Health state monitoring method and system based on big data multi-dimensional evaluation

Abstract

The invention discloses a health state monitoring method and system based on big data multi-dimensional evaluation, and relates to the cross technical field of health monitoring, big data processing and quantum machine learning. Through multi-source health data collection and in combination with quantization coding and high-dimensional data dimension reduction technologies, core low-dimensional feature vectors are effectively extracted, further, nonlinear correlation features in the feature vectors are deeply mined through a quantum neural network model, comprehensive and accurate evaluation of the health state of the user is achieved, and the user health state evaluation accuracy is improved. The multi-dimensional and deep evaluation mode can reflect the health state of the user more accurately, and provides a more reliable basis for subsequent early warning and intervention.

Description

Technical Field

[0001] This invention relates to the interdisciplinary fields of health monitoring, big data processing, and quantum machine learning, specifically a health status monitoring method and system based on multi-dimensional big data assessment. Background Technology

[0002] With the increasing awareness of health and the advancement of technology, health monitoring has become an important means of preventing diseases and improving quality of life. Traditional health monitoring mainly relies on a single data source or limited-dimensional data collection, such as simple physiological indicator measurements or questionnaires. These methods are difficult to comprehensively and accurately reflect an individual's health status.

[0003] Traditional health status monitoring methods have several limitations: First, they rely on a single data source and lack comprehensive analysis of multi-dimensional data, leading to biased assessment results that fail to accurately reflect an individual's true health status. Second, they have limited data processing capabilities; when faced with high-dimensional and complex health data, traditional methods often struggle to effectively extract core features, resulting in information loss and inaccurate assessments. Third, they lack adaptive optimization mechanisms; traditional models struggle to dynamically adjust parameters to maintain assessment accuracy in the face of individual differences and changes in health status.

[0004] In view of the limitations of traditional health status monitoring methods, the present invention proposes a health status monitoring method and system based on big data multi-dimensional assessment, which is of particular importance. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a health status monitoring method and system based on big data multi-dimensional assessment. It effectively extracts core low-dimensional feature vectors through multi-source health data acquisition and preprocessing, combined with quantum coding and high-dimensional data dimensionality reduction techniques. Furthermore, it deeply mines the nonlinear correlation features in the feature vectors using a quantum neural network model, achieving a comprehensive and accurate assessment of the user's health status. Simultaneously, the system incorporates a model optimization and dynamic update mechanism, which can dynamically adjust model parameters based on user health status feedback data, ensuring long-term accuracy and stability of the assessment. Moreover, by optimizing the model structure and employing a quantum state compression algorithm, the system achieves efficient operation at the edge, meeting the needs of real-time health monitoring.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: On one hand, a health status monitoring method based on multi-dimensional assessment using big data, the specific steps of which are as follows: S1. Multi-source health data acquisition and preprocessing: Acquire users' physiological indicator data, gene data, environmental parameter data and behavioral data through multi-source data acquisition equipment, and obtain standardized high-dimensional health datasets through cleaning, outlier correction and format standardization. S2, Quantum Coding and Dimensionality Reduction of High-Dimensional Data: The standardized high-dimensional health dataset is mapped to quantum superposition state data through a quantum bit coding algorithm. A quantum dimensionality reduction model is constructed using the quantum entanglement property. The feature space of the quantum superposition state data is compressed to extract the core low-dimensional feature vector. S3. Quantum Neural Network Feature Mining: Input low-dimensional feature vectors into a pre-trained quantum neural network model to mine the hidden nonlinear correlation features in the feature vectors and output a deep feature matrix. S4. Health Status Assessment and Risk Warning: Calculate the user's health status score and the contribution of each risk factor based on the deep feature matrix. When the warning conditions are met, generate targeted warning information and intervention suggestions. S5. Model Optimization and Dynamic Update: Based on user health status feedback data, dynamically adjust the parameters of the quantum dimensionality reduction model and the quantum neural network model.

