant due to device detachment. By comparing the calculated time difference with the third preset value, it can be determined whether the suspected constant segment is more likely to be invalid data caused by device detachment.

[0077] In one embodiment, a third preset value is set to 25 minutes, and the calculated time difference of the suspected constant segment is 30 minutes. Since 30 minutes is greater than 25 minutes, it indicates that the suspected constant segment is likely invalid data caused by device detachment. If the time difference is 20 minutes and less than 25 minutes, it may be a normal physiological heart rate stabilization period. This numerical comparison method can effectively filter out potentially invalid data.

[0078] K3: If the time difference is greater than the third preset value, then the segment is confirmed as invalid data caused by the device falling off, and the confirmation result is used for data rejection decision.

[0079] In this embodiment, when the time difference is greater than a third preset value, there is a high probability that the suspected constant segment is due to abnormal heart rate data caused by device detachment. After confirming that the segment is invalid data, the result will be used to determine whether to remove the segment from the original heart rate sequence.

[0080] In one embodiment, the third preset value is 30 minutes. A suspected constant time difference is 35 minutes. Since 35 minutes is greater than 30 minutes, this segment is confirmed as invalid data caused by device detachment. In subsequent processing, this data segment is removed from the original heart rate sequence to ensure the validity of the heart rate data. By identifying and removing invalid data, the heart rate data can more accurately reflect the user's true physiological state.

[0081] K4: If the time difference is less than or equal to the third preset value, the segment is retained as valid physiological data, wherein the retention operation maintains the integrity of the original sequence.

[0082] In this embodiment of the application, when the time difference is less than or equal to a third preset value, it indicates that the suspected constant segment may be a normal physiological heart rate stabilization period, and retaining the segment as valid physiological data can maintain the integrity of the original heart rate sequence.

[0083] In one embodiment, the third preset value is 25 minutes, and the time difference of a suspected constant segment is 20 minutes. Since 20 minutes is less than 25 minutes, this segment is retained as valid physiological data. In this way, the integrity of the original heart rate sequence is maintained, and this segment of data can be used for subsequent health analysis. By retaining valid physiological data, useful information in the heart rate data can be fully utilized, improving the reliability of health analysis.

[0084] K5: Write the confirmation result to the data quality tag field for use in subsequent steps, where the data quality tag field indicates the validity status of the data for that time period.

[0085] In this embodiment, the data quality marker field is used to record the data validity status. Writing the confirmation result into this field allows subsequent steps to quickly understand the validity of the data for that time period, providing a reference for subsequent data processing and analysis.

[0086] In this embodiment of the application, by writing a data quality tag field, the sharing and transmission of data validity information can be realized, thereby improving the efficiency and accuracy of data processing.

[0087] In one embodiment, for a suspected constant segment, after confirming it as invalid data, the "invalid" status is written to the data quality flag field. During subsequent data analysis, the system can automatically ignore invalid data and only process valid data based on this flag. If the result is confirmed as valid data, the "valid" status is written to the field. This method ensures that subsequent steps can correctly use the data and avoids analysis errors caused by data validity issues.

[0088] S235: If the duration exceeds a third preset value, the segment is marked as a constant value segment and removed from the original sequence, wherein the removal operation generates a corrected heart rate sequence.

[0089] In this embodiment, when the duration of a suspected constant segment exceeds a third preset value, it is marked as a constant segment and removed from the original heart rate sequence. This removes invalid heart rate data and improves the quality of the heart rate sequence. The corrected heart rate sequence generated after the removal operation contains only valid physiological heart rate data, making it more suitable for subsequent health analysis.

[0090] In one embodiment, an inspection revealed a suspected constant segment lasting 45 minutes, exceeding a third preset value of 30 minutes. This segment was then marked as a constant segment. Assuming the original heart rate sequence contains 200 sampling points, the constant segment corresponds to sampling points 50 through 90. These 40 sampling points are removed from the original sequence, generating a corrected heart rate sequence containing 160 sampling points. By marking and removing constant segments, the heart rate sequence becomes cleaner, avoiding interference from invalid data in health risk analysis results and improving the accuracy of the analysis.

[0091] S24: Set the daily record corresponding to the marked mutation point and the removed constant value segment to an invalid state, wherein the invalid state prevents the record from participating in subsequent fusion.

[0092] In this embodiment, the marked mutation points are anomalous data points identified in body composition data, and the removed constant value segments are invalid data segments detected in the heart rate sequence of physiological activity data. Setting the daily records corresponding to these anomalous data to an invalid state is to avoid the impact of these anomalous data on subsequent multi-source data fusion and analysis. An invalid state means that the record will not participate in subsequent calculations and analysis, thereby ensuring the accuracy and reliability of the data.

[0093] In one embodiment, suppose that a weight data point for a particular day is marked as a mutation point in the body composition data, and a segment of constant values ​​for that day is removed from the heart rate sequence of the physiological activity data. Then, the entire daily record, composed of the statistical summary of the body composition data and physiological activity data for that day, is set to invalid. For example, in a daily alignment mapping table, the record for that day is marked as invalid, and the system will automatically ignore these invalid records during subsequent data fusion and analysis. By invalidating the entire daily record corresponding to abnormal data, it is possible to ensure that the data participating in subsequent fusion and analysis is reliable and accurate, thereby improving the precision and reliability of health risk analysis.

