Big data-based intelligent analysis method and system for resident health

Abstract

The application discloses a resident health intelligent analysis method and system based on big data. The method comprises the following steps: acquiring historical data of multiple health indexes of residents and performing time alignment and cleaning, presetting a sliding window to calculate short-term fluctuation degree, obtaining a trend baseline, and combining the short-term fluctuation degree to generate a health trajectory deviation trend. Key life events are mapped into event impact vectors, and the health trajectory deviation trend is combined into a fusion deviation trend vector. A historical state vector is constructed based on historical data, short-term fluctuation degree, health trajectory deviation trend and event impact vector, a forgetting strategy is used to update the state vector, an individual health risk is output by inputting a time series prediction model, and hierarchical early warning is performed, thereby effectively improving the accuracy and timeliness of health risk prediction. Through the technologies of event impact, state vector, time series model and hierarchical early warning, the application significantly improves the accuracy and timeliness of health risk prediction, and realizes efficient closed-loop management of monitoring and intervention.

Description

Technical Field

[0001] This invention belongs to the field of big data, specifically relating to a method and system for intelligent analysis of residents' health based on big data. Background Technology

[0002] Current resident health risk assessments mostly employ cross-sectional methods based on single physical examinations or a limited number of recent measurements. These methods can only interpret physiological states at a specific point in time and struggle to systematically utilize the evolutionary trajectories and periodic, seasonal, and individualized rhythms of indicators such as blood glucose, blood pressure, and body mass index over longer timescales. Such methods neglect time-series correlations and the inertia effect of health status, easily mistaking short-term noise for anomalies or slow deterioration for normal fluctuations, resulting in insufficient sensitivity and stability in risk identification. With the increasing prevalence of home monitoring and wearable devices, the problem of data density increasing while modeling remains at the cross-sectional level has become even more pronounced.

[0003] On the other hand, key events in real life, such as diagnosis and follow-up visits, surgery or hospitalization, medication initiation or adjustment, changes in exercise and dietary habits, and major stress, often cause sudden shifts or sustained deviations in health trajectories. Existing protocols generally lack mechanisms to jointly model these discrete events and continuous physiological indicators within a unified framework, making it difficult to express baseline shifts, fluctuation amplification, and recovery processes before and after the event. The direct consequences of this include delayed early warning, inaccurate individual risk stratification, and decreased efficiency in the allocation of follow-up and intervention resources. Summary of the Invention

[0004] To address the technical problems of existing health risk assessment schemes lacking a mechanism for jointly modeling discrete events and continuous physiological indicators within a unified framework, which makes it difficult to accurately express the impact of events on health trajectories, resulting in delayed early warnings, inaccurate individual risk stratification, and reduced efficiency in follow-up and intervention resource allocation, this invention provides the following technical solution.

[0005] In the first aspect, the big data-based intelligent analysis method for residents' health includes: acquiring historical data of multiple health indicators of residents and performing time alignment and cleaning; pre-setting a sliding window for the historical data and calculating the short-term volatility of each health indicator within the sliding window to characterize the short-term stability of each indicator and obtain the trend baseline of the historical data of each health indicator; generating a health trajectory deviation trend by combining the deviation of the resident's current health indicator from the trend baseline with the short-term volatility; mapping key life events into event impact vectors according to event type and forming a fusion deviation trend vector with multiple health trajectory deviation trends; constructing a historical state vector based on historical data, short-term volatility, health trajectory deviation trends, and event impact vectors; updating the state vector by attenuating the historical state vector using a forgetting strategy and superimposing it with the fusion deviation trend vector; and inputting the sequence of updated state vectors into a time-series prediction model to output individual health risks and perform stratified early warning.

[0006] Preferably, the time alignment includes time zone standardization, sampling frequency unification, resampling to a unified time grid, outlier quantile pruning, and window interpolation of missing segments, and adds confidence markers to the interpolated samples.

[0007] Preferably, the short-term volatility includes:

[0008] The calculation is performed within a sliding window centered on the target time. The window width is adaptively set according to the indicator sampling frequency or business rules. Based on each preset scale window, intraday fluctuations and weekly fluctuations are obtained.

[0009] Preferably, the step of obtaining the deviation trend of the health trajectory includes:

[0010] The deviation of the health trajectory from the trend is determined by the difference between the current health indicators and the trend baseline, combined with the amplification factor of short-term volatility.

