A remote nursing management method based on data feature mining

Abstract

The present application relates to the technical field of data feature mining, and particularly relates to a remote nursing management method based on data feature mining.The method comprises the following steps: obtaining patient specific conditions and demand data; selecting and configuring data sources according to the patient specific conditions and demand data to obtain a data source configuration table; collecting multi-dimensional data according to the data source configuration table and pre-processing the multi-dimensional data to obtain comprehensive patient data; modeling individual baselines for the comprehensive patient data to obtain an individual baseline model; standardizing real-time data of the comprehensive patient data to obtain standardized real-time data; calculating deviations according to the standardized real-time data and the individual baseline model to obtain a deviation value sequence; and analyzing the correlation of indexes for the deviation value sequence to obtain an index correlation matrix.The present application improves the utilization of data feature mining, makes remote nursing management more accurate and personalized, and ultimately improves nursing quality and efficiency.

Description

Technical Field

[0001] This invention relates to the field of data feature mining technology, and in particular to a remote nursing management method based on data feature mining. Background Technology

[0002] Early remote nursing management methods relied primarily on monitoring simple physiological parameters such as blood pressure and heart rate, and the data analysis methods were also relatively simple. Current methods have gradually evolved from simple threshold judgments to statistical model-based analyses, such as linear regression and logistic regression.

[0003] However, these methods still suffer from insufficient feature utilization, mainly in the following aspects:

[0004] 1. Lack of exploration of nonlinear relationships: Traditional statistical models typically assume linear relationships between features. However, in reality, many physiological indicators and health conditions exhibit complex nonlinear relationships. For example, the relationship between blood glucose levels and insulin secretion is not a simple linear one. Ignoring these nonlinear relationships leads to reduced predictive accuracy of the model.

[0005] 2. Failure to effectively utilize higher-order features: Higher-order features are features composed of combinations of multiple basic features, which can more comprehensively reflect a patient's health status. For example, combining the three basic features of blood pressure, heart rate, and body temperature can construct a new higher-order feature, which is more effective in predicting a patient's health risk than using any one of the basic features alone. However, many existing methods do not effectively utilize higher-order features. Summary of the Invention

[0006] Therefore, it is necessary to provide a remote nursing management method based on data feature mining to solve at least one of the above-mentioned technical problems.

[0007] To achieve the above objectives, a remote nursing management method based on data feature mining includes the following steps:

[0008] Step S1: Obtain patient's specific situation and needs data; select and configure data sources based on patient's specific situation and needs data to obtain data source configuration table; collect multi-dimensional data based on data source configuration table and perform multi-dimensional data preprocessing to obtain comprehensive patient data;

[0009] Step S2: Perform individualized baseline modeling on the patient's comprehensive data to obtain an individual baseline model;

[0010] Step S3: Perform real-time data standardization on the comprehensive patient data to obtain standardized real-time data; calculate the deviation based on the standardized real-time data and the individual baseline model to obtain the deviation value sequence; perform indicator correlation analysis on the deviation value sequence to obtain the indicator correlation matrix; perform comprehensive deviation assessment analysis based on the indicator correlation matrix and the deviation value sequence to obtain the deviation assessment report.

[0011] Step S4: Acquire patient's historical health data; integrate the deviation assessment report and patient's historical health data to obtain an integrated dataset; extract deviation features, process historical data features, and extract environmental features from the integrated dataset to obtain a deviation feature table, a historical feature table, and an environmental feature table, respectively; construct a high-order combined feature based on the deviation feature table, historical feature table, and environmental feature table to obtain a combined feature table; select predictive features from the deviation feature table, historical feature table, environmental feature table, and combined feature table to obtain a predictive feature set; construct a risk prediction model based on the predictive feature set to obtain a risk prediction model; perform risk prediction processing using the risk prediction model to obtain a risk prediction report; trigger early warnings and send information based on the risk prediction report to obtain early warning information, thereby realizing remote nursing management tasks.

[0012] This invention, by acquiring individualized patient information and needs, and selecting appropriate data sources and performing data preprocessing based on this information, ensures that the collected data is targeted, complete, and of high quality. This lays a solid data foundation for subsequent personalized baseline modeling, deviation analysis, and risk prediction. By performing personalized baseline modeling on comprehensive patient data, the unique physiological and behavioral patterns of each patient can be learned, constructing a personalized baseline model. This avoids errors caused by using reference ranges based on group averages, improving the accuracy of subsequent deviation analysis and risk prediction. By standardizing real-time data and combining it with the individual baseline model to calculate the deviation value sequence, the degree of deviation between the patient's current physiological state and the baseline state can be accurately quantified. Furthermore, through indicator correlation analysis, the complex interactions between different physiological indicators can be revealed, thereby more comprehensively assessing the patient's health status and generating a deviation assessment report, providing a more accurate basis for subsequent risk prediction. Integrating historical health data, deviation assessment reports, and environmental data, and performing feature extraction, combination, and selection, a feature set containing rich information can be constructed for training the risk prediction model. By utilizing multi-source data and higher-order combined features, various factors affecting patient health can be considered more comprehensively, improving the accuracy and personalization of risk prediction. Ultimately, risk level classification and early warning triggering based on risk prediction results enable remote real-time monitoring and alerts, promptly reminding medical staff and patients' families, thereby effectively preventing and controlling the occurrence and development of diseases and improving nursing quality and efficiency. Therefore, this invention provides a remote nursing management method based on data feature mining. Through techniques such as indicator correlation analysis, higher-order combined feature construction, deviation value sequence analysis, and multi-dimensional data integration, it more fully mines data features, thereby improving the accuracy of risk prediction and solving the problem of insufficient feature utilization in existing methods. This makes remote nursing management more precise and personalized, ultimately improving nursing quality and efficiency.

[0013] Preferably, step S1 includes the following steps:

[0014] Step S11: Obtain patient's specific situation and needs data; select and configure data sources based on patient's specific situation and needs data to obtain data source configuration table;

[0015] Step S12: Collect data in real time according to the data source configuration table to obtain the raw data stream;

[0016] Step S13: Perform data preprocessing on the original data stream to obtain a preprocessed data stream;

[0017] Step S14: Extract data features from the preprocessed data stream to obtain a feature data stream;

[0018] Step S15: Integrate the feature data streams to obtain comprehensive patient data.

[0019] This invention collects data on patients' specific conditions and needs, and uses this data to select and configure data sources, generating a personalized data source configuration table. This ensures that the collected data matches the patient's actual situation and needs, avoiding the collection of unnecessary data or the omission of key data, improving the efficiency and relevance of data collection, and laying the foundation for subsequent personalized analysis and risk prediction. Real-time data collection based on the data source configuration table ensures a standardized and regulated data collection process, acquiring various types of data to form a raw data stream, providing sufficient data sources for subsequent data preprocessing and feature extraction. Preprocessing the raw data stream, including data cleaning, noise reduction, and missing value imputation, effectively improves data quality, removes noise and outliers from subsequent analysis, and fills in missing data, thereby ensuring the accuracy and reliability of subsequent analysis. Extracting meaningful features from the preprocessed data stream transforms the raw data into more representative and interpretable feature data, reducing data dimensionality, computational complexity, and highlighting important physiological patterns and trends, providing more valuable information for subsequent baseline modeling and deviation analysis. Integrating feature data streams from different sources according to timestamps to form Patient Integrated Data (PID) can fuse multi-dimensional data into a unified data structure, facilitating subsequent analysis and processing, and providing more comprehensive patient health information, thereby improving the accuracy and comprehensiveness of risk prediction.

[0020] Preferably, step S2 includes the following steps:

[0021] Step S21: Perform data segmentation on the comprehensive patient data to obtain the training dataset and the validation dataset;

[0022] Step S22: Perform feature selection on the training dataset to obtain the filtered feature set;

[0023] Step S23: Train the baseline model based on the training dataset and the selected feature set to obtain the initial baseline model;

[0024] Step S24: Validate and optimize the initial baseline model using the validation dataset to obtain the optimized baseline model;

[0025] Step S25: Update the optimized baseline model based on the patient's comprehensive data to obtain the individual baseline model.

[0026] This invention effectively avoids model overfitting and improves the model's generalization ability by dividing comprehensive patient data into training and validation datasets. This ensures the model maintains good predictive performance even on unseen data, thus more accurately reflecting the patient's true physiological baseline. Feature selection identifies key features relevant to the baseline model, removing redundant features and noise, simplifying the model structure, improving training efficiency and interpretability, and reducing the risk of overfitting, making the baseline model more targeted and accurate. Training the model using the training dataset and the selected feature set allows the model to learn the distribution patterns of individual patient data and the relationships between features, building an initial baseline model that provides a foundation for subsequent model optimization and updates. Validating and optimizing the initial baseline model using the validation dataset allows for performance evaluation. Based on the evaluation results, model parameters can be adjusted or a more suitable model can be selected, improving the model's predictive accuracy and stability, making it more accurately reflect the patient's individual physiological baseline. Regularly updating the optimized baseline model with the latest comprehensive patient data allows the model to adapt to dynamic changes in patient physiological data, maintaining the timeliness of the baseline model and improving its predictive ability for future physiological data changes, thus more accurately reflecting the patient's current physiological state and potential health risks.