[0094] S25: Perform cross-source logical consistency checks on the remaining valid records, exclude contradictory combinations such as decreased body fat percentage but significant weight gain, and generate a multi-source fusion dataset, wherein the multi-source fusion dataset contains only logically consistent daily records.

[0095] In this embodiment, cross-source logical consistency checking refers to a comprehensive analysis of statistical summaries of body composition data and physiological activity data to check whether the data from different data sources are logically consistent. A decrease in body fat percentage coupled with a significant increase in weight is physiologically unreasonable and constitutes a contradictory combination. By eliminating these contradictory combinations, it can be ensured that the data in the multi-source fused dataset is logically self-consistent.

[0096] In one embodiment, for the remaining valid records, a significant weight gain threshold of 2 kg is set. When body fat percentage decreases on a given day while weight gain exceeds 2 kg, the record is considered an inconsistent combination. These inconsistent combination records are excluded from the dataset, and the remaining records form a multi-source fusion dataset. For example, in a dataset containing 50 valid records, cross-source logical consistency checks identify 5 records showing a decrease in body fat percentage but a significant increase in weight. These 5 records are excluded, resulting in a multi-source fusion dataset containing 45 records. By performing cross-source logical consistency checks and excluding inconsistent combinations, the logical consistency of the multi-source fusion dataset is ensured, improving the accuracy and reliability of health risk analysis.

[0097] In one embodiment, reference Figure 5 Step S30 may include steps S31-S35, which will be described in detail below: S31: Extract a data sequence of N consecutive days from the multi-source fusion dataset, where N is an integer greater than or equal to three, and construct individual observation windows, wherein the individual observation windows are arranged in chronological order.

[0098] In this embodiment, the individual observation window is a time range used to observe and analyze changes in a user's health status. Constructing the individual observation window from a continuous N-day data sequence extracted from a multi-source fusion dataset allows for focusing on changes in a user's health data over a period of time. Setting N to an integer greater than or equal to three ensures that the observation window has sufficient length to reflect data trends; if the number of days is too small, it may be impossible to accurately identify the coordinated change patterns of body composition fluctuations and deviations in physiological activity.

[0099] In one embodiment, assume that the multi-source fusion dataset contains a statistical summary of a user's body composition and physiological activity data for 30 consecutive days. Setting N to 5, starting from day one of the dataset, a 5-day data sequence is extracted as the first individual observation window, containing data from day 1 to day 5. The window is then moved forward one day, extracting data from day 2 to day 6 as the second individual observation window, and so on, until all possible 5-day data sequences have been extracted. For example, in a dataset containing 30 days of data, a total of 26 individual observation windows can be constructed. By constructing individual observation windows arranged chronologically, it is possible to systematically analyze changes in a user's health status over different time periods, providing an effective data organization method for identifying patterns of coordinated change and helping to more accurately capture the dynamic changes in body composition and physiological activity indicators.

[0100] S32: Calculate the slope of the change trend of the volume component index within the observation window, and mark the index that is monotonically rising or falling, wherein the slope of the change trend is determined by the cumulative direction of the difference between adjacent days.

[0101] In this embodiment, the slope of the trend is used to measure the direction and extent of change of body composition indicators within the observation window. By calculating the cumulative direction of the difference between adjacent days to determine the slope of the trend, it is possible to intuitively reflect whether the indicator is trending upward or downward. Monotonically increasing or decreasing indicators refer to those whose values ​​continuously increase or decrease within the observation window. Labeling these indicators helps to identify stable trends in body composition, providing a basis for subsequent analysis of the synergistic relationship between body composition fluctuations and deviations in physiological activity.

[0102] In one embodiment, for body fat percentage, a body composition indicator, within a 5-day individual observation window, the body fat percentage values ​​for days 1 to 5 are recorded as 20%, 21%, 22%, 23%, and 24%, respectively. The difference between adjacent days is calculated: the difference between day 2 and day 1 is 21% - 20% = 1%, the difference between day 3 and day 2 is 22% - 21% = 1%, and so on, resulting in a difference sequence of [1%, 1%, 1%, 1%]. All elements in the difference sequence are positive, indicating that body fat percentage shows a monotonically increasing trend within the observation window. By statistically analyzing the length of consecutive positive segments with non-zero signs in the difference sequence, it is found that the entire difference sequence is continuously positive, and its length accounts for 100% of the total number of days in the observation window, exceeding a preset trend significance threshold (e.g., 80%). Therefore, body fat percentage is marked as a monotonically increasing indicator. By calculating the slope of the trend and marking monotonically changing indicators, the changing trend of body composition indicators can be accurately identified.

[0103] In one embodiment, reference Figure 6 Step S32 may include sub-steps from steps S321 to S325, which will be described in detail below.

[0104] S321: For the daily values ​​of body component indicators within the observation window, calculate the difference between two adjacent days to form a difference sequence, wherein the difference sequence represents the amount of diurnal variation.

[0105] In this embodiment, the difference between body composition indices on two adjacent days is calculated and a difference sequence is formed, which can intuitively reflect the changes in body composition indices between days. Each element in the difference sequence represents the change in body composition indices between two adjacent days. By analyzing this sequence, the short-term fluctuations of body composition indices can be understood.

