Health data analysis risk early warning system and method for the elderly

Abstract

The application discloses an old person health data analysis risk early warning system and method, relates to the technical field of health data analysis, and comprises the following steps: obtaining a first monitoring frequency based on multi-scale smoothing and minimum redundant point searching; obtaining a first fusion window scale based on redundant reduction determination; applying a plurality of data analysis control sets to health data analysis risk early warning respectively, and calculating a first evaluation value of each data analysis control set; constructing a data analysis control set display model based on the first evaluation value, the monitoring frequency and the fusion window scale, and selecting an optimal data analysis control set from the first surface according to analysis requirements, which is applied to health data analysis risk early warning. The application solves the problem that in the complex physiological signal change scene of the old people, the trade-off between data integrity and real-time performance is lacking, and it is difficult to effectively support accurate and timely health early warning.

Description

Technical Field

[0001] This invention relates to the field of health data analysis technology, and more specifically, to a risk warning system and method for health data analysis of the elderly. Background Technology

[0002] Existing health monitoring systems typically rely on multi-source data acquisition at fixed frequencies and use sliding windows to fuse physiological data over short time periods to determine the fluctuation trends of an individual's health status. These systems are mostly used in the general adult population and focus on the continuous fluctuation characteristics of routine physiological parameters such as heart rate, blood pressure, and activity level. With a moderate frequency of change, they can maintain the real-time nature of data processing while avoiding the waste of resources caused by over-computation.

[0003] However, for the elderly, their physiological data exhibit low fluctuations and slow changes most of the time, but can change drastically in an instant during sudden abnormal events (such as falls, sudden changes in heart rate, and apnea). Due to their suddenness and short intervention window, health monitoring systems for the elderly need to have higher sensitivity and robustness in terms of response speed and status recognition accuracy compared to the general population, especially to avoid missed or false alarms caused by data delays or fuzzy fusion.

[0004] Monitoring frequency refers to the time interval at which the system collects physiological signals or behavioral data, such as once per second or once every 10 minutes. A higher frequency can capture data changes more promptly, which is beneficial for anomaly early warning, but it will lead to increased device power consumption and data processing pressure. While a low frequency is energy-saving and has less computational pressure, it is easy to miss rapidly developing risk events, and is especially unsuitable for handling sudden abnormalities in the elderly.

[0005] The fusion window scale refers to the time period range selected by the system when performing feature integration analysis on multi-source data, and is often used in sliding window algorithms. For example, data from the past 5 minutes, 30 minutes, or 1 hour can be fused and modeled. A large window scale helps smooth short-term fluctuations, improves stability, and is suitable for analyzing long-term trends, but has a higher response delay; a small window scale is beneficial for rapid response to sudden anomalies and improves early warning sensitivity, but is susceptible to occasional noise interference and may produce false alarms.

[0006] The aforementioned publicly disclosed technical solutions have at least the following technical problems: In health risk early warning for the elderly, the monitoring frequency and the fusion window scale are mutually restrictive and synergistic.

[0007] On the one hand, while excessively high monitoring frequency can help improve the sensitivity of anomaly response, it may also lead to a large amount of redundant data, resulting in increased computational burden and affecting the system's real-time performance and stability.

[0008] On the other hand, a fusion window that is too small may not be able to effectively smooth out fluctuations in physiological signals, leading to an increased false alarm rate, while a window that is too large may mask sudden anomalies and reduce the sensitivity of risk identification.

[0009] Existing technologies lack a mechanism for dynamic collaborative optimization between monitoring frequency and fusion window scale, and cannot suppress data redundancy while ensuring sensitivity and stability. Especially in scenarios where physiological signals of the elderly change in complexity, there is a lack of balance between data integrity and real-time performance, making it difficult to effectively support accurate and timely health warnings.

[0010] To address the above problems, this invention proposes a solution. Summary of the Invention

[0011] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a risk warning system and method for health data analysis of the elderly. By analyzing and selecting combinations of different monitoring frequencies and fusion window scales, a combination that satisfies both real-time performance and data integrity is obtained and applied to the risk warning analysis process. This addresses the problem that in scenarios with complex physiological signal changes in the elderly, there is a lack of balance between data integrity and real-time performance, making it difficult to effectively support accurate and timely health warnings.