[0011] Preferably, obtaining the event impact vector includes:

[0012] A weighted mapping is performed based on event type, severity, and temporal proximity. The events include diagnosis or follow-up visits, surgery or hospitalization, medication initiation or adjustment, and exercise or dietary habits.

[0013] Preferably, the updated state vector includes:

[0014] When updating the historical state vector, a forgetting factor is applied to the historical state to reflect the decay effect, and it is superimposed with the composite amount of the fused deviation trend vector and the event impact vector.

[0015] Preferably, the time-series prediction model is a time-series deep learning model of a long short-term memory network, and cross-validation and early stopping strategies are used to suppress overfitting, while weighted loss or resampling is used for correction.

[0016] The beneficial effects of this invention are:

[0017] This invention addresses inconsistencies in timestamps and sequence breaks through time alignment and cleaning, creating usable continuous time series; it overcomes the masking effect of mean on anomalies by highlighting abnormal fluctuation signals through short-term volatility; it utilizes long-term trajectories and seasonal patterns to reveal slowly deteriorating trends; it combines deviation and volatility to reduce false positives and false negatives in thresholding; it incorporates key events such as diagnosis, surgery, and medication into unified quantification through event impact analysis; it uses state vectors and forgetting mechanisms to balance historical memory and new information responses, suppressing noise interference; it employs a time series model to model cross-time dependencies and nonlinearities, improving the accuracy of risk prediction; and it integrates hierarchical early warning and linkage mechanisms to transform risk probabilities into actionable actions, creating a closed loop from monitoring to intervention. Attached Figure Description

[0018] Figure 1 This is a flowchart of steps S1-S4 in the big data-based intelligent analysis method for resident health in this embodiment of the invention.

[0019] Figure 2 This is a structural block diagram of the resident health intelligent analysis system based on big data, according to an embodiment of the present invention. Detailed Implementation

[0020] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments.

[0021] Reference Figure 1 The big data-based intelligent analysis method for residents' health includes steps S1-S4, as detailed below:

[0022] S1: Obtain historical data on multiple health indicators of residents and perform time alignment and cleaning.

[0023] We collect raw data from multiple sources, including electronic medical records, physical examination records, wearable devices, and questionnaires, which includes multidimensional continuous indicators and discrete event sets. We aggregate and deduplicate data by user identifier and timestamp, and unify the coding, units, and field naming.

[0024] All timestamps are unified to the base time zone while retaining the original time zone field for auditing purposes; the recorded times are sorted, verified, and concurrent conflicts at the same second are eliminated to form a strictly monotonic timeline.

[0025] A unified time grid is set based on the sampling frequency of each data source. Linear interpolation or forward filling resampling strategies are applied to multidimensional continuous indicators to align multiple variables on the same time axis. Median filtering is used to smooth sudden high-frequency noise.

[0026] For obvious measurement errors and outliers, quantile pruning is performed and anomaly markers are recorded; window imputation is performed on short missing segments, and model imputation is performed on long missing segments with the confidence weight of the segment reduced, ensuring sequence integrity while controlling error propagation.

[0027] The discrete event set is dictionary-based, unifying event types, severity, and direction of impact. A configurable mapping table from events to impact vectors is established and version-managed, supporting quantitative expression and traceability during subsequent fusion.

[0028] Desensitize personally identifiable information and implement least privilege access control, enable audit logs and data lineage records to ensure that subsequent training and inference processes are compliant and traceable.

[0029] S2: Preset a sliding window for historical data and calculate the short-term volatility of each health indicator within the sliding window to characterize the short-term stability of each indicator and obtain the trend baseline of historical data for each health indicator; combine the deviation of residents' current health indicators from the trend baseline with the short-term volatility to generate a health trajectory deviation trend.

[0030] A sliding window is preset for historical data, and the short-term volatility of each health indicator within the sliding window is calculated to characterize the short-term stability of each indicator.

[0031] With target time Build a center with a width of sliding window Calculate the index within the window. Short-term volatility is used to measure stability and to detect abnormal volatility signals.

[0032] Specifically, short-term volatility satisfies the following relationship:

[0033] ;

[0034] In the formula, Indicates the first Each health indicator at the target time Short-term volatility Represents a sliding window Number of data points within, express, Indicates the first Health indicators at a given time point The measured value, Indicates the first Several health indicators in the sliding window The average value within, Indicates a point in time In the sliding window Inside.

[0035] Trend baseline estimation includes: performing robust regression or smoothing (linear, polynomial, spline or kernel methods) on each indicator based on the complete historical series to obtain the long-term trend, and combining seasonal decomposition when necessary to separate cyclical components and reduce the impact of short-term disturbances on trend estimation.