[0106] In one embodiment, assuming the observation window is a continuous 7 days, and the body composition index is weight, the daily weight values ​​are 60 kg, 61 kg, 60.5 kg, 62 kg, 62.5 kg, 63 kg, and 64 kg. The difference between adjacent days is calculated: the difference between day 2 and day 1 is 61 - 60 = 1 kg, the difference between day 3 and day 2 is 60.5 - 61 = -0.5 kg, and so on, resulting in a difference sequence [1, -0.5, 1.5, 0.5, 0.5, 1]. This difference sequence clearly shows the daily changes in weight; for example, weight may decrease slightly between day 2 and day 3, while increasing at other times. Calculating the difference between adjacent days to form a difference sequence provides a quantitative basis for analyzing short-term changes in body composition index, helping to more accurately grasp the dynamic changes in body composition index.

[0107] S322: For each element in the difference sequence, determine whether its sign is positive, negative or zero, and generate a sign sequence, wherein the sign sequence reflects the daily change direction.

[0108] In this embodiment, the sign sequence is generated by determining the sign of each element in the difference sequence. It can concisely reflect the direction of change of the body composition index from day to day. A positive sign indicates that the index is rising, a negative sign indicates that the index is falling, and zero indicates that the index is unchanged.

[0109] In one embodiment, for the difference sequence [1, -0.5, 1.5, 0.5, 0.5, 1] ​​obtained above, the sign of each element is determined, resulting in a sign sequence [+, -, +, +, +, +]. The direction of weight change on different dates can be quickly observed from the sign sequence; for example, weight decreases from day 2 to day 3, and then increases continuously from day 3 onwards. By generating the sign sequence, the direction of change in body composition indicators can be more clearly displayed, providing a convenient way to determine the monotonic trend of these indicators.

[0110] S333: The length of a continuous segment of non-zero symbols in a statistical symbol sequence, wherein the continuous segment is a subsequence of symbols that are identical and appear consecutively.

[0111] In this embodiment, consecutive unidirectional segments are subsequences in a symbol sequence where the symbols are identical and appear consecutively. Statistical analysis of their length can reveal the continuous changes in volume component indices over a period of time. Longer consecutive unidirectional segments indicate that the indices exhibit a relatively stable trend of change within that time period.

[0112] In one embodiment, for the symbol sequence [+, -, +, +, +, +], two consecutive segments in the same direction can be found: one is a negative segment (-) of length 1, and the other is a positive segment (+, +, +, +) of length 4. By statistically analyzing the length of these consecutive segments, it can be found that body weight shows an upward trend for most of the time, especially a relatively long continuous upward segment from day 3 to day 7. Statistical analysis of the length of these consecutive segments allows for a more in-depth analysis of the changing trends of body composition indicators, providing data support for accurately identifying monotonic changes.

[0113] S324: If the length of the longest continuous segment in the same direction accounts for more than the proportion of the total number of days in the observation window, the indicator is determined to have a monotonic trend, wherein the fourth preset value defines the proportion threshold for the significance of the trend.

[0114] In this embodiment, the fourth preset value is a proportional threshold used to determine whether an indicator exhibits a monotonic trend. When the proportion of the longest continuous segment in the same direction to the total number of days in the observation window exceeds the fourth preset value, it indicates that the indicator has a significant monotonic trend within the observation window.

[0115] In one embodiment, the fourth preset value is set to 60%. The total observation window is 7 days, and the longest continuous unidirectional segment is 4 days. The proportion of this segment to the total observation window is 4 / 7 ≈ 57%, which is less than 60%, so the indicator is not considered to have a monotonic trend. If another observation window has a total of 6 days and the longest continuous unidirectional segment is 4 days, the proportion is 4 / 6 ≈ 67%, which is greater than 60%, then the indicator is considered to have a monotonic trend. In this way, based on the actual changes in the indicator, it is possible to reasonably determine whether it has a monotonic trend, providing a more accurate basis for subsequent health risk analysis.

[0116] S325: Based on the sign of the longest consecutive segment in the same direction, mark the indicator as monotonically increasing or monotonically decreasing.

[0117] In this embodiment, the monotonic change direction of the volume component index can be determined based on the sign of the longest consecutive segment in the same direction. If the sign of the longest consecutive segment in the same direction is positive, the index is marked as monotonically increasing; if the sign is negative, it is marked as monotonically decreasing.

[0118] In one embodiment, if the longest consecutive segment in the sign sequence of a certain body composition indicator is found to be positive, it indicates that the indicator shows an upward trend for most of the observation window, and the indicator is marked as monotonically increasing. For example, for the weight indicator, if the longest consecutive segment in the sign sequence is (+, +, +, +), then the weight indicator is marked as monotonically increasing. By marking the monotonically changing direction of the indicator, the changing characteristics of the body composition indicator can be clearly displayed, which helps to analyze the correlation between multi-source data more deeply and provide more targeted information for health risk warning.

[0119] S33: Simultaneously calculate the degree of deviation of physiological activity indicators within the same observation window, and identify indicators that are lower or higher than the historical average for multiple consecutive days, wherein the historical average is calculated based on the user's data over the past thirty days.