[0027] Preferably, step S3 includes the following steps:

[0028] Step S31: Perform real-time data standardization on the comprehensive patient data to obtain standardized real-time data; extract baseline distribution parameters from the individual baseline model to obtain baseline distribution parameters;

[0029] Step S32: Calculate the real-time data probability density based on the baseline distribution parameters and standardized real-time data to obtain the real-time data probability density;

[0030] Step S33: Perform probability density transformation on the real-time data probability density to obtain the deviation value; sort the deviation value by time series to obtain the deviation value sequence;

[0031] Step S34: Perform correlation analysis on the deviation value sequence to obtain the correlation matrix;

[0032] Step S35: Perform a comprehensive deviation assessment based on the correlation matrix of the indicators and the deviation value sequence to obtain a deviation assessment score; generate a deviation assessment report based on the deviation assessment score and the deviation value sequence.

[0033] This invention standardizes real-time data and extracts baseline distribution parameters from individual baseline models, eliminating dimensional differences between different indicators, making the data comparable, and providing necessary input for subsequent deviation and probability density calculations. This allows for a more accurate assessment of the deviation between the patient's real-time data and the baseline model. By calculating the probability density of standardized real-time data under the baseline distribution, the degree of deviation between the real-time data and the baseline model can be quantified. The lower the probability density value, the greater the deviation, providing a basis for subsequent deviation value calculations and thus more precisely depicting changes in the patient's health status. Converting the probability density into deviation values ​​and sorting the deviation values ​​by time series provides a more intuitive reflection of the degree of deviation between the real-time data and the baseline model, while retaining time series information. This facilitates subsequent correlation analysis of indicators and comprehensive deviation assessment, thus more clearly demonstrating the changing trends of the patient's physiological data. Correlation analysis of the deviation value series reveals the relationships between different physiological indicators, such as the correlation between heart rate and blood pressure, and helps identify which indicators' deviation values ​​change synchronously. This provides a more comprehensive assessment of the patient's health status and important evidence for subsequent comprehensive deviation assessments. By combining the correlation matrix of indicators and the sequence of deviation values ​​to conduct a comprehensive deviation assessment and generate a deviation assessment report, the deviation information of multiple indicators can be integrated into a single deviation assessment score, which can more comprehensively reflect the patient's overall health status and present the assessment results in a visual way. This makes it easier for medical staff to quickly understand the patient's condition and provides important input for subsequent risk prediction.

[0034] Preferably, step S34 includes the following steps:

[0035] Step S341: Construct a multidimensional time series matrix from the deviation value sequence to obtain the deviation value matrix;

[0036] Step S342: Divide the deviation value matrix into time windows to obtain the deviation value window sequence;

[0037] Step S343: Extract the single-window deviation matrix from the deviation value window sequence to obtain the single-window deviation value matrix;

[0038] Step S344: Calculate the linear correlation between indicators on the single-window deviation value matrix to obtain a linear correlation matrix sequence; calculate the nonlinear correlation between indicators on the single-window deviation value matrix to obtain a nonlinear correlation matrix sequence.

[0039] Step S345: Perform linear and nonlinear correlation fusion on the linear correlation matrix sequence and the nonlinear correlation matrix sequence to obtain the comprehensive correlation matrix sequence;

[0040] Step S346: Perform time window comprehensive analysis on the comprehensive correlation matrix sequence to obtain the index correlation matrix.

[0041] This invention constructs a multi-dimensional time series matrix, integrating the deviation value sequences of various indicators into a unified matrix structure. This facilitates subsequent time window segmentation and correlation analysis, laying a data foundation for more effectively uncovering the correlations between indicators. Segmenting the deviation value matrix according to time windows decomposes long-term deviation value sequences into multiple short-term data segments, allowing for the analysis of the correlations between indicators within each time period, thereby capturing the dynamic relationships between indicators at different time scales. Extracting the deviation value matrix of a single time window from the deviation value window sequence provides independent data units for subsequent linear and nonlinear correlation calculations, ensuring the accuracy and independence of the calculation results. Calculating the linear and nonlinear correlations between indicators separately captures the complex relationships between them more comprehensively, considering not only simple linear relationships but also potential nonlinear correlations, thus delving deeper into the interactions between indicators. Merging the linear and nonlinear correlation matrix sequences comprehensively considers different types of correlations between indicators, obtaining more comprehensive and accurate indicator correlation information, thereby improving the accuracy of subsequent comprehensive deviation assessments. By performing a time window comprehensive analysis on the comprehensive correlation matrix sequence, the correlation information of different time windows can be integrated to obtain a final indicator correlation matrix. This matrix reflects the overall correlation between indicators over a period of time, and is more stable and representative, providing a more reliable basis for subsequent risk prediction.

[0042] Preferably, step S344 specifically includes:

[0043] The linear correlation between each pair of indicators within a single window is calculated on the single-window deviation value matrix to obtain the single-window linear correlation matrix.

[0044] Linear correlation matrix aggregation is performed on the single-window linear correlation matrix to obtain a sequence of linear correlation matrices;

[0045] The single-window deviation value matrix is ​​discretized to obtain the discretized deviation value matrix.

[0046] The joint probability distribution matrix is ​​obtained by calculating the joint probability distribution based on the discretized deviation value matrix; the univariate probability distribution data is obtained by calculating the univariate probability distribution based on the discretized deviation value matrix.

[0047] The mutual information value is calculated based on the joint probability distribution matrix and the univariate probability distribution data to obtain the single-window mutual information matrix.

[0048] The nonlinear correlation matrix sequence is constructed by using the deviation value window sequence and the single-window mutual information matrix.

[0049] This invention quantifies the strength and direction of linear relationships between indicators within a single time window by calculating the pairwise linear correlations between indicators, thus constructing a single-window linear correlation matrix and providing a foundation for subsequent linear correlation analysis. Aggregating the linear correlation matrices from each time window into a sequence preserves linear correlation information from different time windows, facilitating subsequent analysis of the dynamic changes in linear correlations between indicators. Discretizing the continuous deviation values ​​in the single-window deviation matrix simplifies data representation, reduces computational complexity, and provides the necessary input data format for subsequent nonlinear correlation calculations (e.g., mutual information). Calculating the joint probability distribution and univariate probability distribution provides the necessary probabilistic basis for subsequent mutual information calculations, used to quantify the strength of nonlinear associations between indicators. By calculating the mutual information values ​​between indicators, the strength of nonlinear associations between indicators within the current time window can be quantified, constructing a single-window mutual information matrix and providing a foundation for subsequent nonlinear correlation analysis. Aggregating the mutual information matrices from each time window into a sequence of nonlinear correlation matrices preserves nonlinear correlation information from different time windows, facilitating subsequent analysis of the dynamic changes in nonlinear correlations between indicators and supporting a more comprehensive understanding of the interactions between indicators.

[0050] Preferably, step S4 includes the following steps:

[0051] Step S41: Obtain patient's historical health data; integrate the deviation assessment report and patient's historical health data to obtain an integrated dataset;

[0052] Step S42: Extract deviation features from the integrated dataset to obtain a deviation feature table;

[0053] Step S43: Perform historical data feature processing on the integrated dataset to obtain a historical feature table;

[0054] Step S44: Extract environmental features from the integrated dataset to obtain an environmental feature table;

[0055] Step S45: Construct a high-order combined feature table based on the deviation feature table, historical feature table, and environmental feature table;

[0056] Step S46: Select predictive features from the deviation feature table, historical feature table, environmental feature table, and combined feature table to obtain the predictive feature set;

[0057] Step S47: Construct a risk prediction model based on the prediction feature set and the preset labeled dataset to obtain the risk prediction model; use the risk prediction model to make risk predictions on standardized real-time data and generate a risk prediction report to obtain the risk prediction report.

[0058] Step S48: Classify the risk levels according to the risk prediction report to obtain a risk level list; trigger the early warning and send the information according to the risk level list and the preset early warning rule base to obtain the early warning information.

[0059] This invention integrates patients' historical health data and deviation assessment reports to form a unified integrated dataset. This dataset combines patients' static health information with dynamic physiological data trends, providing a more comprehensive data foundation for subsequent feature extraction and risk prediction, thereby improving the accuracy of risk prediction. Extracting deviation features from the integrated dataset quantifies the dynamic trends of patients' physiological data, such as short-term fluctuations and long-term trends, and transforms these trends into features usable for risk prediction, thus improving the sensitivity of the risk prediction model. Structured processing of patients' historical health data and its transformation into features usable for risk prediction allows the integration of static health information such as past medical history, medication records, and allergy history into the risk prediction model, thereby improving the personalization and accuracy of risk prediction. Extracting environmental features from the integrated dataset incorporates the influence of environmental factors such as temperature, humidity, and air quality into the risk prediction model, thus considering factors affecting patients' health more comprehensively and improving the accuracy of risk prediction. By constructing high-order combined features, the interactions between different features can be captured, such as the interaction between deviation features and historical features, and the interaction between environmental features and historical features, thereby uncovering deeper risk factors and further improving the performance of the risk prediction model. Feature selection for prediction can remove redundant and irrelevant features, reduce model complexity, improve training efficiency and generalization ability, and avoid overfitting, thus making the risk prediction model more robust and stable. The risk prediction model is trained using a predictive feature set and a labeled dataset, and then used to predict risks from real-time patient data, generating a risk prediction report. The extracted features can be used for actual risk prediction, and the prediction results are visualized. Based on the risk prediction report, risk levels are classified, and corresponding warning information is triggered according to preset warning rules. This transforms the risk prediction results into specific risk levels and promptly notifies relevant personnel, thereby achieving early warning.