[0036] The health trajectory deviation trend generation includes: combining the degree of deviation of residents' current health indicators from the trend baseline with short-term volatility to generate the health trajectory deviation trend.

[0037] ;

[0038] In the formula, Indicates the first Each health indicator at the target time The health trajectory deviates from the trend. Indicates the first Each health indicator at the target time The actual measured value, Indicates the first Each health indicator at the target time The trend baseline value, Indicates the magnification factor. Indicates the first Each health indicator at the target time Short-term volatility Indicates the first The average short-term volatility of a health indicator.

[0039] Multi-indicator synthesis and scale alignment include: integrating various dimensions Composition vector Distribution standardization and robust scaling are applied to indicators with different dimensions to ensure comparability across indicators; minimum weights are assigned to weak signal dimensions to prevent them from being completely submerged.

[0040] S3: Map key life events to event impact vectors based on event type, and combine them with multiple health trajectory deviation trends to form a fused deviation trend vector; construct a historical state vector based on historical data, short-term volatility, health trajectory deviation trends, and event impact vectors.

[0041] Event impact vector synthesis includes:

[0042] Key life events are mapped to event impact vectors based on event type, and then combined with multiple health trajectory deviation trends to form a fused deviation trend vector.

[0043] The event is retrieved or calculated from the mapping table based on event type, severity, and temporal proximity. When multiple events exist at the same time, they are combined into an equivalent impact based on importance and time weights to ensure that acute events have an immediate impact and chronic events have a sustained impact.

[0044] The state vector update includes: applying a forgetting factor to the historical state to reflect time decay, and then superimposing it with the current deviation potential and event impact to generate a new state representation, which is used as the input sequence for the downstream time series model.

[0045] Specifically, the state vector satisfies the following relationship:

[0046] ;

[0047] In the formula, Indicates the target time The state vector, Indicates the forgetting factor, , Indicates the previous moment The state vector, Indicates the current time The health trajectory deviates from the trend. Indicates the current time A collection of all key life events, Indicates an event The event impact vector.

[0048] In other words, in updating the resident health status vector At that time, firstly through the forgetting factor Regarding the state vector of the previous time step The decay process is applied to reflect the time decay effect; then, the deviation trend of the current health trajectory is added. This reflects the deviation of current health indicators from the trend baseline; finally, it includes the event impact vectors of all key life events at the current moment. The sum of these factors reflects the impact of events on health status. This comprehensive process ensures that the health status vector dynamically reflects residents' health status while taking into account both historical information and newly occurring events.

[0049] Smoothing and hysteresis control includes: adjusting the resident health state vector to reduce drastic fluctuations caused by short-term noise. Alternatively, it can use an index-weighted moving average based on the risk projection; it responds immediately to rising signals and sets a hysteresis threshold for falling signals to avoid frequent repetitions.

[0050] Interpretable records include: in the generation Simultaneously, it outputs summaries of key indicators, critical events, and contributions, forming a decision-making chain of "indicator → deviation → event → status" for auditing and intervention review.

[0051] S4: The historical state vector is decayed by using a forgetting strategy and then superimposed with the fused deviation trend vector to update the state vector; the sequence of updated state vectors is input into the time series prediction model according to the time series to output individual health risk and perform stratified early warning.

[0052] Training data preparation and partitioning include: Samples are constructed using the input sequence and risk labels within a future window as supervision signals. Training, validation, and test sets are partitioned according to the user dimension to avoid information leakage, and weighted losses or resampling are applied to imbalanced data to ensure the balance and effectiveness of model training.

[0053] Choose LSTM, GRU, or an equivalent time series model as the base model, and configure hyperparameters such as the number of layers, hidden units, learning rate, and dropout rate; use grid or Bayesian optimization to tune the parameters on the validation set, and enable early stopping and weight decay to suppress overfitting.

[0054] Sliding window sampling and mini-batch training are employed to monitor training and validation loss, AUC, recall, and calibration accuracy in real time. When calibration deviation is too large, temperature scaling or equivalent techniques are introduced for post-processing to ensure the model's accuracy and reliability.

[0055] Maintaining the most recent window in online services Triggered or timed refresh of risk output; ensure end-to-end latency meets business thresholds, and downsample non-critical indicators to ensure real-time performance when resources are limited.