[0120] In this embodiment, the deviation of a physiological activity indicator refers to the degree of difference between the indicator's value within the observation window and its historical mean. By calculating the deviation, it is possible to understand the comparison between the user's current physiological activity state and its average state over a past period. The historical mean is calculated based on the user's data from the past thirty days, reflecting the user's average physiological activity level over a longer period. Identifying indicators that are consistently below or above the historical mean for several days helps to detect abnormal changes in the user's physiological activity, which may indicate changes in health status.

[0121] In one embodiment, taking the daily average heart rate as an example of physiological activity indicators, it is assumed that the historical average daily heart rate calculated based on the user's data over the past 30 days is 72 beats / minute. Within a 5-day individual observation window, the daily average heart rates are 78, 79, 80, 81, and 82 beats / minute, respectively. The difference between the daily average heart rate and the historical average for each day is calculated to be 6, 7, 8, 9, and 10 beats / minute, respectively, indicating that the daily average heart rate is consistently higher than the historical average within the observation window. By identifying this situation of consecutive days exceeding the historical average, it can be inferred that the user's body may be in a state of stress, such as increased exercise, physical fatigue, or potential health problems. By simultaneously calculating the deviation of physiological activity indicators and identifying abnormal indicators, abnormal changes in physiological activity can be detected in a timely manner. Combined with the analysis of body composition indicators, this improves the accuracy of health risk assessment.

[0122] S34: Cross-compare the changing trends of body composition indicators with the deviation directions of physiological activity indicators to screen out windows that simultaneously meet specific combination conditions, wherein the specific combination conditions are defined as an increase in body fat percentage and a continuous increase in resting heart rate.

[0123] In this embodiment, cross-comparison refers to comprehensively comparing the changing trend (increasing or decreasing) of body composition indicators with the deviation direction (above or below historical average) of physiological activity indicators. Specific combination conditions are set based on medical research and clinical experience; a combination of increased body fat percentage and continuously elevated resting heart rate may indicate potential health risks for the user, such as an increased risk of cardiovascular disease. Screening for windows that meet specific combination conditions allows focusing on time periods with a higher probability of health risk.

[0124] In one embodiment, within a series of individual observation windows, for each window, the trend of body fat percentage change (a body composition indicator) and the deviation of resting heart rate (a physiological activity indicator) are considered simultaneously. Assume that within a 7-day individual observation window, body fat percentage increases monotonically from 18% on day 1 to 20% on day 7; resting heart rate continuously increases from 68 beats / minute on day 1 to 75 beats / minute on day 7, consistently exceeding the historical average (assuming the historical average is 65 beats / minute). Since this window simultaneously meets the specific combination of increasing body fat percentage and continuously increasing resting heart rate, it is selected. By cross-referencing and filtering multiple individual observation windows, time periods with potential health risks can be identified, providing a basis for subsequent health risk warnings and improving the targeting and accuracy of warnings.

[0125] S35: Mark the selected windows as potential risk evolution segments and extract their corresponding collaborative change patterns, wherein the collaborative change patterns are characterized by a combination of indicators and their change direction.

[0126] In this embodiment, the potential risk evolution segment refers to individual observation windows that meet specific combinations of conditions. Changes in body composition and physiological activity indicators within these windows may indicate that potential health risks are evolving. The synergistic change pattern describes the interrelationships and patterns of change between body composition and physiological activity indicators. Using combinations of indicators and their changing directions as characteristic features, it clearly expresses the synergistic relationships between different indicators.

[0127] In one embodiment, multiple individual observation windows that meet the criteria of increasing body fat percentage and continuously increasing resting heart rate are selected through cross-comparison and labeled as potential risk evolution segments. For each potential risk evolution segment, its corresponding co-variation pattern is extracted, for example, body fat percentage increases by X%, and resting heart rate continuously increases by Y beats / minute. Taking a 6-day potential risk evolution segment as an example, if body fat percentage increases from 19% to 21% (an increase of 2%) and resting heart rate continuously increases from 70 beats / minute to 76 beats / minute (an increase of 6 beats / minute), then the co-variation pattern corresponding to this potential risk evolution segment can be represented as (body fat percentage increases by 2%, resting heart rate continuously increases by 6 beats / minute). By labeling potential risk evolution segments and extracting co-variation patterns, key information from multi-source data can be extracted, facilitating subsequent matching with preset health risk triggering rules and improving the effectiveness of health risk early warning.

[0128] In one embodiment, reference Figure 7 Step S40 may include steps S41-S45, which will be described in detail below: S41: Compare the combination of indicators in the collaborative change mode with the rule entries in the health risk trigger rule base one by one to find the matching rule identifier, wherein the rule entry uses the indicator pair and its change direction as the index key.

[0129] In this embodiment, the health risk triggering rule base is a pre-defined set of rules used to determine whether a coordinated change pattern constitutes a health risk. Rule entries are indexed using indicator pairs and their changing directions for easy and quick searching and matching. The indicator combinations in the coordinated change pattern reflect the coordinated change relationship between body composition indicators and physiological activity indicators; comparing these with rule entries in the rule base determines whether a matching rule exists.