[0060] Preferably, step S42 includes the following steps:

[0061] Step S421: Define time windows of different lengths for bias feature extraction based on the integrated dataset to obtain a list of bias time windows;

[0062] Step S422: Group the integrated dataset into deviation scores based on time window grouping according to the deviation time window list to obtain the time window deviation score set;

[0063] Step S423: Calculate the statistical characteristics of the time window deviation on the time window deviation fraction set to obtain the statistical characteristic set of the time window deviation;

[0064] Step S424: Calculate the time window deviation trend characteristics of the time window deviation fraction set to obtain the time window deviation trend characteristic set;

[0065] Step S425: Merge the time window deviation statistical feature set and the time window deviation trend feature set to obtain the deviation feature table.

[0066] This invention, by defining time windows of varying lengths, captures deviations in patient physiological data across multiple time scales, including short-term fluctuations, medium-term changes, and long-term trends, providing a more comprehensive temporal perspective for subsequent deviation feature extraction. Grouping deviation scores according to the defined time windows separates deviation information from different time scales, preparing for the subsequent calculation of statistical and trend characteristics for each time window. Calculating the statistical characteristics of deviation scores within each time window, such as mean, standard deviation, maximum, and minimum values, quantifies the distribution of deviations across different time scales, providing richer feature information for risk prediction. Calculating the trend characteristics of deviation scores within each time window, such as slope and rate of change, captures the changing trends of patient physiological data deviations, such as whether the deviation is gradually increasing or decreasing, providing more dynamic feature information for risk prediction. Merging the statistical and trend characteristics of different time windows into a deviation feature table integrates deviation information from multiple time scales, forming a more comprehensive deviation feature set, thereby improving the accuracy and robustness of the risk prediction model.

[0067] Preferably, step S44 includes the following steps:

[0068] Step S441: Define time windows of different lengths for environmental feature extraction based on the integrated dataset to obtain a list of environmental time windows;

[0069] Step S442: Group the integrated dataset into environmental data based on time window grouping according to the environmental time window list to obtain the time window environmental dataset;

[0070] Step S443: Perform time window environmental statistical features on the time window deviation score set to obtain the time window environmental statistical feature set;

[0071] Step S444: Calculate the environmental change features of the time window based on the environmental statistical feature set and the environmental dataset of the time window to obtain the environmental change feature set of the time window;

[0072] Step S445: Merge the time window environment statistical feature set and the time window environment change feature set to obtain the environment feature table.

[0073] This invention, by defining environmental time windows of varying lengths, allows for a multi-faceted examination of the impact of environmental factors on patients, including short-term and long-term effects, thereby more comprehensively capturing the relationship between environmental factors and patient health. Grouping environmental data according to different time windows separates environmental information at different time scales, preparing for subsequent calculations of environmental statistical and variation characteristics for each time window. Calculating environmental statistical characteristics within each time window, such as average temperature, maximum temperature, minimum temperature, and average humidity, quantifies environmental conditions at different time scales, providing richer environmental information for risk prediction. Calculating environmental variation characteristics within each time window, such as temperature and humidity change rates, captures trends in environmental factors, such as whether temperature is gradually increasing or decreasing, providing more dynamic environmental information for risk prediction. Merging the environmental statistical and variation characteristics from different time windows into an environmental feature table integrates environmental information from multiple time scales, forming a more comprehensive environmental feature set, thereby improving the accuracy and sensitivity of the risk prediction model to environmental factors.

[0074] Preferably, step S45 includes the following steps:

[0075] Step S451: Prepare the feature matrix for the deviation feature table, historical feature table, and environmental feature table to obtain the feature matrix;

[0076] Step S452: Generate basic combined features from the feature matrix to obtain the basic combined feature set;

[0077] Step S453: Use deep feature synthesis technology to synthesize deep features from the basic combined feature set to obtain deep features;

[0078] Step S454: Optimize the deep features using a genetic algorithm to obtain an optimized feature set;

[0079] Step S455: Generate a combined feature table based on the optimized combined feature set to obtain the combined feature table.

[0080] This invention integrates different types of features into a unified data structure by merging the deviation feature table, historical feature table, and environmental feature table into a single feature matrix. This facilitates subsequent combined feature generation and feature selection, and ensures consistent processing of all features in subsequent analysis. Basic combined feature generation, such as addition, subtraction, multiplication, and division operations between features, allows for preliminary exploration of feature interactions and the generation of new features, expanding the feature space and laying the foundation for subsequent deep feature synthesis. Deep feature synthesis technology automatically generates more complex and deeper combined features from the basic combined feature set, thereby more comprehensively capturing the nonlinear relationships and interactions between features and significantly improving their expressive and predictive capabilities. Optimizing deep features using a genetic algorithm can filter out the subset of features that contribute most to risk prediction from a large number of combined features, reducing feature redundancy, improving model efficiency and generalization ability, and preventing overfitting. Organizing the optimized combined feature set into a combined feature table provides a structured feature dataset for subsequent feature selection and model training, facilitating use and management in subsequent steps and ensuring that the features ultimately used for model training are the optimal subset. Attached Figure Description

[0081] Figure 1 This is a flowchart illustrating the steps of a remote nursing management method based on data feature mining.

[0082] Figure 2 for Figure 1 A detailed flowchart illustrating the implementation steps of step S3.

[0083] Figure 3 for Figure 1 A detailed flowchart illustrating the implementation steps of step S4.

[0084] The objectives, features, and advantages of this invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0085] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0086] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods.

[0087] It should be understood that although the terms "first," "second," etc., may be used herein to describe various units, these units should not be limited by these terms. These terms are used merely to distinguish one unit from another. For example, without departing from the scope of the exemplary embodiments, a first unit may be referred to as a second unit, and similarly, a second unit may be referred to as a first unit. The term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0088] To achieve the above objectives, please refer to Figures 1 to 3 A remote nursing management method based on data feature mining includes the following steps:

[0089] Step S1: Obtain patient's specific situation and needs data; select and configure data sources based on patient's specific situation and needs data to obtain data source configuration table; collect multi-dimensional data based on data source configuration table and perform multi-dimensional data preprocessing to obtain comprehensive patient data;

[0090] Step S2: Perform individualized baseline modeling on the patient's comprehensive data to obtain an individual baseline model;

[0091] Step S3: Perform real-time data standardization on the comprehensive patient data to obtain standardized real-time data; calculate the deviation based on the standardized real-time data and the individual baseline model to obtain the deviation value sequence; perform indicator correlation analysis on the deviation value sequence to obtain the indicator correlation matrix; perform comprehensive deviation assessment analysis based on the indicator correlation matrix and the deviation value sequence to obtain the deviation assessment report.

[0092] Step S4: Acquire patient's historical health data; integrate the deviation assessment report and patient's historical health data to obtain an integrated dataset; extract deviation features, process historical data features, and extract environmental features from the integrated dataset to obtain a deviation feature table, a historical feature table, and an environmental feature table, respectively; construct a high-order combined feature based on the deviation feature table, historical feature table, and environmental feature table to obtain a combined feature table; select predictive features from the deviation feature table, historical feature table, environmental feature table, and combined feature table to obtain a predictive feature set; construct a risk prediction model based on the predictive feature set to obtain a risk prediction model; perform risk prediction processing using the risk prediction model to obtain a risk prediction report; trigger early warnings and send information based on the risk prediction report to obtain early warning information, thereby realizing remote nursing management tasks.

[0093] In this embodiment of the invention, reference is made to Figure 1 The diagram shown illustrates the steps of the remote nursing management method based on data feature mining according to the present invention. In this example, the remote nursing management method based on data feature mining includes the following steps:

[0094] Step S1: Obtain patient's specific situation and needs data; select and configure data sources based on patient's specific situation and needs data to obtain data source configuration table; collect multi-dimensional data based on data source configuration table and perform multi-dimensional data preprocessing to obtain comprehensive patient data;

[0095] In this embodiment of the invention, patient information, health status, nursing needs, and desired monitoring indicators are first acquired through various means. Then, based on this data and preset rules, appropriate data acquisition devices and sensors are selected, acquisition parameters are configured, and a data source configuration table is generated. Finally, multi-dimensional data is acquired according to the configuration table, and the acquired raw data is preprocessed, including data cleaning, noise reduction, missing value imputation, and format standardization, ultimately yielding comprehensive patient data (PID) containing multi-dimensional information.

[0096] Step S2: Perform individualized baseline modeling on the patient's comprehensive data to obtain an individual baseline model;

[0097] In this embodiment of the invention, a personalized baseline model (IBM) is constructed using Patient Comprehensive Data (PID). First, the PID is divided into a training dataset and a validation dataset in chronological order. Then, feature selection is performed on the training dataset to identify key features relevant to the target variable. Next, a suitable machine learning algorithm (e.g., AR, GMM, RNN, etc.) is used to train the model on the training dataset, obtaining an initial baseline model. Afterward, the initial model is validated and optimized using the validation dataset, adjusting model parameters or trying different models to obtain an optimized baseline model. Finally, the optimized model is periodically updated using the latest patient comprehensive data to obtain the final individual baseline model (IBM).

[0098] Step S3: Perform real-time data standardization on the comprehensive patient data to obtain standardized real-time data; calculate the deviation based on the standardized real-time data and the individual baseline model to obtain the deviation value sequence; perform indicator correlation analysis on the deviation value sequence to obtain the indicator correlation matrix; perform comprehensive deviation assessment analysis based on the indicator correlation matrix and the deviation value sequence to obtain the deviation assessment report.