[0056] Risk stratification and threshold setting include: setting multi-level thresholds based on the performance curve of the validation set to map risk probabilities to stratification levels; applying differentiated thresholds and alarm routes to highly sensitive populations or specific disease subgroups to achieve refined management.

[0057] The early warning linkage and intervention strategy includes: linking stratified results with follow-up, retesting, medication review or lifestyle intervention; recording the execution and feedback of results according to the strategy to form closed-loop data to support subsequent model and threshold retraining.

[0058] Perform incremental training periodically or on a trigger-based basis, compare the performance of the new and old models in offline and low-traffic gray-scale tests, and complete the version switch after meeting the preset threshold; retain version files of the model, dictionary and threshold for backtracking.

[0059] Continuously monitor service availability, latency, failure rate, and data drift; automatically alert to abnormal fluctuations and switch to a security degradation strategy; retain audit logs and explanatory summaries throughout the entire process to meet compliance requirements.

[0060] Establish fault tolerance mechanisms and fallback models for data gaps, model unavailability, or external dependency failures to ensure that degraded but usable risk scores can still be provided during anomalies, and automatically switch back and fill in missing records after service recovery.

[0061] It should be noted that those skilled in the art can make various modifications and improvements without departing from the inventive concept, and these all fall within the scope of protection of this invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0062] This invention also provides a smart resident health analysis system based on big data. For example... Figure 2 As shown, the system includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the big data-based intelligent analysis method for resident health according to the first aspect of the present invention.

[0063] The system also includes other components well known to those skilled in the art, such as communication buses and communication interfaces, the settings and functions of which are known in the art and will not be described in detail here.

[0064] In the description of this specification, "multiple" or "several" means at least two, such as two, three or more, unless otherwise explicitly specified.

[0065] While this specification has shown and described numerous embodiments of the invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in the practice of this invention.

Claims

1. A method for intelligent analysis of residents' health based on big data, characterized in that, include: Obtain historical data on multiple health indicators of residents and perform time alignment and cleaning; A sliding window is preset for historical data, and the short-term volatility of each health indicator within the sliding window is calculated to characterize the short-term stability of each indicator and obtain the trend baseline of the historical data of each health indicator. The deviation of residents' current health indicators from the trend baseline is combined with short-term volatility to generate a health trajectory deviation trend. Key life events are mapped to event impact vectors based on event type, and then combined with multiple health trajectory deviation trends to form a fused deviation trend vector; a historical state vector is constructed based on historical data, short-term volatility, health trajectory deviation trends, and event impact vectors. A forgetting strategy is used to decay the historical state vector, which is then superimposed with the fused deviation trend vector to update the state vector. The sequence of updated state vectors is input into the time series prediction model to output individual health risks and provide stratified early warning.

2. The method for intelligent analysis of resident health based on big data according to claim 1, characterized in that, The time alignment process includes time zone standardization, sampling frequency unification, resampling to a unified time grid, outlier quantile pruning, and window interpolation of missing segments, with confidence markers added to the interpolated samples.

3. The method for intelligent analysis of resident health based on big data according to claim 1, characterized in that, The short-term volatility includes: The calculation is performed within a sliding window centered on the target time. The window width is adaptively set according to the indicator sampling frequency or business rules. Based on each preset scale window, intraday fluctuations and weekly fluctuations are obtained.

4. The method for intelligent analysis of resident health based on big data according to claim 1, characterized in that, The steps for obtaining the deviation trend of the health trajectory include: The deviation of the health trajectory from the trend is determined by the difference between the current health indicators and the trend baseline, combined with the amplification factor of short-term volatility.

5. The intelligent analysis method for resident health based on big data according to claim 1, characterized in that, Obtaining the event impact vector includes: A weighted mapping is performed based on event type, severity, and temporal proximity. The events include diagnosis or follow-up visits, surgery or hospitalization, medication initiation or adjustment, and exercise or dietary habits.

6. The method for intelligent analysis of resident health based on big data according to claim 1, characterized in that, The updated state vector includes: When updating the historical state vector, a forgetting factor is applied to the historical state to reflect the decay effect, and it is superimposed on the fused deviation trend vector.

7. The intelligent analysis method for resident health based on big data according to claim 1, characterized in that, The time-series prediction model is a time-series deep learning model of a long short-term memory network, and it employs cross-validation and early stopping strategies to suppress overfitting, and uses weighted loss or resampling for correction.

8. A resident health intelligent analysis system based on big data, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions that, when executed by the processor, implement the big data-based intelligent analysis method for resident health according to any one of claims 1-7.

Images