[0130] In one embodiment, it is assumed that the health risk triggering rule base contains the following rule entries: Rule 1 (increased body fat percentage and continuously rising resting heart rate), Rule 2 (decreased body fat percentage and significantly shortened sleep duration), etc. An extracted co-change pattern is an increase in body fat percentage and a continuous increase in resting heart rate. This indicator combination is compared one by one with the rule entries in the rule base. A match with Rule 1 is found, and this rule is identified as "Rule 1". This comparison method allows for rapid identification of rules matching the co-change pattern, providing a basis for subsequent risk assessment. By comparing the co-change pattern with the rule base, potential health risks can be accurately identified, improving the accuracy and targeting of health risk warnings.

[0131] S42: For a successfully matched rule entry, verify whether its corresponding time-series persistence condition is met, wherein the time-series persistence condition requires that the coordinated change last for no less than a preset number of days.

[0132] In this embodiment, the duration condition is a crucial component of the rule entries. It specifies the minimum number of days that the coordinated change needs to last to ensure that the coordinated change is not a random fluctuation but a health risk signal with a certain degree of stability and persistence. Verifying whether the duration condition is met allows for further screening out truly risky situations.

[0133] In one embodiment, reference Figure 8 Step S42 may include sub-steps from steps S421 to S424, which will be described in detail below.

[0134] S421: Extract the start and end dates of the cooperative change pattern corresponding to the matching rule entry in the observation window, and calculate the duration in days, wherein the duration in days is the calendar interval between the end date and the start date plus one.

[0135] In this embodiment, extracting the start and end dates of the cooperative change pattern within the observation window clarifies the time range of the pattern. The calculation of the duration by adding one to the calendar interval between the start and end dates is used to accurately include both dates, thus obtaining the actual number of days the cooperative change pattern lasts.

[0136] In one embodiment, assuming the matching rule entry is "increased body fat percentage and continuously increasing resting heart rate," the corresponding coordinated change pattern has a start date of October 5, 2024, and an end date of October 9, 2024, within the observation window. The calendar interval between the end date and the start date is calculated: October 9 minus October 5 equals 4 days, plus 1 day, resulting in a duration of 5 days. Accurately calculating the duration provides accurate data support for subsequently determining whether the coordinated change pattern meets the time-series duration condition.

[0137] S422: Compare the calculated duration with the minimum duration defined in the rule entry, where the minimum duration is an inherent parameter of the rule.

[0138] In this embodiment, the minimum duration of days defined in the rule entry is the rule's requirement for the duration of the collaborative change pattern, and is an inherent parameter of the rule. By comparing the calculated duration of days with the minimum duration of days, it can be determined whether the collaborative change pattern meets the temporal duration condition of the rule.

[0139] In one embodiment, the minimum duration defined in the rule entry is 3 days, and the calculated duration of the collaborative change pattern is 5 days. Since 5 days is greater than 3 days, it indicates that the collaborative change pattern meets the temporal duration condition of the rule. If the calculated duration is 2 days, which is less than 3 days, then the temporal duration condition is not met. This comparison method clearly determines whether the collaborative change pattern meets the rule requirements, providing a basis for subsequent risk assessment.

[0140] S423: If the duration is greater than or equal to the minimum duration, the time series duration condition is determined to be met, and the determination result triggers the rule activation; if the duration is less than the minimum duration, the match is determined to be a momentary fluctuation and does not constitute a valid risk path, and the momentary fluctuation is excluded from the risk path set.

[0141] In this embodiment, when the duration is greater than or equal to the minimum duration, it indicates that the collaborative change pattern has sufficient duration in time, meets the timing requirements of the rule, and the timing duration condition is determined to be met. The determination result triggering rule activation means that the health risk path corresponding to the rule has been confirmed to exist.

[0142] In one embodiment, the minimum duration of a rule entry is 4 days, and the duration of the coordinated change pattern is also 4 days. Since the duration equals the minimum duration, the time-series duration condition is met. At this point, activating the risk assessment flag for the rule entry indicates that the health risk path corresponding to the coordinated change pattern is valid. For example, for the rule "increased body fat percentage and continuously increased resting heart rate," activating the rule when the duration requirement is met means that the user may have related health risks and a warning is needed.

[0143] In this embodiment, when the duration is less than the minimum duration, it indicates that the coordinated change pattern may only be a short-term fluctuation, lacking sufficient stability and persistence, and is therefore judged as an instantaneous fluctuation, not constituting a valid risk path. Excluding instantaneous fluctuations from the risk path set can avoid misjudging short-term normal physiological changes as health risks.

[0144] In one embodiment, the minimum duration of a rule entry is 5 days, and the duration of a co-change pattern is 3 days. Since the duration is less than the minimum duration, the match is determined to be a transient fluctuation. For example, body fat percentage and resting heart rate may rise simultaneously for a short period of time, but the duration is short-lived and may be due to short-term changes caused by factors such as diet and exercise on that day, and does not constitute a valid health risk path. Excluding such transient fluctuations from the risk path set ensures that the paths in the risk path set all have actual health risk significance.

[0145] S424: Repeat the above verification process for all matching rules to generate a final list of activated rules, wherein the list of activated rules contains only the entries that have passed the time-series verification.

[0146] In this embodiment, the verification process is repeated for all matching rules, ensuring accurate temporal persistence condition verification for each matching rule. The generated final list of activated rules contains only entries that have passed temporal verification, accurately reflecting rules with valid health risk paths.