[0099] In this embodiment of the invention, the real-time data in the patient's comprehensive data is first standardized, for example, using the Z-score standardization method. Then, based on the baseline distribution parameters extracted from the Individual Baseline Model (IBM) (e.g., the mean, variance, and weights of the GMM), the probability density value of the standardized real-time data under the baseline distribution is calculated and converted into a deviation value, for example, using a negative logarithmic transformation. The deviation values ​​of each indicator are arranged in chronological order to form a deviation value sequence. Next, correlation analysis is performed on the deviation value sequence to calculate the linear correlation (e.g., Pearson correlation coefficient) and non-linear correlation (e.g., mutual information) between indicators, and the two are fused to obtain an indicator correlation matrix. Finally, based on the indicator correlation matrix and the deviation value sequence, the deviation of the patient's physiological data is comprehensively evaluated, a deviation assessment score is calculated, and a deviation assessment report (DAR) is generated.

[0100] Step S4: Acquire patient's historical health data; integrate the deviation assessment report and patient's historical health data to obtain an integrated dataset; extract deviation features, process historical data features, and extract environmental features from the integrated dataset to obtain a deviation feature table, a historical feature table, and an environmental feature table, respectively; construct a high-order combined feature based on the deviation feature table, historical feature table, and environmental feature table to obtain a combined feature table; select predictive features from the deviation feature table, historical feature table, environmental feature table, and combined feature table to obtain a predictive feature set; construct a risk prediction model based on the predictive feature set to obtain a risk prediction model; perform risk prediction processing using the risk prediction model to obtain a risk prediction report; trigger and send early warnings based on the risk prediction report to obtain early warning information, thereby realizing remote nursing management tasks.

[0101] In this embodiment of the invention, the patient's historical health data is acquired and integrated with a Deviation Assessment Report (DAR) to form an integrated dataset. Then, feature extraction is performed on the integrated dataset, including extracting deviation features (e.g., mean, standard deviation, trend) from the Deviation Assessment scores, processing historical health data to obtain historical features (e.g., disease diagnoses, medication records), and extracting environmental features (e.g., temperature, humidity). Next, high-order combined features are constructed based on the extracted features, for example, through mathematical or logical operations between features. Then, all features are selected to obtain a predicted feature set. Next, a risk prediction model (e.g., random forest, SVM, GBDT, etc.) is trained using the labeled dataset and the predicted feature set. The trained model is used to predict the risk of the patient's real-time data, generating a Risk Prediction Report (RPR). Finally, based on the risk prediction results in the RPR, risk levels are classified, and corresponding warning information is triggered according to a preset warning rule base, which is then sent to doctors and family members.

[0102] Preferably, step S1 includes the following steps:

[0103] Step S11: Obtain patient's specific situation and needs data; select and configure data sources based on patient's specific situation and needs data to obtain data source configuration table;

[0104] Step S12: Collect data in real time according to the data source configuration table to obtain the raw data stream;

[0105] Step S13: Perform data preprocessing on the original data stream to obtain a preprocessed data stream;

[0106] Step S14: Extract data features from the preprocessed data stream to obtain a feature data stream;

[0107] Step S15: Integrate the feature data streams to obtain comprehensive patient data.

[0108] In this embodiment of the invention, firstly, specific patient information and needs data are obtained through electronic health record systems, patient questionnaires, and doctor interviews. This includes data such as age, gender, past medical history, current medication use, allergy history, lifestyle habits, nursing needs, and desired physiological indicators to be monitored. Then, based on this data, combined with predefined disease-data source mapping rules and the patient's individual needs, the system automatically selects appropriate data collection sources and configures collection parameters. For example, for patients with hypertension and sleep apnea, the system automatically selects a smart blood pressure monitor, heart rate monitor, and sleep apnea monitor, and configures the sampling frequency and data transmission method. Finally, a data source configuration table is generated, recording the data source type, collection frequency, data type, and data transmission protocol required for each patient. This configuration table guides the subsequent data collection process, ensuring that multi-dimensional data that meets the individual needs of the patient is collected.

[0109] Based on the "data source configuration table," corresponding wearable devices and smart home sensors, such as smartwatches, smart blood pressure monitors, blood glucose meters, sleep monitors, and ambient temperature and humidity sensors, are activated to collect real-time physiological, environmental, and behavioral data from patients. The data collection process follows preset parameters in the data source configuration table, such as sampling frequency, data type, and data transmission protocol. The collected data is transmitted and stored as a raw data stream composed of timestamps, data types, and values. For example, a heart rate data stream contains a series of timestamps and corresponding heart rate values.

[0110] After acquiring the raw data stream, a series of data preprocessing techniques are employed to improve data quality. First, data cleaning is performed using preset rules and thresholds to remove obviously erroneous data, such as data exceeding the normal range of physiological indicators. Second, signal processing techniques such as Kalman filtering or wavelet transform are applied to denoise the data, removing interference and noise. Third, missing data is filled using linear interpolation or model-based interpolation methods. Finally, data from different data sources are standardized in format and converted in units; for example, blood pressure data collected from different devices are converted to mmHg. The preprocessed data retains its data stream format, forming a preprocessed data stream that provides a high-quality data foundation for subsequent feature extraction.

[0111] Meaningful features are extracted from preprocessed data streams for subsequent modeling and analysis. For example, heart rate variability (HRV) features, such as RMSSD, SDNN, and pNN50, can be extracted from heart rate data streams; sleep duration, deep / light sleep ratio, sleep efficiency, and number of apneas can be extracted from sleep data streams; and steps, activity duration, activity intensity, and energy expenditure can be extracted from activity data streams. Furthermore, features such as average temperature, humidity, and light intensity can be extracted from environmental data. The extracted features are presented as a feature data stream in the form of timestamps and feature values. For example, an HRV feature data stream contains a series of timestamps and corresponding RMSSD, SDNN, and pNN50 feature values.

[0112] Multiple feature data streams are aligned and integrated according to timestamps to form Patient Integrated Data (PID). PIDs are stored in a structured format, containing multi-dimensional information about the patient's physiology, environment, and behavior. For example, a record in a PID might contain information such as heart rate, blood pressure, blood oxygen saturation, sleep stage, activity level, ambient temperature, and humidity at a specific time point. The structured storage of PIDs facilitates subsequent operations such as baseline modeling, bias analysis, and risk prediction. Timestamp alignment ensures the temporal consistency of data from different data sources, avoiding time deviations during data fusion.

[0113] Preferably, step S2 includes the following steps:

[0114] Step S21: Perform data segmentation on the comprehensive patient data to obtain the training dataset and the validation dataset;

[0115] Step S22: Perform feature selection on the training dataset to obtain the filtered feature set;

[0116] Step S23: Train the baseline model based on the training dataset and the selected feature set to obtain the initial baseline model;

[0117] Step S24: Validate and optimize the initial baseline model using the validation dataset to obtain the optimized baseline model;

[0118] Step S25: Update the optimized baseline model based on the patient's comprehensive data to obtain the individual baseline model.

[0119] In this embodiment of the invention, the patient comprehensive data (PID) is segmented chronologically. For example, the first 80% of the data is used as the training dataset for model training, and the last 20% is used as the validation dataset for model evaluation and parameter tuning. To ensure that the model can adapt to changes in patient data, a sliding window mechanism is used. For example, the training and validation datasets are updated daily, adding the latest day's data to the training dataset and removing the oldest day's data, while keeping the dataset size constant.

[0120] Feature selection is performed on the training dataset to remove redundant and irrelevant features, improving model efficiency and generalization ability. A filter-based feature selection method is employed. For example, the Pearson correlation coefficient between each feature and the target variable (e.g., heart rate) is calculated, and the features are ranked according to the absolute value of the correlation coefficient. The top N features, or features with an absolute correlation coefficient greater than a preset threshold, are selected to form a filtered feature set. For example, the 10 features most correlated with heart rate, such as resting heart rate, heart rate variability, sleep duration, and activity level, are selected to form the filtered feature set for the heart rate baseline model.

[0121] Based on the training dataset and feature set, a personalized baseline model is trained. Taking the heart rate baseline model as an example, an autoregressive (AR) model is used for training. Using historical heart rate data from the training dataset and other features selected from the feature set, such as sleep duration and activity level, the AR model is trained to learn the pattern of heart rate data changes over time. During training, the parameters of the AR model are estimated using the least squares method, and the optimal model order is selected using the Akaike Information Content Criterion (AIC) or the Bayesian Information Content Criterion (BIC). After training, an initial baseline model is obtained, which can predict future heart rate values ​​based on the patient's historical data and relevant features.

[0122] The initial baseline model is validated and optimized using a validation dataset. The mean squared error (MSE) or mean absolute error (MAE) of the model on the validation dataset is calculated to evaluate the model's prediction accuracy. If the model's prediction error is large, model parameters are adjusted, such as the order or regularization coefficient of the AR model, or other models are tried, such as Gaussian mixture models (GMMs) or recurrent neural networks (RNNs), to improve the model's prediction accuracy. Through iterative model validation and parameter tuning, an optimized baseline model is finally obtained, which has high prediction accuracy on the validation dataset.