[0007] Furthermore, the synchronization guarantee process for multi-source data acquisition in S1 is as follows: First, all data acquisition devices are calibrated with the system clock using the Network Time Protocol (NTP) every 5 minutes to ensure that the timestamp error of the data acquired by each device does not exceed 10ms. Second, for wearable devices and environmental sensor real-time acquisition devices, a local caching method at the edge and timed upload is adopted, with a caching period of 1 minute. The upload includes a timestamp, device identifier, and data verification code, and data real-time performance is ensured through 5G / Wi-Fi dual-mode transmission. For non-real-time data such as gene testing devices and medical examination records, a batch upload and timestamp completion mechanism is adopted. The accurate timestamp is completed according to the acquisition period corresponding to the data, and it is matched with the real-time data of the same period through a data association algorithm. Then, when the system receives data, it verifies the integrity of data transmission through a CRC32 checksum. If the verification fails, the device is triggered to re-upload. Finally, a data synchronization verification module is constructed to perform logical consistency verification on multi-source data in the same period. Data that passes the verification is stored in a standardized dataset, and data that fails the verification is marked as suspicious data and stored separately to ensure the time synchronization, integrity, and logical consistency of the acquired data, providing a reliable data foundation for subsequent processing.

[0008] Furthermore, the specific implementation process of data preprocessing in S1 is as follows: First, an improved 3σ criterion is used to identify outliers in the original data. For numerical data such as physiological indicator data and environmental parameter data, the mean μ and standard deviation σ are calculated. Data exceeding the interval [μ-3.5σ, μ+3.5σ] are marked as outliers. For continuous time-series outliers, a correction method based on local weighted regression is used for smoothing. For isolated outliers, the statistical median of the same type of data is used for replacement. Second, for missing values, if the missing rate is less than 5%, an adaptation method is selected according to the data type: physiological time-series data... Linear interpolation based on adjacent time-time data is used to supplement the data, while mode imputation based on historical similar scenarios is used to supplement the behavioral classification data. If the missing rate is higher than 5%, the secondary data acquisition mechanism of the data acquisition device is activated, and the target device is triggered to re-acquire the corresponding time period data through the device wake-up command. Finally, the format is standardized, the base sequence of the gene data is converted into numerical code, and all heterogeneous data are uniformly mapped to standardized values ​​in the range [0, 1]. Data type verification algorithm is used to ensure that the output standardized high-dimensional health dataset has no format conflicts and no logical contradictions, providing high-quality data input for subsequent quantization processing.

[0009] Furthermore, the qubit encoding algorithm in S2 employs a data reliability-weighted quantum mapping algorithm, with the following formula: ,in For the first The quantum superposition state corresponding to each classical data point These are the standardized classic data values. , These are the global minimum and maximum values ​​for this type of data, respectively. The data credibility weight is determined by a weighted sum of three factors: the accuracy level of the data acquisition device, the bit error rate during data transmission, and the stability of the data acquisition scenario. This formula integrates data credibility into the quantum coding process, ensuring that high-credibility data occupies a higher weight in the quantum state, thereby improving the accuracy of subsequent feature extraction. Unlike existing quantum coding, which only focuses on the numerical mapping of data, this formula achieves a deep binding between the coding process and data quality.

[0010] Furthermore, the quantum dimensionality reduction model in S2 is constructed based on the quantum entanglement-enhanced principal component analysis algorithm, and the formula is: ,in For the first The quantum state projection values ​​of a low-dimensional eigenvector. The number of qubits in the original high-dimensional data. For the first The density matrix of qubits, For the first The quantum unitary transformation matrix corresponding to each principal component The quantum entanglement weight coefficients are determined as follows: the correlation strength between each qubit is calculated by quantum entanglement entropy, and combined with the physiological / environmental / genetic priority of the data type, the priority is preset to 0.4 for physiological data, 0.3 for genetic data, 0.2 for environmental data, and 0.1 for behavioral data. The gradient descent algorithm is used to optimize and obtain the weight coefficients that maximize the feature variance after dimensionality reduction. This formula breaks through the limitation of traditional quantum principal component analysis relying only on linear projection. By integrating data correlation characteristics and priorities through quantum entanglement weights, it achieves dimensionality compression while retaining the core features that are key to health assessment to the greatest extent. The computational efficiency is improved by more than 300% compared with traditional classical dimensionality reduction algorithms.