[0012] To achieve the above objectives, the present invention provides the following technical solution:

[0013] A health data analysis risk warning system and method for the elderly includes the following steps: obtaining a first monitoring frequency based on multi-scale smoothing and minimum redundancy point search; obtaining a first fusion window scale based on redundancy reduction; defining an array combination of a monitoring frequency and a fusion window scale as a data analysis control set; alternately advancing the first monitoring frequency and the first fusion window scale towards information gain and information reduction directions by a preset step size, iterating several times, and randomly combining the obtained data to generate several sets of data analysis control sets; applying several sets of data analysis control sets to health data analysis risk warning, and collecting response capability characteristics, data redundancy characteristics, and warning quality characteristics, and weightedly calculating the first evaluation value of each set of data analysis control sets; constructing a data analysis control set display model based on the first evaluation value, monitoring frequency, and fusion window scale; mapping each set of data analysis control sets and the corresponding first evaluation value to the data analysis control set display model, and fitting a first surface; selecting the best data analysis control set from the first surface according to the analysis requirements, and applying it to health data analysis risk warning.

[0014] In a preferred embodiment, obtaining the first monitoring frequency based on multi-scale smoothing and minimum redundancy point search specifically involves: time-series processing of collected historical health data and defining the mean differential entropy of the time-series health data as information density; defining the mean autocorrelation coefficient of the time-series health data as data redundancy; obtaining information density and data redundancy at different monitoring frequencies; and selecting the monitoring frequency with the maximum information density and minimum redundancy per unit amount of health data as the initial monitoring frequency.

[0015] In a preferred embodiment, obtaining the first fusion window scale based on redundancy reduction specifically involves: defining a fusion window scale, i.e., the fusion window scale is equal to the time-domain window length; calculating the mutual information within each time-domain window length under different time-domain window lengths; gradually increasing the time-domain window length and calculating the mutual information gain under each window length; calculating the rate of change of information increment between adjacent window scales to form the marginal benefit curve of information gain; and using the inflection point of the marginal benefit curve of information gain as the first fusion window scale.

[0016] In a preferred embodiment, the step of alternately advancing the first monitoring frequency and the first fusion window scale in the directions of information gain and information attenuation by a preset step size is as follows: the advancement in the direction of information gain is to start from the first monitoring frequency and the first fusion window scale, increase the monitoring frequency by a preset step size each time, and record the corresponding monitoring frequency and fusion window scale after each increase; the advancement in the direction of information attenuation is to start from the first monitoring frequency and the first fusion window scale, decrease the monitoring frequency by a preset step size each time, and record the corresponding monitoring frequency and fusion window scale.

[0017] In a preferred embodiment, the iteration is repeated several times, and the obtained data is randomly combined to generate several sets of data analysis control sets. Specifically, several monitoring frequencies and several fusion window scales are randomly combined to obtain several sets of data analysis control sets, each of which contains a monitoring frequency and a fusion window scale.

[0018] In a preferred embodiment, the specific method for obtaining the response capability characteristics is as follows: collecting the actual occurrence time and corresponding early warning issuance time of all abnormal events of the target monitoring object within the monitoring period; calculating the response delay of each abnormal event to obtain the average response delay; and calculating the response capability characteristics based on the ratio of the average response delay to the monitoring frequency set by the system, combined with the monitoring frequency set by the system.

[0019] In a preferred embodiment, the specific method for obtaining the data redundancy is as follows: acquiring time-series health data of the target monitoring object within a certain monitoring period, calculating the autocorrelation coefficient of the data sequence, calculating the autocorrelation coefficients corresponding to all lag orders from 1 to the preset maximum lag value, and averaging the autocorrelation coefficients of all lag orders to obtain the data redundancy as a data redundancy characteristic value.

[0020] In a preferred embodiment, the method for obtaining the warning quality feature is as follows: for each time window, calculate the recall rate of the risk warning; calculate the difference between adjacent recall rates and extract the symbol change sequence; count the number of alternations of positive and negative symbols from the symbol change sequence and use it as the warning quality feature value.