[0123] To enable the baseline model to adapt to changes in patient physiological data, it is regularly updated and optimized using the latest comprehensive patient data. For example, incremental learning is performed daily using newly added comprehensive patient data. Specifically, new data is added to the training dataset, and the model parameters are fine-tuned using the updated training dataset. The model update frequency can be adjusted based on changes in the patient's condition and data. The updated model better reflects the patient's current physiological state, improving the accuracy and timeliness of the baseline model, ultimately resulting in an individual baseline model (IBM).

[0124] Preferably, step S3 includes the following steps:

[0125] Step S31: Perform real-time data standardization on the comprehensive patient data to obtain standardized real-time data; extract baseline distribution parameters from the individual baseline model to obtain baseline distribution parameters;

[0126] Step S32: Calculate the real-time data probability density based on the baseline distribution parameters and standardized real-time data to obtain the real-time data probability density;

[0127] Step S33: Perform probability density transformation on the real-time data probability density to obtain the deviation value; sort the deviation value by time series to obtain the deviation value sequence;

[0128] Step S34: Perform correlation analysis on the deviation value sequence to obtain the correlation matrix;

[0129] Step S35: Perform a comprehensive deviation assessment based on the correlation matrix of the indicators and the deviation value sequence to obtain a deviation assessment score; generate a deviation assessment report based on the deviation assessment score and the deviation value sequence.

[0130] As an example of the present invention, reference is made to... Figure 2 As shown, step S3 in this example includes:

[0131] Step S31: Perform real-time data standardization on the comprehensive patient data to obtain standardized real-time data; extract baseline distribution parameters from the individual baseline model to obtain baseline distribution parameters;

[0132] In this embodiment of the invention, firstly, the real-time data in the Patient Integrated Data (PID) is standardized. The Z-score standardization method is used, subtracting the mean of each indicator in the baseline model from the real-time data, and then dividing by the standard deviation of that indicator in the baseline model. The standardized data follows a standard normal distribution with a mean of 0 and a standard deviation of 1. For example, subtracting the average heart rate value of the heart rate baseline model from the patient's real-time heart rate value, and then dividing by the standard deviation of the heart rate baseline model, yields the standardized real-time heart rate value. Simultaneously, baseline distribution parameters are extracted from the Individual Baseline Model (IBM). Assuming that the IBM is constructed using a Gaussian Mixture Model (GMM), the mean, variance, and weights of the GMM corresponding to each indicator are extracted to form the baseline distribution parameters.

[0133] Step S32: Calculate the real-time data probability density based on the baseline distribution parameters and standardized real-time data to obtain the real-time data probability density;

[0134] In this embodiment of the invention, the probability density value of real-time data under the baseline distribution is calculated based on baseline distribution parameters and standardized real-time data. Taking heart rate as an example, if the heart rate baseline model uses a Gaussian mixture model (GMM), the probability density value of the standardized real-time heart rate value under the GMM is calculated based on the mean, variance, and weight parameters of the GMM. A corresponding probability density value is calculated for real-time data at each time point.

[0135] Step S33: Perform probability density transformation on the real-time data probability density to obtain the deviation value; sort the deviation value by time series to obtain the deviation value sequence;

[0136] In this embodiment of the invention, the probability density value of real-time data is converted into a deviation value. A negative logarithmic transformation method is used to take the negative logarithm of the probability density value to obtain the deviation value. The smaller the probability density value, the greater the deviation between the real-time data and the baseline model, and the larger the corresponding deviation value. The deviation values ​​of each indicator are sorted according to their timestamps to form a deviation value sequence. For example, the deviation values ​​corresponding to the real-time heart rate data at each time point are arranged in chronological order to form a heart rate deviation value sequence.

[0137] Step S34: Perform correlation analysis on the deviation value sequence to obtain the correlation matrix;

[0138] In this embodiment of the invention, correlation analysis is performed on the deviation value sequences. First, a multidimensional time series matrix is ​​constructed from the deviation value sequences of all indicators. Then, a sliding window technique is used to divide the multidimensional time series matrix into multiple time windows. For the deviation value data within each time window, the Pearson correlation coefficient between the indicators is calculated to measure the linear correlation between the indicators. Simultaneously, the mutual information between the indicators is calculated to measure the nonlinear correlation between the indicators. The linear and nonlinear correlations are weighted and averaged to obtain the final indicator correlation matrix.

[0139] Step S35: Perform a comprehensive deviation assessment based on the indicator correlation matrix and the deviation value sequence to obtain a deviation assessment score; generate a deviation assessment report based on the deviation assessment score and the deviation value sequence.

[0140] In this embodiment of the invention, a comprehensive assessment of the patient's physiological data deviation is performed based on an indicator correlation matrix and a deviation value sequence. A weighted summation method is used, multiplying the deviation value of each indicator by its weight in the correlation matrix, and then summing the results to obtain a deviation assessment score. The weights reflect the importance of the indicator and its correlation with other indicators. A higher deviation assessment score indicates a greater degree of deviation in the patient's physiological data and a higher potential health risk. Finally, a Deviation Assessment Report (DAR) is generated based on the deviation assessment score, the deviation value sequence, and the indicator correlation matrix. The DAR includes information such as the deviation value of each indicator, deviation trend, indicator correlation, and comprehensive assessment results, presented visually, such as in tables and charts, to facilitate medical staff's quick understanding of the patient's health status.

[0141] Preferably, step S34 includes the following steps:

[0142] Step S341: Construct a multidimensional time series matrix from the deviation value sequence to obtain the deviation value matrix;

[0143] Step S342: Divide the deviation value matrix into time windows to obtain the deviation value window sequence;

[0144] Step S343: Extract the single-window deviation matrix from the deviation value window sequence to obtain the single-window deviation value matrix;

[0145] Step S344: Calculate the linear correlation between indicators on the single-window deviation value matrix to obtain a linear correlation matrix sequence; calculate the nonlinear correlation between indicators on the single-window deviation value matrix to obtain a nonlinear correlation matrix sequence.

[0146] Step S345: Perform linear and nonlinear correlation fusion on the linear correlation matrix sequence and the nonlinear correlation matrix sequence to obtain the comprehensive correlation matrix sequence;

[0147] Step S346: Perform time window comprehensive analysis on the comprehensive correlation matrix sequence to obtain the index correlation matrix.

[0148] In this embodiment of the invention, the deviation value sequences of various indicators are integrated into a multi-dimensional time series matrix. Each row of the matrix represents a time point, and each column represents the deviation value of an indicator. For example, the first column is the heart rate deviation value sequence, the second column is the blood pressure deviation value sequence, the third column is the blood oxygen saturation deviation value sequence, and so on. This ensures that the deviation value sequences of all indicators are aligned in time; for example, the first row of the matrix contains the deviation values ​​of all indicators at the first time point.

[0149] The multidimensional time series matrix is ​​divided according to a preset time window size. For example, the time window size is set to 5 minutes, and the sliding window is divided in 1-minute increments. The deviation value data for each 5-minute period is used as a time window to form a series of deviation value window data. The data within each time window constitutes a sub-matrix, with the number of rows equal to the time window size (e.g., 5) and the number of columns equal to the number of indicators. All sub-matrices constitute the deviation value window sequence.

[0150] The deviation value data for each individual time window is extracted from the deviation value window sequence to form a single-window deviation value matrix. Each single-window deviation value matrix represents multi-dimensional deviation information within a time window. For example, the deviation value data for the first time window is extracted to form a matrix of 5 rows (5 minutes) x 1 column of indicator quantity.

[0151] For each single-window deviation value matrix, the linear and non-linear correlations between the indicators are calculated. Linear correlation is calculated using the Pearson correlation coefficient. For each pair of indicator data in the matrix (e.g., heart rate deviation and blood pressure deviation), a Pearson correlation coefficient is calculated, resulting in a coefficient between -1 and 1, reflecting the degree of linear correlation between the two indicators. The Pearson correlation coefficients between all pairs of indicators are used to construct a linear correlation matrix. Non-linear correlation is calculated using the mutual information method. First, the data in the single-window deviation value matrix is ​​discretized; for example, the deviation value of each indicator is divided into high, medium, and low levels. Then, the mutual information value between each pair of indicators is calculated to measure the degree of non-linear correlation. The mutual information values ​​between all pairs of indicators are used to construct a non-linear correlation matrix. The above calculations are performed for each single-window deviation value matrix, ultimately yielding a sequence of linear and non-linear correlation matrices.

[0152] The linear correlation matrix sequence and the nonlinear correlation matrix sequence are fused to obtain the comprehensive correlation matrix sequence. A weighted average method is used for fusion. For example, the weight of linear correlation is set to 0.5, and the weight of nonlinear correlation is set to 0.5. The linear correlation matrix and nonlinear correlation matrix for the corresponding time window are multiplied by their respective weights and then summed to obtain the comprehensive correlation matrix. The same operation is performed for each time window to finally obtain the comprehensive correlation matrix sequence.

[0153] A time-window comprehensive analysis is performed on the overall correlation matrix sequence to obtain the final indicator correlation matrix. The average correlation coefficient between each pair of indicators across all time windows is calculated, yielding the final indicator correlation matrix. This matrix reflects the overall correlation between indicators over a period of time. For example, the average correlation coefficient between heart rate and blood pressure across all time windows is calculated as the final correlation coefficient between heart rate and blood pressure in the indicator correlation matrix.