[0147] In one embodiment, assuming there are 5 matching rules, a verification process is performed on the corresponding collaborative change pattern of each rule, including calculating the duration in days and comparing it with the minimum duration in days. After verification, 3 rules are found to meet the time-series duration condition, and these 3 rules are added to the final list of activated rules. For example, rules 1, 3, and 5 pass the time-series verification, and the final list of activated rules is [rule 1, rule 3, rule 5]. By generating the final list of activated rules, all rules with health risks can be accurately identified, providing comprehensive information for subsequent health risk warnings and interventions.

[0148] S43: If the timing continuity condition is met, the risk determination flag of the rule entry is activated, wherein the risk determination flag indicates that there is a valid risk path.

[0149] In this embodiment, the risk determination flag is an identifier used to mark whether a rule entry has been triggered. When a successfully matched rule entry meets the time-series continuity condition, the risk determination flag of that rule entry is activated, meaning that the situation corresponding to the collaborative change pattern constitutes a valid health risk path.

[0150] In one embodiment, for a rule entry that successfully matches and meets the time-series continuity condition, "body fat percentage increases and resting heart rate increases continuously for at least 5 days," when it is found that body fat percentage and resting heart rate do indeed increase simultaneously for 6 consecutive days in a potential risk evolution segment, the risk assessment flag for this rule entry is activated. For example, a Boolean risk assessment flag is set for this rule entry in the rule base, with an initial value of false, and its value is changed to true when the condition is met. After the risk assessment flag is activated, the rule entry is confirmed to have a valid risk path, and subsequent processing and early warning operations can be performed based on this flag. By activating the risk assessment flag, situations with health risks can be accurately identified.

[0151] S44: Summarize all activated risk assessment flags to generate a risk path set, wherein the risk path set contains at least one valid risk path.

[0152] In this embodiment, the risk path set is a collection of all rule entries that have activated the risk assessment flag. Each activated rule entry represents a valid risk path, and the set contains at least one valid risk path.

[0153] In one embodiment, after performing time-series continuous condition verification on multiple successfully matched rule entries, it was found that the risk assessment markers of three rule entries were activated: Rule A (increased body fat percentage and significantly shortened sleep duration), Rule B (decreased skeletal muscle mass and continuously reduced total steps), and Rule C (decreased basal metabolic rate and increased resting heart rate). These three activated rule entries are summarized to generate a risk path set. This set can be represented in list form, i.e., [Rule A, Rule B, Rule C]. By generating the risk path set, all existing effective risk paths can be clearly displayed, providing a comprehensive information foundation for subsequent health risk warnings and interventions.

[0154] S45: If the set of risk paths is not empty, then it is determined that there exists a dynamic evolution path that satisfies the risk conditions.

[0155] In this embodiment, a non-empty risk path set indicates that at least one rule entry is activated, meaning that at least one valid risk path exists. A dynamic evolution path refers to the change process of body composition indicators and physiological activity indicators over time. When the risk path set is non-empty, it means that this change process meets the preset health risk conditions.

[0156] In one embodiment, the generated risk path set is [Rule D (increased body fat percentage and continuously increasing resting heart rate)]. Since the set is not empty, it indicates the existence of a dynamic evolution path that meets the risk conditions. This means that, based on the analysis of multi-source data, the changes in the user's body composition and physiological activity indicators have reached the preset health risk standards, requiring timely warnings. This judgment method can quickly and accurately identify potential health risks, providing timely prompts for users to take appropriate health management measures.

[0157] Accordingly, to better implement the above methods, this application also provides a dynamic health risk early warning system based on the fusion of multi-source data from a body fat scale and a fitness tracker. For example... Figure 9 As shown, the health risk dynamic early warning system 80 based on the fusion of multi-source data from a body fat scale and a fitness tracker includes: The acquisition module 801 is used to simultaneously acquire body composition data measured by the user on the body fat scale and physiological activity data recorded by the fitness tracker. The body composition data includes weight, body fat percentage, skeletal muscle mass and basal metabolic rate, and the physiological activity data includes heart rate sequence, cumulative step count and sleep duration. Data fusion module 802 is used to perform timestamp alignment and outlier removal on the body composition data and physiological activity data to generate a multi-source fusion dataset, wherein the multi-source fusion dataset contains valid data records aligned to daily granularity and verified for consistency. The change recognition module 803 is used to construct an individualized health status trajectory based on the multi-source fusion dataset and identify the coordinated change pattern of body composition fluctuation and physiological activity deviation over multiple consecutive days. The coordinated change pattern includes a combination of features such as increased body fat percentage accompanied by a sustained increase in resting heart rate or a significant reduction in sleep duration. The matching and judgment module 804 is used to match and judge the collaborative change pattern with the preset health risk triggering rules to determine whether there is a dynamic evolution path that meets the risk conditions, wherein the health risk triggering rules define the joint deviation threshold relationship between body composition and physiological activity indicators in the time dimension. The early warning processing module 805 is used to generate a health risk early warning identifier and output it to the user terminal when there is a dynamic evolution path that meets the risk conditions, wherein the health risk early warning identifier is a graded early warning signal corresponding to the dynamic evolution path.

[0158] The implementation details of each module are provided in the preceding method embodiments and will not be repeated here. The technical effects achieved by each module and device are described in the foregoing method embodiments.