[0021] In a preferred embodiment, the step of constructing a data analysis control set display model based on the first evaluation value, monitoring frequency, and fusion window scale, and mapping each data analysis control set and its corresponding first evaluation value to the data analysis control set display model, specifically involves: constructing a curved empty display model as the data analysis control set display model, where the first evaluation value is the vertical axis, and the monitoring frequency and fusion window scale are the two bottom coordinates of the three-dimensional coordinate system, with arbitrary positions; and mapping the monitoring frequency, fusion window scale, and corresponding first evaluation value of each data analysis control set to a data point in the data analysis control set display model.

[0022] A risk warning system for health data analysis of the elderly, characterized by comprising a monitoring frequency initialization module, a fusion window scale initialization module, a group data analysis control set generation module, a control evaluation module, and a control set optimization module; the monitoring frequency initialization module is used to obtain a first monitoring frequency based on multi-scale smoothing and minimum redundancy point search; the fusion window scale initialization module is used to obtain a first fusion window scale based on redundancy reduction; the group data analysis control set generation module is used to alternately advance the first monitoring frequency and the first fusion window scale in the directions of information gain and information reduction by a preset step size, iterate several times, and randomly combine the obtained data to generate a control set. The system comprises: a control set for data analysis; a control evaluation module, which applies several control sets to health data analysis risk warning and collects response capability characteristics, data redundancy characteristics, and warning quality characteristics, and calculates a weighted first evaluation value for each control set; and a control set optimization module, which constructs a data analysis control set display model based on the first evaluation value, monitoring frequency, and fusion window scale, maps each control set and its corresponding first evaluation value to the data analysis control set display model, fits a first surface, and selects the optimal control set from the first surface according to the analysis requirements for application in health data analysis risk warning.

[0023] The technical effects and advantages of the elderly health data analysis risk early warning system and method of this invention are as follows:

[0024] This invention provides a risk warning method for elderly health data analysis. It dynamically and adaptively determines the optimal combination of monitoring frequency and fusion window scale based on historical health data characteristics, achieving synergistic optimization of data acquisition and fusion strategies. Compared to traditional fixed-parameter monitoring systems, this method fully considers the dynamic characteristics of "stationary + abrupt changes" in the physiological characteristics of the elderly. By introducing information density, data redundancy, and mutual information gain marginal benefit analysis, it effectively suppresses redundant data and improves response timeliness. Simultaneously, by combining the sign change frequency analysis of warning recall rate, it enhances the ability to measure the stability of warning quality. By establishing a three-dimensional display model and fitting an evaluation surface, it enables intuitive comparison and optimal selection of the performance of multiple strategies, thereby improving the system's sensitivity, accuracy, and resource efficiency in high-risk scenarios involving the elderly, demonstrating good practicality and promotional value. Attached Figure Description

[0025] Figure 1 This is a flowchart illustrating the risk warning method for health data analysis of the elderly according to the present invention.

[0026] Figure 2 This is a schematic diagram of the elderly health data analysis and risk warning system of the present invention. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0028] Example 1, Figure 1 The present invention provides a method for risk early warning through health data analysis of the elderly, comprising the following steps:

[0029] S1, the first monitoring frequency is obtained based on multi-scale smoothing and search for minimum redundancy points;

[0030] In this embodiment, the first monitoring frequency is obtained based on multi-scale smoothing and a minimum redundancy point search, specifically as follows:

[0031] The collected historical health data is time-series analyzed, and the mean differential entropy of the time-series health data is defined as information density.

[0032] The mean autocorrelation coefficient of time-series health data is defined as data redundancy.

[0033] Information density and data redundancy are obtained at different monitoring frequencies;

[0034] The monitoring frequency with the highest information density and lowest redundancy per unit of health data was selected as the initial monitoring frequency.

[0035] It should be noted that information density measures the amount of effective information contained in health data per unit time. In this embodiment, it is calculated from the mean differential entropy of time-series health data. Differential entropy reflects the degree of uncertainty in the data; a higher value indicates richer information and more complex signal fluctuations. Since the physiological signals of the elderly typically change slowly but are prone to sudden changes, information density can serve as an important basis for detecting potential abnormalities. The information density is described using... Calculate, where, For windows; For window w, i represents the number of health data points in time; i is the index used to traverse each health data point within the window. The variance is the moving variance calculated based on the local time series, reflecting the data volatility at time i. The differential entropy expression for a Gaussian distribution. This is the normalization coefficient.