[0154] Preferably, step S344 specifically includes:

[0155] The linear correlation between each pair of indicators within a single window is calculated on the single-window deviation value matrix to obtain the single-window linear correlation matrix.

[0156] Linear correlation matrix aggregation is performed on the single-window linear correlation matrix to obtain a sequence of linear correlation matrices;

[0157] The single-window deviation value matrix is ​​discretized to obtain the discretized deviation value matrix.

[0158] The joint probability distribution matrix is ​​obtained by calculating the joint probability distribution based on the discretized deviation value matrix; the univariate probability distribution data is obtained by calculating the univariate probability distribution based on the discretized deviation value matrix.

[0159] The mutual information value is calculated based on the joint probability distribution matrix and the univariate probability distribution data to obtain the single-window mutual information matrix.

[0160] The nonlinear correlation matrix sequence is constructed by using the deviation value window sequence and the single-window mutual information matrix.

[0161] In this embodiment of the invention, a single-window deviation value matrix is ​​obtained. For each pair of indicators in this matrix, such as heart rate deviation and blood pressure deviation, their Pearson correlation coefficient is calculated. The formula for calculating the Pearson correlation coefficient is: covariance (X, Y) / (standard deviation (X) * standard deviation (Y)). Here, X and Y represent the deviation value vectors of the two indicators. The calculated Pearson correlation coefficient ranges from -1 to 1; a value closer to 1 indicates a stronger positive correlation, a value closer to -1 indicates a stronger negative correlation, and a value close to 0 indicates a weak correlation. The Pearson correlation coefficients between all indicator pairs are stored in a matrix, forming a single-window linear correlation matrix. This matrix is ​​a symmetric matrix, with each diagonal element being 1.

[0162] The single-window linear correlation matrices calculated for each single-window deviation value matrix are arranged in chronological order to form a sequence of linear correlation matrices. This sequence contains information on the linear correlation between various indicators within each time window.

[0163] The deviation values ​​of each indicator in the single-window deviation value matrix are discretized. For example, using the equal-frequency binning method, the deviation values ​​of each indicator are divided into three intervals: high, medium, and low. Specifically, the deviation values ​​of each indicator are sorted, and then the deviation values ​​are evenly divided into three intervals. The deviation value in each interval is represented by a corresponding discrete value; for example, the high interval is represented by "2", the medium interval by "1", and the low interval by "0". The discretized values ​​replace the original deviation values ​​to obtain the discretized deviation value matrix.

[0164] Based on the discretized deviation value matrix, the joint probability distribution between every two indicators is calculated. For example, the joint probability of heart rate deviation and blood pressure deviation falling into the "high," "medium," and "low" intervals is calculated. For example, P(heart rate = high, blood pressure = low) represents the probability that the heart rate deviation is in the "high" interval and the blood pressure deviation is in the "low" interval. The joint probability distributions of all indicator pairs are stored in a matrix to form a joint probability distribution matrix. Simultaneously, the univariate probability distribution of each indicator is calculated, i.e., the probability that each indicator falls into the "high," "medium," or "low" intervals. For example, P(heart rate = high) represents the probability that the heart rate deviation is in the "high" interval.

[0165] Based on the joint probability distribution matrix and univariate probability distribution data obtained in the previous step, calculate the mutual information value between each pair of indicators. The formula for calculating the mutual information value is: ΣP(X,Y)*log(P(X,Y) / (P(X)*P(Y))), where X and Y represent the discrete values ​​of the two indicators, and "Σ" represents the summation over all possible X and Y values. The larger the mutual information value, the stronger the nonlinear correlation between the two indicators. Store the mutual information values ​​between all indicator pairs in a matrix to form a single-window mutual information matrix.

[0166] The single-window mutual information matrices calculated for each single-window deviation value matrix are arranged in chronological order to form a sequence of nonlinear correlation matrices. This sequence contains nonlinear correlation information between various indicators within each time window.

[0167] Preferably, step S4 includes the following steps:

[0168] Step S41: Obtain patient's historical health data; integrate the deviation assessment report and patient's historical health data to obtain an integrated dataset;

[0169] Step S42: Extract deviation features from the integrated dataset to obtain a deviation feature table;

[0170] Step S43: Perform historical data feature processing on the integrated dataset to obtain a historical feature table;

[0171] Step S44: Extract environmental features from the integrated dataset to obtain an environmental feature table;

[0172] Step S45: Construct a high-order combined feature table based on the deviation feature table, historical feature table, and environmental feature table;

[0173] Step S46: Select predictive features from the deviation feature table, historical feature table, environmental feature table, and combined feature table to obtain the predictive feature set;

[0174] Step S47: Construct a risk prediction model based on the prediction feature set and the preset labeled dataset to obtain the risk prediction model; use the risk prediction model to make risk predictions on standardized real-time data and generate a risk prediction report to obtain the risk prediction report.

[0175] Step S48: Classify the risk levels according to the risk prediction report to obtain a risk level list; trigger the early warning and send the information according to the risk level list and the preset early warning rule base to obtain the early warning information.

[0176] As an example of the present invention, reference is made to... Figure 3 As shown, step S4 in this example includes:

[0177] Step S41: Obtain patient's historical health data; integrate the deviation assessment report and patient's historical health data to obtain an integrated dataset;

[0178] In this embodiment of the invention, historical health data of patients are obtained from electronic health record (EHR) systems, hospital information systems (HIS), or other medical databases, including past medical history, diagnostic information, medication records, allergy history, family history, and laboratory test results. This historical health data is then integrated with the data in the deviation assessment report (DAR) generated in step S3. During the integration process, the DAR data and historical health data are aligned according to timestamps, and data from different sources are converted to a unified format and units. The integrated data is stored in an integrated dataset, containing the patient's real-time physiological data deviation information, historical health information, and other relevant information.

[0179] Step S42: Extract deviation features from the integrated dataset to obtain a deviation feature table;

[0180] In this embodiment of the invention, features related to deviation in the integrated dataset are analyzed to construct a deviation feature table. For different time windows (e.g., the past 24 hours, the past 7 days, and the past 30 days), statistical characteristics of the deviation value sequence for each indicator are calculated, such as mean, standard deviation, maximum value, minimum value, median, skewness, kurtosis, and specific quantile values. Furthermore, the changing trend of the deviation value sequence is calculated, such as slope and rate of change. These statistical and trend characteristics are stored as deviation features in the deviation feature table.

[0181] Step S43: Perform historical data feature processing on the integrated dataset to obtain a historical feature table;

[0182] In this embodiment of the invention, feature processing is performed on the historical health data of patients in the integrated dataset to construct a historical feature table. Information such as the patient's past medical history, medication records, and allergy history is converted into structured features. For example, one-hot encoding is used to convert disease diagnosis information into binary feature vectors, where 1 indicates the patient has the disease and 0 indicates the patient does not have the disease. The patient's medication records are converted into features of drug dosage and administration time. The patient's allergy history is converted into binary feature vectors of allergens. The processed historical data features are stored in the historical feature table.

[0183] Step S44: Extract environmental features from the integrated dataset to obtain an environmental feature table;

[0184] In this embodiment of the invention, environmental features relevant to patients are extracted from the integrated dataset to construct an environmental feature table. For different time windows (e.g., the past 24 hours, the past 7 days, and the past 30 days), statistical features of the environmental data are calculated, such as average temperature, maximum temperature, minimum temperature, average humidity, and average light intensity. Furthermore, the changing trends of the environmental data are calculated, such as the rate of temperature change and the rate of humidity change. These statistical features and trend features are stored as environmental features in the environmental feature table.

[0185] Step S45: Construct a high-order combined feature table based on the deviation feature table, historical feature table, and environmental feature table;

[0186] In this embodiment of the invention, high-order combined features are constructed based on a deviation feature table, a historical feature table, and an environmental feature table to capture the interactions between different features. For example, a new combined feature is constructed by multiplying the mean of the heart rate deviation values ​​by the average temperature. Alternatively, another combined feature is constructed by calculating the ratio of the standard deviation of the blood pressure deviation values ​​to the average activity level over the past 7 days. The constructed high-order combined features are stored in the combined feature table. High-order combined features can be automatically constructed using deep feature synthesis technology.

[0187] Step S46: Select predictive features from the deviation feature table, historical feature table, environmental feature table, and combined feature table to obtain the predictive feature set;

[0188] In this embodiment of the invention, feature selection is performed on all features in the deviation feature table, historical feature table, environmental feature table, and combined feature table to remove redundant and irrelevant features, thereby improving the efficiency and generalization ability of the model. A wrapper-based feature selection method, such as Recursive Feature Eradication (RFE), is employed, combined with a machine learning model (e.g., Random Forest) to evaluate the importance of features and progressively eliminate features with low importance. Finally, the subset of features with the highest predictive power is selected to constitute the prediction feature set.

[0189] Step S47: Construct a risk prediction model based on the prediction feature set and the preset labeled dataset to obtain the risk prediction model; use the risk prediction model to make risk predictions on standardized real-time data and generate a risk prediction report to obtain the risk prediction report.