[0011] Furthermore, the pre-training process of the quantum neural network model in S3 is as follows: First, a hybrid training set of a large-scale public health dataset and a private health dataset is constructed. The public dataset contains multi-source health data from over 100,000 users, while the private dataset contains historical health data of the target user group. Second, a federated learning architecture is used for pre-training. The public dataset is deployed on a cloud server, and the private dataset is deployed on a local edge node. Collaborative training of the model is achieved through parameter federation and sharing. During training, only the gradients of the model parameters are transmitted, without leaking the original data. Then, the pre-training is divided into two stages: the first stage trains a basic model based on the public dataset to learn the general characteristics and patterns of health data; the second stage fine-tunes the basic model based on the private dataset to adapt to the feature distribution of the target user group. Finally, after pre-training is completed, the model is lightweighted using a quantum state compression algorithm to remove redundant parameters, ensuring that the model can run efficiently at the edge with an inference latency controlled within 100ms to meet the needs of real-time health monitoring. At the same time, the federated learning architecture ensures data privacy and security, avoiding the risk of leakage of sensitive health data.

[0012] Furthermore, the specific implementation process of the quantum neural network model in S3 is as follows: First, a three-layer network architecture of quantum entanglement layer - quantum convolution layer - measurement layer is constructed. The quantum entanglement layer uses alternating combinations of two-qubit CNOT gates and three-qubit Tooffoli gates to construct multi-order correlation mappings between features. Each layer contains 8-12 quantum gate combination units, and the number of units is adaptively adjusted according to the dimension of the low-dimensional feature vector. Second, the quantum convolution layer adopts a quantum convolution kernel design, and frequency domain extraction of features is achieved through quantum Fourier transform. The size of the convolution kernel is dynamically set to 2×2 or 3×3 according to the dimension of the feature vector, and the convolution stride is set to 1. Parallel convolution calculation is achieved using the quantum state superposition property. In the post-model training process, quantum state fidelity is used as the loss function. That is, the fidelity value between the model output quantum state and the labeled quantum state is calculated, and the quantum gate parameters are adjusted by quantum gradient descent algorithm. The number of training iterations is set to 100-200 times, and training stops when the fidelity value is higher than 0.95. Finally, the measurement layer converts the quantum state data into a classical deep feature matrix through quantum measurement operators. During the measurement process, multiple measurements are taken and averaged to reduce quantum measurement noise and ensure the stability and accuracy of the output feature matrix. Compared with traditional quantum neural networks, this network architecture improves the feature mining capability by more than 40% and can effectively capture complex nonlinear correlation features such as gene-environment and physiological-behavioral relationships.

[0013] Furthermore, the specific calculation process for the health status score in S4 is as follows: First, an evaluation index system containing 5 primary indicators and 20 secondary indicators is constructed, with each secondary indicator corresponding to a specific feature dimension in the deep feature matrix; second, the weights of the primary indicators are determined using the analytic hierarchy process (AHP): physiological function 0.3, genetic risk 0.25, environmental exposure 0.2, behavioral habits 0.15, and psychological state 0.1. The dynamic weights of the secondary indicators are determined using the entropy weight method, where the entropy weights are calculated based on the information entropy of the current batch of data to ensure that the weights are adapted to the data distribution characteristics; then, the weights of each secondary indicator are calculated... The scoring of indicators: the scores of secondary indicators are converted into quantitative values ​​of 0-100 by comparing the feature values ​​with preset thresholds. The scores of primary indicators are obtained by weighted summation of the scores of secondary indicators and the corresponding entropy weights. The total health status score is obtained by weighted summation of the scores of primary indicators and the weights of the analytic hierarchy process. Finally, the contribution of risk factors is calculated using the elasticity coefficient method, that is, the degree of influence of small changes in each feature dimension on the total health score is calculated. After ranking, the top 5 features with the highest degree of influence are selected as core risk factors to clarify the contribution ratio of each factor and provide precise targeting basis for subsequent early warning and intervention.

[0014] Furthermore, the specific formula for dynamically adjusting the model parameters in S5 is as follows: ,in For the updated model parameters, These are the current model parameters. The learning rate is dynamically set to 0.001-0.01, and adaptively decreases based on the number of training iterations. The gradient of the loss function. The feedback data validity coefficient is determined by multiplying the degree of implementation of the user intervention plan, the significance of changes in health indicators, and the completeness of the feedback data. This formula incorporates the validity assessment of feedback data to avoid model parameter drift caused by invalid feedback, ensuring that the model continuously adapts to changes in the individual user's health status and improving the accuracy and stability of long-term assessment.