[0036] It should be noted that data redundancy measures the degree of repetition between adjacent sampling points in continuous health data. In this embodiment, it is calculated from the mean autocorrelation coefficient of the health data sequence. A stronger autocorrelation indicates smaller changes in the data and higher redundancy. Excessive data redundancy not only increases the system's processing burden but may also mask subtle but important abrupt changes, reducing the sensitivity and response efficiency of the early warning system. The data redundancy is measured using... Quantization, where K is the preset maximum lag step. In the formula This is the autocorrelation coefficient; ,in This represents the sample mean of the time series. This represents time-series health data, where t is a time point.

[0037] S2, the first fusion window scale is obtained based on redundancy reduction;

[0038] In this embodiment, obtaining the first fusion window scale based on redundancy reduction specifically involves:

[0039] Define the fusion window scale, which is equal to the temporal window length;

[0040] Calculate mutual information within each time-domain window length under different time-domain window lengths;

[0041] As the time-domain window length is gradually increased, the mutual information gain under each window length is statistically analyzed.

[0042] It also calculates the rate of change of information increment between adjacent window scales to form the marginal benefit curve of information gain;

[0043] The inflection point of the marginal benefit curve of information gain is used as the first fusion window metric.

[0044] It should be noted that the inflection point is the critical point in the marginal benefit curve of mutual information gain where information gain shifts from "rapid growth" to "slow increase or even stabilization". The window size corresponding to the inflection point represents the position where the information gain benefits begin to decrease. Expanding the window further may slightly increase information, but it will reduce real-time performance, make data fusion more resource-intensive, and may smooth out abnormal signals. Therefore, this inflection point is regarded as the balance point between information benefits and latency costs, and is used to set the first fusion window size.

[0045] It should be noted that the time-domain window length refers to the length of the time segment used for fusion analysis, measured in seconds or the number of sampling points. It indicates that the data collected within this time range will be regarded as a fusion analysis unit. For example, if the system samples at 1Hz, a window length of 30 means that the data from the past 30 seconds will be used for fusion.

[0046] In health monitoring of the elderly, the typical time window length can be set from 10s to 600s (adjusted according to the target event, such as a fall, sudden change in heart rate, sleep stage, etc.).

[0047] It should be noted that the mutual information within each time-domain window refers to the mutual information value calculated between different health data (such as heart rate, blood pressure, body movement, etc.) within a specific window length. It is used to measure the degree of information dependence between variables within that time period. The higher the mutual information obtained, the greater the fusion potential within that window length. The specific method for calculating mutual information is to use the Shannon entropy formula after discretization, that is:

[0048] In the formula For mutual information between variables X and Y in health data, Let X and Y be the joint probability distribution. and It is distributed at the edge.

[0049] It should be noted that mutual information gain is the increment of mutual information between two adjacent fusion window scales, used to evaluate whether increasing the window scale brings a significant information gain, i.e.: In the formula This represents the mutual information value at the k-th fusion window scale.

[0050] S3 defines a data analysis control set as an array combination of a monitoring frequency and a fusion window scale.

[0051] It should be noted that the "data analysis control set" refers to a binary combination consisting of a monitoring frequency value and a fusion window scale value, which is essentially a parameter setting for the data acquisition and processing strategy.

[0052] S4, the first monitoring frequency and the first fusion window scale are alternately advanced by a preset step size in the direction of information gain and information reduction, respectively, and iterated several times. The obtained data are randomly combined to generate several sets of data analysis control sets.

[0053] In this embodiment, the first monitoring frequency and the first fusion window scale are alternately advanced by a preset step size in the direction of information gain and information attenuation, respectively, specifically as follows:

[0054] The information gain direction advancement is to start with the first monitoring frequency and the first fusion window scale, increase the monitoring frequency, and increase it by a preset step size each time. After each increase, the corresponding monitoring frequency and fusion window scale are recorded.

[0055] The information attenuation direction is advanced by starting from the first monitoring frequency and the first fusion window scale, reducing the monitoring frequency by a preset step size each time, and recording the corresponding monitoring frequency and fusion window scale.