[0190] In this embodiment of the invention, a risk prediction model is trained using a pre-defined labeled dataset (containing patient characteristics and corresponding risk labels) and a prediction feature set. The labeled dataset is obtained through retrospective analysis of historical case data. A suitable machine learning model, such as random forest, support vector machine, or gradient boosting decision tree (GBDT), is selected and trained using the labeled dataset. After training, the trained risk prediction model is used to predict the risk of patients based on standardized real-time data, predicting the probability of a patient developing a specific health risk (e.g., heart failure, fall, infection, etc.) in the future. The prediction results, along with other relevant information such as the patient's personal information, current physiological indicators, and the type and probability of the predicted risk, are integrated into a Risk Prediction Report (RPR). The RPR presents the prediction results in a clear and easy-to-understand manner, such as using charts, tables, and other visualizations, to facilitate healthcare professionals' quick understanding of the patient's risk status.

[0191] Step S48: Classify the risk levels according to the risk prediction report to obtain a risk level list; trigger and send early warnings according to the risk level list and the preset early warning rule base to obtain early warning information;

[0192] In this embodiment of the invention, patients' risks are classified into different levels, such as low risk, medium risk, and high risk, based on the risk probability predicted in the Risk Prediction Report (RPR). The threshold for classifying risk levels can be adjusted according to specific application scenarios and risk types. For example, patients with a heart failure risk probability below 0.1 are classified as low risk, those with a probability between 0.1 and 0.5 are classified as medium risk, and those with a probability above 0.5 are classified as high risk. Patient risk level information is stored in a risk level list. Then, based on the risk level list and a preset warning rule base, a corresponding warning mechanism is triggered. The warning rule base predefines the warning methods and information content corresponding to different risk levels. For example, for high-risk patients, the system will immediately send warning information to doctors and family members; for medium-risk patients, the system will periodically send reminder information to doctors, suggesting enhanced monitoring; for low-risk patients, the system will not send warning information. Warning information can be sent through various methods such as SMS, email, and app notifications.

[0193] Preferably, step S42 includes the following steps:

[0194] Step S421: Define time windows of different lengths for bias feature extraction based on the integrated dataset to obtain a list of bias time windows;

[0195] Step S422: Group the integrated dataset into deviation scores based on time window grouping according to the deviation time window list to obtain the time window deviation score set;

[0196] Step S423: Calculate the statistical characteristics of the time window deviation on the time window deviation fraction set to obtain the statistical characteristic set of the time window deviation;

[0197] Step S424: Calculate the time window deviation trend characteristics of the time window deviation fraction set to obtain the time window deviation trend characteristic set;

[0198] Step S425: Merge the time window deviation statistical feature set and the time window deviation trend feature set to obtain the deviation feature table.

[0199] In this embodiment of the invention, time windows of different lengths are defined based on the characteristics of the integrated dataset and the analytical requirements to extract deviation features. For example, three time windows are defined: 24 hours, 7 days, and 30 days. The length values ​​of these time windows are stored in a deviation time window list, for example, deviation time window list = [24, 168, 720], in hours. The selection of these time windows is based on clinical experience and the needs of data analysis; for example, a 24-hour window can reflect short-term changes in patients, while a 30-day window can reflect long-term trends.

[0200] Based on the list of deviation time windows, the deviation assessment scores in the integrated dataset are grouped according to time windows. For example, for a 24-hour time window, the deviation assessment scores of the past 24 hours are grouped together; for a 7-day time window, the deviation assessment scores of the past 7 days are grouped together; and for a 30-day time window, the deviation assessment scores of the past 30 days are grouped together. Each time window corresponds to a set of deviation assessment scores, and these sets constitute the time window deviation score set.

[0201] For each time window's set of deviation scores, calculate the corresponding statistical characteristics. For example, calculate the mean, standard deviation, maximum, minimum, median, skewness, kurtosis, and 25th and 75th percentiles of the deviation assessment scores within each time window. Store the calculated statistical characteristics in the time window deviation statistical characteristic set. Each time window corresponds to a set of statistical characteristics.

[0202] For each time window's set of deviation scores, calculate the corresponding trend features. For example, use linear regression to fit the deviation assessment scores within each time window, calculate the slope of the regression line to represent the trend of change in the deviation assessment scores. Alternatively, calculate the rate of change of deviation assessment scores within a time window, such as (last value - first value) / first value. Store the calculated trend features in the time window deviation trend feature set. Each time window corresponds to a set of trend features.

[0203] The statistical feature set and trend feature set of time window deviation are merged to form a deviation feature table. Each row in the deviation feature table represents the deviation feature of a time window, and each column represents a feature, such as 24-hour mean, 7-day standard deviation, 30-day slope, etc.

[0204] Preferably, step S44 includes the following steps:

[0205] Step S441: Define time windows of different lengths for environmental feature extraction based on the integrated dataset to obtain a list of environmental time windows;

[0206] Step S442: Group the integrated dataset into environmental data based on time window grouping according to the environmental time window list to obtain the time window environmental dataset;

[0207] Step S443: Perform time window environmental statistical features on the time window deviation score set to obtain the time window environmental statistical feature set;

[0208] Step S444: Calculate the environmental change features of the time window based on the environmental statistical feature set and the environmental dataset of the time window to obtain the environmental change feature set of the time window;

[0209] Step S445: Merge the time window environment statistical feature set and the time window environment change feature set to obtain the environment feature table.

[0210] In this embodiment of the invention, time windows of different lengths are defined based on the characteristics and analytical needs of the integrated dataset to extract environmental features. The setting of the environmental time window should match the cycle of changes in the patient's physiological data and the cycle of influence of environmental factors. For example, three time windows are defined: 24 hours, 7 days, and 30 days, to capture short-term, medium-term, and long-term environmental change trends. The length values ​​of these time windows are stored in an environmental time window list, for example, environmental time window list = [24, 168, 720], in hours.

[0211] Based on the list of environmental time windows, the environmental data in the integrated dataset, such as temperature, humidity, light intensity, and noise, are grouped according to time windows. For example, temperature and humidity data from the past 24 hours are grouped into one group; temperature and humidity data from the past 7 days are grouped into another group; and temperature and humidity data from the past 30 days are grouped into yet another group. Each time window corresponds to an environmental dataset containing all environmental data within that time window. These datasets constitute the time window environmental dataset.

[0212] For each time window's environmental dataset, calculate the corresponding environmental statistical features. For example, calculate the mean, maximum, minimum, and standard deviation of temperature; the mean, maximum, minimum, and standard deviation of humidity; the mean and standard deviation of light intensity; and the mean and standard deviation of noise, etc. Store the calculated statistical features in a time window environmental statistical feature set. Each time window corresponds to a set of environmental statistical features.

[0213] Beyond statistical characteristics, further analysis of environmental data trends is conducted. For example, the rate of temperature change within each time window can be calculated as (last temperature value - first temperature value) / first temperature value; the rate of humidity change can also be calculated; and the rate of light intensity change can be calculated. Linear regression can also be used to fit environmental data, such as temperature, within each time window, and the slope of the regression line can be calculated to represent the temperature trend. The calculated trend features are stored in a time window environmental change feature set. Each time window corresponds to a set of environmental change features.

[0214] The environmental statistical feature set of the time window and the environmental change feature set of the time window calculated in step S444 are combined to form an environmental feature table. Each row in the environmental feature table represents the environmental features of a time window, and each column represents a feature, such as 24-hour average temperature, 7-day temperature change rate, 30-day average humidity, etc.

[0215] Preferably, step S45 includes the following steps:

[0216] Step S451: Prepare the feature matrix for the deviation feature table, historical feature table, and environmental feature table to obtain the feature matrix;

[0217] Step S452: Generate basic combined features from the feature matrix to obtain the basic combined feature set;

[0218] Step S453: Use deep feature synthesis technology to synthesize deep features from the basic combined feature set to obtain deep features;

[0219] Step S454: Optimize the deep features using a genetic algorithm to obtain an optimized feature set;

[0220] Step S455: Generate a combined feature table based on the optimized combined feature set to obtain the combined feature table.

[0221] In this embodiment of the invention, the feature data from the deviation feature table, the historical feature table, and the environmental feature table are merged into a feature matrix. Each row of the feature matrix represents a sample (e.g., a point in time or a time period), and each column represents a feature. For example, the first column could be the 24-hour average heart rate deviation, the second column could be the 7-day maximum blood pressure deviation, the third column could be whether there is a history of hypertension, and the fourth column could be the 24-hour average temperature, etc.

[0222] Based on the feature matrix, basic combined features are generated. Basic combined features are new features obtained by performing simple mathematical operations (such as addition, subtraction, multiplication, division, squaring, square root, etc.) or logical operations (such as AND, OR, NOT) on the original features. For example, multiplying the 24-hour average heart rate deviation by the 24-hour average temperature yields a new combined feature representing the interaction between heart rate deviation and temperature. All generated combined features are stored in the basic combined feature set.

[0223] Deep feature synthesis (DFS) technology can be used to automatically generate more complex deep features from a basic set of combined features. DFS is a graph-based feature generation method that iteratively applies predefined transformation operators (such as addition, subtraction, multiplication, division, average, maximum, minimum, etc.) to the original and generated features to generate new features. DFS can automatically explore the feature space and generate deep features with stronger expressive power. For example, DFS can first generate a product feature of "24-hour heart rate deviation mean" and "7-day average temperature", and then perform a logical operation on this feature with "history of hypertension" to generate a new deep feature.

[0224] Genetic algorithms are used to optimize deep features, selecting the most effective feature subset. A genetic algorithm is an optimization algorithm that simulates the process of natural selection. It progressively optimizes feature subsets by simulating operations such as gene crossover, mutation, and selection. Specifically, deep features are used as genes to construct a population of feature subsets. Then, the fitness of each feature subset is evaluated by calculating its performance (e.g., AUC value) on a risk prediction model. Feature subsets with high fitness are selected for crossover and mutation operations to generate new feature subsets. This process is iteratively performed until the optimal feature subset is found. The optimal feature subset is then stored in an optimized feature set.