[0159] like Figure 10As shown, this application embodiment also provides a computer device 90 (which may be an intelligent body fat scale), which includes a processor 901 and a memory 902, wherein the memory 902 stores a computer program, and when the computer program is executed by the processor 901, the processor 901 performs the steps of any of the methods described above.

[0160] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application are still within the scope of this application.

Claims

1. A health risk early warning method based on body fat scale and fitness tracker data, characterized in that, include: The system simultaneously collects body composition data measured by the user on the body fat scale and physiological activity data recorded by the fitness tracker. The body composition data includes weight, body fat percentage, skeletal muscle mass, and basal metabolic rate, while the physiological activity data includes heart rate sequence, cumulative step count, and sleep duration. The body composition data and physiological activity data are time-stamped and outliers are removed to generate a multi-source fusion dataset, wherein the multi-source fusion dataset contains valid data records aligned to daily granularity and verified for consistency. Based on the multi-source fusion dataset, an individualized health status trajectory is constructed to identify the synergistic change pattern of body composition fluctuations and physiological activity deviations over multiple consecutive days. The synergistic change pattern includes a combination of features such as increased body fat percentage accompanied by a sustained increase in resting heart rate or a significant reduction in sleep duration. The collaborative change pattern is matched with the preset health risk triggering rules to determine whether there is a dynamic evolution path that meets the risk conditions. The health risk triggering rules define the joint deviation threshold relationship between body composition and physiological activity indicators in the time dimension. When a dynamic evolution path that meets the risk conditions exists, a health risk warning sign is generated and output to the user terminal, wherein the health risk warning sign is a graded warning signal corresponding to the dynamic evolution path.

2. The method according to claim 1, characterized in that, The data collected from the user's body composition measured on the body fat scale and the physiological activity data recorded by the fitness tracker include: After the user completes the body fat scale measurement, the data from the fitness tracker within a set time window before and after the measurement time is extracted, and the body composition data is linked to the physiological activity data of the corresponding time window to establish an association index, wherein the association index is based on the measurement time as the reference anchor point. Units and formats are standardized for each indicator in the body composition data, and the original measured values ​​are converted into structured fields, which are stored in data containers according to indicator type. Physiological activity data are segmented and aggregated according to sampling frequency to generate a daily statistical summary, which includes the average daily heart rate, total steps, and effective sleep duration. The structured volume component data is bound to the daily statistical summary through user identity to form an initial multi-source data pair, wherein the initial multi-source data pair retains the original timestamp information; Perform integrity checks on the initial multi-source data pairs, remove records that are missing from either data source, and generate a data set to be processed, wherein the data set to be processed contains only entries where both source data are complete.

3. The method according to claim 2, characterized in that, The process of aligning the body composition data and physiological activity data with timestamps and removing outliers to generate a multi-source fusion dataset includes: Using the measurement date of the body composition data in the dataset to be processed as the primary key, and matching the statistical summary of the physiological activity data on the same day, a day-level alignment mapping table is established, wherein the day-level alignment mapping table ensures that there is only one fusion record per day. Perform sliding window median filtering on each indicator in the body composition data to identify and mark abrupt change points, wherein the abrupt change points are values ​​that deviate from the median of the window by more than a preset proportion; The heart rate sequence in physiological activity data is continuously detected, and constant value segments caused by device detachment are removed. The constant value segments are heart rate subsequences whose duration exceeds a threshold and whose value remains unchanged. Set the daily records corresponding to the marked mutation points and the removed constant value segments to an invalid state; Perform cross-source logical consistency checks on the remaining valid records to exclude contradictory combinations such as decreased body fat percentage but significant weight gain, and generate a multi-source fusion dataset, wherein the multi-source fusion dataset contains only logically consistent daily records.

4. The method according to claim 1, characterized in that, The step of constructing an individualized health status trajectory based on the multi-source fusion dataset and identifying the coordinated change pattern of body composition fluctuations and physiological activity deviations over multiple consecutive days includes: Extract data sequences of N consecutive days from a multi-source fusion dataset, where N is an integer greater than or equal to three, and construct individual observation windows, wherein the individual observation windows are arranged in chronological order; The slope of the change trend of the volume component index within the observation window is calculated, and the indexes that show a monotonically rising or falling trend are marked. The slope of the change trend is determined by the cumulative direction of the difference between adjacent days. Simultaneously calculate the degree of deviation of physiological activity indicators within the same observation window, and identify indicators that are below or above the historical average for multiple consecutive days, wherein the historical average is calculated based on the user's data over the past thirty days; The trend of changes in body composition indicators is cross-compared with the deviation direction of physiological activity indicators to screen out windows that simultaneously meet specific combination conditions, wherein the specific combination conditions are defined as an increase in body fat percentage and a continuous increase in resting heart rate. The selected windows are marked as potential risk evolution segments, and their corresponding collaborative change patterns are extracted, wherein the collaborative change patterns are characterized by the combination of indicators and their change direction.