[0056] The iterations are repeated several times, and the obtained data are randomly combined to generate several sets of data analysis control sets, specifically:

[0057] Several monitoring frequencies and several fusion window scales are randomly combined to obtain several sets of data analysis control sets. Each data analysis control set contains one monitoring frequency and one fusion window scale.

[0058] It should be noted that the purpose of initializing the monitoring frequency and fusion window scale to obtain the first monitoring frequency and the first fusion window scale is to establish a reasonable time granularity and observation framework for subsequent data monitoring and feature fusion. This ensures that the collected data has sufficient information density to reflect the changing trend of health status, while also being controlled within an acceptable range of redundancy. It avoids increased computational burden or information duplication and redundancy caused by excessively high frequency or excessively wide window, thereby laying the foundation for subsequent data compression, feature extraction and risk identification, and improving the sensitivity and resource utilization efficiency of the monitoring system.

[0059] The initial monitoring frequency and fusion window size settings directly affect the timeliness and discriminative ability of time-series health data. Setting the initial frequency too high can lead to increased data redundancy and a heavier computational load on the system; setting it too low may cause key state transition information to be missed, affecting the timeliness of fault warnings. Similarly, an excessively large fusion window size will mask local anomalies, while an excessively small one may result in drastic feature fluctuations and poor stability. Therefore, properly initializing these parameters not only helps in building a monitoring mechanism with appropriate information density but also determines the accuracy and reliability of subsequent redundancy calculations, feature extraction, and risk modeling.

[0060] S5 applies several sets of data analysis control sets to health data analysis risk warning, and collects response capability characteristics, data redundancy characteristics, and warning quality characteristics, and calculates the first evaluation value of each set of data analysis control sets by weight.

[0061] In this embodiment, the responsiveness feature is primarily used to measure the system's real-time identification and response capabilities to potential health risks, corresponding to the referenced "real-time responsiveness" dimension. Its core principle is that a higher monitoring frequency allows the system to capture potential risk signals earlier, thereby improving the immediacy of warnings. However, excessively high monitoring frequencies can also lead to data redundancy and increased noise, affecting the accuracy of warning judgments. Therefore, this feature emphasizes a balanced assessment of "timely response" and "response quality."

[0062] The specific method for obtaining the response capability characteristics is as follows:

[0063] Collect the actual occurrence time and corresponding early warning issuance time of all abnormal events of the target monitoring object within the monitoring period;

[0064] Calculate the response delay for each abnormal event to obtain the average response delay; combine the system's set monitoring frequency with the ratio of the average response delay to the monitoring frequency to calculate the response capability characteristics.

[0065] The specific calculation formula for the response capability feature is as follows:

[0066]

[0067]

[0068] In the formula, N represents the number of abnormal events. To indicate the response latency, i is the event index. For average response delay, To monitor frequency, R is the smoothing constant, and R is the characteristic value of the response capability.

[0069] Analyzing response capability characteristics offers the following advantages in addressing the challenges of selecting control sets and balancing data integrity with real-time performance in scenarios with complex physiological signal changes in the elderly, thus hindering accurate and timely health warnings:

[0070] Response capability characteristics directly reflect the impact of monitoring frequency on health early warning of physiological signals in the elderly. When physiological signals in the elderly change slowly and occasionally exhibit sudden anomalies, a reasonable monitoring frequency ensures that key abnormal events are captured promptly, shortening the early warning response delay and improving the sensitivity and real-time performance of the early warning system. Simultaneously, it avoids the surge of redundant data and waste of computational resources caused by excessively high monitoring frequencies. By quantifying response capability characteristics, a data analysis control set with an appropriate monitoring frequency can be scientifically selected, achieving a balance between sampling density and response speed. This effectively supports the accurate and timely response of the elderly health early warning system in complex physiological signal environments, solving the problems of blindly selecting monitoring frequencies, delayed responses, or wasted resources in traditional solutions.

[0071] Data redundancy characteristics measure the degree of repetition and redundancy of information in collected health data under a specific combination of monitoring frequency and fusion window scale. High redundancy means that a large amount of similar or duplicate data is collected, which may increase the computational burden and storage pressure on the system, but does not necessarily increase the amount of effective information in early warnings. Conversely, low redundancy indicates that the data is more concise and has a higher information density, which helps to improve processing efficiency and the sensitivity of early warnings. Therefore, data redundancy characteristics are an important indicator for evaluating the balance between information effectiveness and resource utilization in monitoring strategies.