[0225] The features from the optimized combined feature set, along with the original features, are stored together in the combined feature table. The combined feature table contains the original features and the combined features optimized by deep feature synthesis and genetic algorithms, which are used for subsequent risk prediction model training.

[0226] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0227] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A remote care management method based on data feature mining, characterized by, The method comprises the following steps: Step S1: obtaining patient-specific condition and demand data; selecting and configuring data sources according to the patient-specific condition and demand data to obtain a data source configuration table; According to the data source configuration table, multi-dimensional data acquisition is performed, and multi-dimensional data preprocessing is performed to obtain patient comprehensive data; Step S2: individual baseline modeling is performed on the patient comprehensive data to obtain an individual baseline model; Step S3: real-time data standardization is performed on the patient comprehensive data to obtain standardized real-time data; deviation calculation is performed according to the standardized real-time data and the individual baseline model to obtain a deviation value sequence; index correlation analysis is performed on the deviation value sequence to obtain an index correlation matrix; deviation comprehensive evaluation analysis is performed according to the index correlation matrix and the deviation value sequence to obtain a deviation evaluation report; Step S4: obtaining patient historical health data; data integration is performed on the deviation evaluation report and the patient historical health data to obtain an integrated data set; deviation feature extraction, historical data feature processing and environment feature extraction are performed on the integrated data set respectively to obtain a deviation feature table, a historical feature table and an environment feature table; high-order combined feature construction is performed according to the deviation feature table, the historical feature table and the environment feature table to obtain a combined feature table; prediction feature selection is performed on the deviation feature table, the historical feature table, the environment feature table and the combined feature table to obtain a prediction feature set; a risk prediction model is constructed according to the prediction feature set to obtain a risk prediction model; risk prediction processing is performed by using the risk prediction model to obtain a risk prediction report; early warning triggering and information sending are performed according to the risk prediction report to obtain early warning information, so as to realize the remote nursing management task.

2. The data-feature-mining-based tele-nursing management method according to claim 1, characterized by, Step S1 comprises the following steps: Step S11: obtaining patient-specific condition and demand data; selecting and configuring data sources according to the patient-specific condition and demand data to obtain a data source configuration table; Step S12: data real-time acquisition is performed according to the data source configuration table to obtain an original data stream; Step S13: data preprocessing is performed on the original data stream to obtain a preprocessed data stream; Step S14: data feature extraction is performed on the preprocessed data stream to obtain a feature data stream; Step S15: data integration is performed on the feature data stream to obtain patient comprehensive data.

3. The data-feature-mining-based tele-nursing management method according to claim 1, characterized by, Step S2 comprises the following steps: Step S21: data segmentation is performed on the patient comprehensive data to obtain a training data set and a verification data set; Step S22: feature selection is performed on the training data set to obtain a filtered feature set; Step S23: baseline model training is performed according to the training data set and the filtered feature set to obtain an initial baseline model; Step S24: model verification and optimization are performed on the initial baseline model by using the verification data set to obtain an optimized baseline model; Step S25: model updating is performed on the optimized baseline model according to the patient comprehensive data to obtain an individual baseline model.

4. The data-feature-mining-based tele-nursing management method according to claim 1, characterized by, Step S3 comprises the following steps: Step S31: real-time data standardization is performed on the patient comprehensive data to obtain standardized real-time data; baseline distribution parameter extraction is performed on the individual baseline model to obtain baseline distribution parameters; Step S32: Real-time data probability density calculation is performed according to the baseline distribution parameters and the standardized real-time data, and real-time data probability density is obtained; Step S33: Probability density conversion is performed on the real-time data probability density, and a deviation value is obtained; time sequence sorting is performed on the deviation value, and a deviation value sequence is obtained; Step S34: Index correlation analysis is performed on the deviation value sequence, and an index correlation matrix is obtained; Step S35: Deviation comprehensive evaluation is performed according to the index correlation matrix and the deviation value sequence, and a deviation evaluation score is obtained; deviation evaluation report generation is performed according to the deviation evaluation score and the deviation value sequence, and a deviation evaluation report is obtained.

5. The data-feature-mining-based tele-nursing management method according to claim 4, characterized by, Step S34 includes the following steps: Step S341: Multi-dimensional time sequence matrix construction is performed on the deviation value sequence, and a deviation value matrix is obtained; Step S342: Time window segmentation is performed on the deviation value matrix, and a deviation value window sequence is obtained; Step S343: Single-window deviation matrix extraction is performed on the deviation value window sequence, and a single-window deviation value matrix is obtained; Step S344: Linear correlation calculation between indexes is performed on the single-window deviation value matrix, and a linear correlation matrix sequence is obtained; nonlinear correlation calculation between indexes is performed on the single-window deviation value matrix, and a nonlinear correlation matrix sequence is obtained; Step S345: Linear and nonlinear correlation fusion is performed on the linear correlation matrix sequence and the nonlinear correlation matrix sequence, and a comprehensive correlation matrix sequence is obtained; Step S346: Time window comprehensive analysis is performed on the comprehensive correlation matrix sequence, and an index correlation matrix is obtained.

6. The data-feature-mining-based telehealth management method according to claim 5, characterized by, Step S344 specifically includes the following steps: Linear correlation calculation between two indexes within a single window is performed on the single-window deviation value matrix, and a single-window linear correlation matrix is obtained; Linear correlation matrix aggregation is performed on the single-window linear correlation matrix, and a linear correlation matrix sequence is obtained; Single-window deviation value discretization is performed on the single-window deviation value matrix, and a discretized deviation value matrix is obtained; Joint probability distribution calculation is performed according to the discretized deviation value matrix, and a joint probability distribution matrix is obtained; single-variable probability distribution calculation is performed according to the discretized deviation value matrix, and single-variable probability distribution data is obtained; Mutual information value calculation is performed according to the joint probability distribution matrix and the single-variable probability distribution data, and a single-window mutual information matrix is obtained; Nonlinear correlation matrix sequence construction is performed according to the deviation value window sequence and the single-window mutual information matrix, and a nonlinear correlation matrix sequence is obtained.

7. The data-feature-mining-based telehealth management method according to claim 1, characterized by, Step S4 includes the following steps: Step S41: Obtain patient historical health data; data integration is performed on the deviation evaluation report and the patient historical health data, and an integrated data set is obtained; Step S42: Deviation feature extraction is performed on the integrated data set, and a deviation feature table is obtained; Step S43: Historical data feature processing is performed on the integrated data set, and a historical feature table is obtained; Step S44: Environment feature extraction is performed on the integrated data set, and an environment feature table is obtained; Step S45: High-order combined feature construction is performed according to the deviation feature table, the historical feature table, and the environment feature table, and a combined feature table is obtained; Step S46: performing prediction feature selection on the deviation feature table, the historical feature table, the environment feature table, and the combined feature table to obtain a prediction feature set; Step S47: performing risk prediction model construction according to the prediction feature set and a preset labeled data set to obtain a risk prediction model; performing risk prediction on standardized real-time data by using the risk prediction model, and generating a risk prediction report to obtain a risk prediction report; Step S48: performing risk level division according to the risk prediction report to obtain a risk level list; performing early warning triggering and information sending according to the risk level list and a preset early warning rule library to obtain early warning information.

8. The data-feature-mining-based telehealth management method according to claim 7, characterized by, Step S42 includes the following steps: Step S421: defining different length time windows for deviation feature extraction according to the integrated data set to obtain a deviation time window list; Step S422: grouping deviation scores based on time window grouping according to the deviation time window list to obtain a time window deviation score set; Step S423: performing time window deviation statistical feature calculation on the time window deviation score set to obtain a time window deviation statistical feature set; Step S424: performing time window deviation trend feature calculation on the time window deviation score set to obtain a time window deviation trend feature set; Step S425: performing window deviation feature merging on the time window deviation statistical feature set and the time window deviation trend feature set to obtain a deviation feature table.

9. The data-feature-mining-based telehealth management method according to claim 7, characterized by, Step S44 includes the following steps: Step S441: defining different length time windows for environment feature extraction according to the integrated data set to obtain an environment time window list; Step S442: grouping environment data based on time window grouping according to the environment time window list to obtain a time window environment data set; Step S443: performing time window environment statistical feature calculation on the time window deviation score set to obtain a time window environment statistical feature set; Step S444: performing time window environment change feature calculation according to the time window environment statistical feature set and the time window environment data set to obtain a time window environment change feature set; Step S445: performing window environment feature merging on the time window environment statistical feature set and the time window environment change feature set to obtain an environment feature table.

10. The data-feature-mining-based telehealth management method of claim 7, wherein, Step S45 includes the following steps: Step S451: performing feature matrix preparation on the deviation feature table, the historical feature table, and the environment feature table to obtain a feature matrix; Step S452: performing basic combined feature generation on the feature matrix to obtain a basic combined feature set; Step S453: performing deep feature synthesis on the basic combined feature set by using a deep feature synthesis technology to obtain deep features; Step S454: performing genetic algorithm optimization on the deep features to obtain an optimized combined feature set; Step S455: generating a combined feature table according to the optimized combined feature set to obtain a combined feature table.

Images