5. The method according to claim 1, characterized in that, The step of matching the collaborative change pattern with preset health risk triggering rules to determine whether there is a dynamic evolution path that meets the risk conditions includes: The indicator combinations in the collaborative change mode are compared one by one with the rule entries in the health risk trigger rule base to find matching rule identifiers, wherein the rule entries are indexed by indicator pairs and their change directions. For a successfully matched rule entry, verify whether its corresponding time-series persistence condition is met, wherein the time-series persistence condition requires that the coordinated change last for no less than a preset number of days; If the timing continuity condition is met, the risk assessment flag of the rule entry is activated, whereby the risk assessment flag indicates that there is a valid risk path; Summarize all activated risk assessment flags to generate a risk path set, wherein the risk path set contains at least one valid risk path; If the set of risk paths is not empty, then it is determined that there exists a dynamic evolution path that satisfies the risk conditions.

6. The method according to claim 3, characterized in that, The step of performing continuity detection on the heart rate sequence in physiological activity data and removing constant value segments caused by device detachment includes: The heart rate sequence is divided into continuous segments in chronological order, and the absolute value of the difference between adjacent sampling points within each segment is calculated, where the absolute value of the difference reflects the amplitude of heart rate fluctuations. For each segment, the percentage of sampling points whose absolute difference is less than a first preset value is counted, where the first preset value represents the threshold value for no physiological fluctuation. If the proportion of a certain segment exceeds the second preset value, the segment is determined to be a suspected constant segment, wherein the second preset value defines the proportion threshold of a high probability constant state; for suspected constant segments, it is further checked whether their duration exceeds the third preset value, wherein the third preset value corresponds to the lower limit of the typical duration of device detachment; If the duration exceeds the third preset value, the segment is marked as a constant value segment and removed from the original sequence.

7. The method according to claim 4, characterized in that, The slope of the change trend of the calculated volume component index within the observation window, marking indices that show a monotonically increasing or decreasing trend, includes: For the daily values ​​of body component indicators within the observation window, the difference between two adjacent days is calculated to form a difference sequence, wherein the difference sequence represents the diurnal variation. For each element in the difference sequence, determine whether its sign is positive, negative, or zero, and generate a sign sequence, wherein the sign sequence reflects the daily change direction; The length of a continuous segment of non-zero symbols in a statistical symbol sequence, wherein the continuous segment of non-zero symbols is a subsequence in which the symbols are identical and appear consecutively; If the length of the longest consecutive segment in the same direction accounts for more than the proportion of the total number of days in the observation window, the indicator is determined to show a monotonic trend, wherein the fourth preset value defines the proportion threshold for the significance of the trend. The index is marked as monotonically increasing or monotonically decreasing based on the sign of the longest consecutive segment in the same direction.

8. The method according to claim 5, characterized in that, The verification of whether the corresponding time-series persistence condition is satisfied for each successfully matched rule entry includes: Extract the start and end dates of the collaborative change patterns corresponding to the matching rule entries in the observation window, and calculate the duration in days, where the duration in days is the calendar interval between the end date and the start date plus one. The calculated duration is compared with the minimum duration defined in the rule entry, where the minimum duration is an inherent parameter of the rule; If the duration is greater than or equal to the minimum duration, the time sequence duration condition is determined to be met, and the determination result triggers the rule. If the duration is less than the minimum duration, the match is determined to be a momentary fluctuation and does not constitute a valid risk path. The momentary fluctuation is excluded from the risk path set. Repeat the above verification process for all matching rules to generate a final list of activation rules, wherein the list of activation rules contains only the entries that have passed the time-series verification.

9. The method according to claim 6, characterized in that, The further inspection of suspected constant segments to determine whether their duration exceeds a third preset value includes: Obtain the start and end timestamps of the suspected constant segment, and calculate the time difference between them, wherein the time difference is in minutes; The time difference is compared with a third preset value, wherein the third preset value is a fixed time length; If the time difference is greater than the third preset value, the segment is confirmed as invalid data caused by device detachment; if the time difference is less than or equal to the third preset value, the segment is retained as valid physiological data, wherein the retention operation maintains the integrity of the original sequence. The confirmation result is written to the data quality tag field for use in subsequent steps, where the data quality tag field indicates the validity status of the data for that time period.

10. A health risk early warning system based on body fat scale and fitness tracker data, characterized in that, The system includes: The data acquisition module is used to simultaneously acquire body composition data measured by the user on the body fat scale and physiological activity data recorded by the fitness tracker. The body composition data includes weight, body fat percentage, skeletal muscle mass and basal metabolic rate, and the physiological activity data includes heart rate sequence, cumulative step count and sleep duration. The data fusion module is used to perform timestamp alignment and outlier removal on the body composition data and physiological activity data to generate a multi-source fusion dataset, wherein the multi-source fusion dataset contains valid data records aligned to daily granularity and verified for consistency. The change recognition module is used to construct an individualized health status trajectory based on the multi-source fusion dataset and identify the coordinated change pattern of body composition fluctuation and physiological activity deviation over multiple consecutive days. The coordinated change pattern includes a combination of features such as increased body fat percentage accompanied by a sustained increase in resting heart rate or a significant reduction in sleep duration. The matching and judgment module is used to match and judge the collaborative change pattern with the preset health risk triggering rules to determine whether there is a dynamic evolution path that meets the risk conditions. The health risk triggering rules define the joint deviation threshold relationship between body composition and physiological activity indicators in the time dimension. The early warning processing module is used to generate a health risk early warning identifier and output it to the user terminal when there is a dynamic evolution path that meets the risk conditions. The health risk early warning identifier is a graded early warning signal corresponding to the dynamic evolution path.

Images