[0072] The specific method for obtaining the data redundancy is as follows:

[0073] Acquire time-series health data of the target monitoring object within a certain monitoring period, calculate the autocorrelation coefficient of the data series, and calculate the autocorrelation coefficients corresponding to all lag orders from 1 to the preset maximum lag value.

[0074] The data redundancy is obtained by averaging the autocorrelation coefficients of all lag orders and used as the characteristic value of data redundancy.

[0075] The specific method for obtaining data redundancy characteristics is the same as the data redundancy quantification described above:

[0076] And the data redundancy characteristic value is represented by P.

[0077] Data redundancy characteristics reflect the degree of similarity between time-series samples in the data. A higher value indicates greater redundancy between data points, suggesting more repetitive information and potentially reduced system processing efficiency. Conversely, a lower value indicates richer data variation and higher information content, but may also indicate higher noise levels. Quantifying data redundancy can effectively assess the redundancy of data under different monitoring frequencies and fusion window scales, providing a basis for parameter optimization in early warning systems.

[0078] In health data analysis and risk early warning, early warning quality features are used to measure the comprehensive effect of different "data analysis control sets" (i.e., combinations of monitoring frequency and fusion window scale) in actual early warning, reflecting the accuracy and stability of risk event identification. These features primarily evaluate the following aspects: precision, recall, F1 score, false alarm rate, and missed detection rate, any of which can be used as analytical parameters for early warning quality features.

[0079] The following are feasible methods for obtaining early warning quality characteristics:

[0080] For each time window, calculate the recall rate of the risk warning;

[0081] Calculate the difference in recall between adjacent sequences and extract the symbol change sequence;

[0082] The number of alternations between positive and negative signs in the sign change sequence is counted and used as a quality characteristic value for early warning.

[0083] It should be noted that the warning quality feature value represents the degree of fluctuation in the recall rate trend, and is used to quantify the stability of historical warning capabilities. Specifically, this feature value is based on the dispersion of the recall rate across multiple time windows, reflecting the consistency of the system's coverage of real alarm events under different conditions. The warning quality feature value is represented by Q.

[0084] Furthermore, the weighted calculation yields the first evaluation value for each data analysis control set, specifically as follows:

[0085] Based on the backpropagation (BP) neural network algorithm, the weighted weights of response capability features, data redundancy features, and early warning quality features are trained to obtain the calculation model for the first evaluation value. The following is a basic model framework:

[0086] In the formula The first evaluation value, , , The corresponding weights are shown below. It should be noted that training weights based on the bp neural network algorithm is an existing mature technology, which will not be elaborated here.

[0087] S6 constructs a data analysis control set display model based on the first evaluation value, monitoring frequency, and fusion window scale. It maps each data analysis control set and its corresponding first evaluation value to the data analysis control set display model and fits the first surface. According to the analysis requirements, it selects the best data analysis control set from the first surface and applies it to health data analysis risk warning.

[0088] In this embodiment, a data analysis control set display model is constructed based on the first evaluation value, monitoring frequency, and fusion window scale. Each data analysis control set and its corresponding first evaluation value are mapped to the data analysis control set display model, specifically as follows:

[0089] Construct a curved empty display model as the data analysis control set display model, where the first evaluation value is the vertical coordinate, and the two bottom coordinates of the three-dimensional coordinates of the monitoring frequency and the fusion window scale are arbitrary in position;

[0090] The monitoring frequency, fusion window scale, and corresponding first evaluation value of each data analysis control set are mapped to a data point in the data analysis control set display model.

[0091] Furthermore, the fitting of the first surface specifically involves:

[0092] The gaps in the data analysis control set display model are filled by interpolation, and the data points are fitted into a surface, which is used as the first surface.

[0093] Furthermore, based on the analysis requirements, the optimal data analysis control set is selected from the first surface and applied to health data analysis risk warning, specifically as follows:

[0094] If there are no specific analytical requirements, the data analysis control set corresponding to the peak value in the surface is selected as the optimal data analysis control set.

[0095] If there are specific analysis requirements, the optimal data analysis control set is selected based on the emphasis of the data analysis requirements on monitoring frequency and fusion window scale, with the largest control parameter within the first preset range of the surface vertical axis and the one with the greater emphasis on the analysis requirements.

[0096] Apply the best data analysis control set to health data analysis risk warning.

[0097] It should be noted that, based on the emphasis placed on monitoring frequency and fusion window scale according to data analysis needs, the optimal data analysis control set is selected when the vertical axis of the surface is within the first preset range, and the control parameter with the greater emphasis on analysis needs is the largest. An example of this is:

[0098] If the analysis requirement focuses on data integrity, then the corresponding focus is on a large fusion window scale. In this case, the data analysis control set corresponding to the point with the largest fusion window scale within the first preset range of the surface's vertical axis is found as the optimal data analysis control set.

[0099] It should be noted that the first preset range is the range of the first evaluation value, which is used to limit the selection range of the data analysis control set.

[0100] Example 2, Figure 2 The system for analyzing health data and providing risk warnings for the elderly includes a monitoring frequency initialization module, a fusion window scale initialization module, a group data analysis control set generation module, a control evaluation module, and a control set optimization module.

[0101] The monitoring frequency initialization module is used to obtain the first monitoring frequency based on multi-scale smoothing and by searching for the minimum redundancy point.

[0102] The fusion window scale initialization module is used to obtain the first fusion window scale based on redundancy reduction.

[0103] The data analysis control set generation module is used to alternately advance the first monitoring frequency and the first fusion window scale in the direction of information gain and information reduction by a preset step size, iterate several times, and randomly combine the obtained data to generate several sets of data analysis control sets.

[0104] The control evaluation module is used to apply several sets of data analysis control sets to health data analysis risk warning, and to collect response capability characteristics, data redundancy characteristics and warning quality characteristics, and to calculate the first evaluation value of each set of data analysis control sets by weighting.

[0105] The control set optimization module is used to construct a data analysis control set display model based on the first evaluation value, monitoring frequency, and fusion window scale. It maps each data analysis control set and its corresponding first evaluation value to the data analysis control set display model and fits it to the first surface. According to the analysis requirements, it selects the best data analysis control set from the first surface for application in health data analysis risk warning.

[0106] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0107] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0108] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0109] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0110] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0111] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for risk early warning through health data analysis of the elderly, characterized in that, Includes the following steps: The first monitoring frequency is obtained based on multi-scale smoothing and a minimum redundancy point search, specifically as follows: The collected historical health data is time-series analyzed, and the mean differential entropy of the time-series health data is defined as information density. The mean autocorrelation coefficient of time-series health data is defined as data redundancy. Information density and data redundancy are obtained at different monitoring frequencies; The monitoring frequency with the highest information density and lowest redundancy per unit of health data was selected as the initial monitoring frequency. The first fusion window scale is obtained based on redundancy reduction; A combination of an array of monitoring frequencies and a fusion window scale is defined as a data analysis control set; The first monitoring frequency and the first fusion window scale are alternately advanced by a preset step size in the direction of information gain and information reduction, respectively, and iterated several times. The obtained data are then randomly combined to generate several sets of data analysis control sets. Several sets of data analysis control sets are applied to health data analysis risk warning, and response capability characteristics, data redundancy characteristics and warning quality characteristics are collected. The weighted calculation yields the first evaluation value of each set of data analysis control sets. A data analysis control set display model is constructed based on the first evaluation value, monitoring frequency, and fusion window scale. Each data analysis control set and its corresponding first evaluation value are mapped to the data analysis control set display model, and a first surface is fitted. According to the analysis requirements, the best data analysis control set is selected from the first surface and applied to health data analysis risk warning. The first fusion window scale obtained based on redundancy reduction is specifically as follows: Define the fusion window scale, which is equal to the temporal window length; Calculate mutual information within each time-domain window length under different time-domain window lengths; As the time-domain window length is gradually increased, the mutual information gain under each window length is statistically analyzed. It also calculates the rate of change of information increment between adjacent window scales to form the marginal benefit curve of information gain; The inflection point of the marginal benefit curve of information gain is used as the first fusion window metric.

2. The method for risk early warning through health data analysis of the elderly according to claim 1, characterized in that, The step of alternately advancing the first monitoring frequency and the first fusion window scale in the directions of information gain and information attenuation by a preset step size is specifically as follows: The information gain direction advancement is to start with the first monitoring frequency and the first fusion window scale, increase the monitoring frequency, and increase it by a preset step size each time. After each increase, the corresponding monitoring frequency and fusion window scale are recorded. The information attenuation direction is advanced by starting from the first monitoring frequency and the first fusion window scale, reducing the monitoring frequency by a preset step size each time, and recording the corresponding monitoring frequency and fusion window scale.

3. The method for risk early warning through health data analysis of the elderly according to claim 2, characterized in that, The iterations are repeated several times, and the obtained data are randomly combined to generate several sets of data analysis control sets, specifically: Several monitoring frequencies and several fusion window scales are randomly combined to obtain several sets of data analysis control sets. Each data analysis control set contains one monitoring frequency and one fusion window scale.

4. The method for risk early warning through health data analysis of the elderly according to claim 3, characterized in that, The specific method for obtaining the response capability characteristics is as follows: Collect the actual occurrence time and corresponding early warning issuance time of all abnormal events of the target monitoring object within the monitoring period; Calculate the response delay for each abnormal event and obtain the average response delay; Based on the monitoring frequency set by the system, the response capability characteristics are calculated according to the ratio of average response delay to monitoring frequency.

5. The method for risk early warning through health data analysis of the elderly according to claim 4, characterized in that, The specific method for obtaining the data redundancy is as follows: Acquire time-series health data of the target monitoring object within a certain monitoring period, calculate the autocorrelation coefficient of the data series, and calculate the autocorrelation coefficients corresponding to all lag orders from 1 to the preset maximum lag value. The data redundancy is obtained by averaging the autocorrelation coefficients of all lag orders and using this as the characteristic value of data redundancy.

6. The method for risk early warning through health data analysis of the elderly according to claim 5, characterized in that, The method for obtaining the early warning quality characteristics is as follows: For each time window, calculate the recall rate of the risk warning; Calculate the difference in recall between adjacent sequences and extract the symbol change sequence; The number of alternations between positive and negative signs in the sign change sequence is counted and used as a quality characteristic value for early warning.

7. The method for risk early warning through health data analysis of the elderly according to claim 6, characterized in that, The data analysis control set display model is constructed based on the first evaluation value, monitoring frequency, and fusion window scale. Each data analysis control set and its corresponding first evaluation value are mapped to the data analysis control set display model, specifically as follows: Construct a curved empty display model as the data analysis control set display model, where the first evaluation value is the vertical coordinate, and the two bottom coordinates of the three-dimensional coordinates of the monitoring frequency and the fusion window scale are arbitrary in position; The monitoring frequency, fusion window scale, and corresponding first evaluation value of each data analysis control set are mapped to a data point in the data analysis control set display model.

8. A system using the elderly health data analysis and risk warning method as described in any one of claims 1-7, characterized in that, It includes a monitoring frequency initialization module, a fusion window scale initialization module, a group data analysis control set generation module, a control evaluation module, and a control set optimization module; The monitoring frequency initialization module is used to obtain the first monitoring frequency based on multi-scale smoothing and by searching for the minimum redundancy point. The fusion window scale initialization module is used to obtain the first fusion window scale based on redundancy reduction. The data analysis control set generation module is used to alternately advance the first monitoring frequency and the first fusion window scale in the direction of information gain and information reduction by a preset step size, iterate several times, and randomly combine the obtained data to generate several sets of data analysis control sets. The control evaluation module is used to apply several sets of data analysis control sets to health data analysis risk warning, and to collect response capability characteristics, data redundancy characteristics and warning quality characteristics, and to calculate the first evaluation value of each set of data analysis control sets by weighting. The control set optimization module is used to construct a data analysis control set display model based on the first evaluation value, monitoring frequency, and fusion window scale. It maps each data analysis control set and its corresponding first evaluation value to the data analysis control set display model and fits it to the first surface. According to the analysis requirements, it selects the best data analysis control set from the first surface for application in health data analysis risk warning.

Images