[0015] On the other hand, a health status monitoring system based on big data multi-dimensional assessment includes: a multi-source data acquisition module, a data preprocessing module, a quantum computing processing module, a health assessment and early warning module, a model optimization module, and a data storage and interaction module; The multi-source data acquisition module consists of a wearable physiological monitoring device, a gene detection module, an environmental sensor array, and a behavior recording terminal, which simultaneously collects multi-dimensional health-related data. The data preprocessing module is connected to the multi-source data acquisition module to clean, correct outliers, and standardize the format of the raw acquired data, and output a standardized high-dimensional health dataset. The quantum computing processing module includes a quantum encoding unit, a quantum dimensionality reduction unit, and a quantum neural network unit. The quantum encoding unit is used to map classical high-dimensional data into quantum superposition state data. The quantum dimensionality reduction unit is used to efficiently reduce the dimensionality of quantum superposition state data by utilizing the properties of quantum entanglement and extract core low-dimensional feature vectors. The quantum neural network unit is used to mine nonlinear correlation features in low-dimensional feature vectors and output a deep feature matrix. The health assessment and early warning module is connected to the quantum computing processing module. It has a built-in health status assessment index system and early warning threshold. It calculates the health status score based on the deep feature matrix, analyzes the contribution of risk factors, and generates early warning information and intervention suggestions. The model optimization module is connected to the health assessment and early warning module and the quantum computing processing module respectively. Based on the user's health status feedback data, it dynamically adjusts the parameters of the quantum dimensionality reduction model and the quantum neural network model to achieve adaptive optimization of the model. The data storage and interaction module stores standardized high-dimensional health datasets, deep feature matrices, health assessment results, early warning information, and model parameters. It also provides a user interaction interface, allowing users to view health reports, early warning information, and intervention suggestions, and supports data interaction with medical institutions and health management platforms.

[0016] Compared with existing technologies, this health status monitoring method and system based on big data multi-dimensional assessment has the following beneficial effects: I. This invention effectively extracts core low-dimensional feature vectors by collecting health data from multiple sources and combining quantum coding and high-dimensional data dimensionality reduction techniques. Furthermore, it delves into the nonlinear correlation features in the feature vectors through a quantum neural network model, achieving a comprehensive and accurate assessment of the user's health status. This multi-dimensional and in-depth assessment method can more accurately reflect the user's health status and provide a more reliable basis for subsequent early warning and intervention.

[0017] Second, this invention incorporates a built-in model optimization and dynamic update mechanism, which can dynamically adjust the parameters of the quantum dimensionality reduction model and the quantum neural network model based on user health status feedback data. This adaptive optimization capability ensures that the model can continuously adapt to changes in the individual user's health status, improving the accuracy and stability of long-term assessment. At the same time, by optimizing the model structure and adopting a quantum state compression algorithm, the system achieves efficient operation at the edge, with inference latency controlled within 100ms, meeting the needs of real-time health monitoring and providing users with timely and effective health warnings and intervention suggestions.

[0018] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0020] Figure 1 A flowchart of a health status monitoring method based on big data and multi-dimensional assessment; Figure 2 A flowchart illustrating the process of ensuring the synchronization of multi-source data acquisition in a health status monitoring method based on big data and multi-dimensional assessment. Figure 3 This is a flowchart of a health status monitoring system based on multi-dimensional big data assessment. Detailed Implementation

[0021] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structure, features and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0022] Example 1 The multi-source data acquisition module consists of a wearable physiological monitoring bracelet, a gene detection module, an office and home environment sensor array, and a behavior recording terminal. To ensure data acquisition synchronization, all devices are calibrated with the system clock via a network time protocol. The wearable bracelet and environmental sensors employ an edge-end local caching + timed upload mode, with a caching period of 1 minute. Uploaded data includes a timestamp, a unique device identifier, and a data verification code, ensuring real-time performance through 5G / Wi-Fi dual-mode transmission. Gene detection data uses a batch upload + timestamp completion mechanism, completing the accurate timestamp based on the acquisition period and matching it with real-time data from the same period. When the system receives data, it verifies its integrity using a CRC32 checksum; if verification fails, the device re-uploads. Finally, the data synchronization verification module performs logical consistency verification of multi-source data from the same period. Data that passes verification is stored in the dataset. Figure 2 As shown.

[0023] In the data preprocessing stage, an improved 3σ criterion is used to identify outliers. The mean μ and standard deviation σ of physiological index data and environmental parameter data are calculated. Data exceeding the interval [μ-3.5σ, μ+3.5σ] are marked as outliers. Continuous time-series outliers are smoothed using a correction method based on local weighted regression, while isolated outliers are replaced with the statistical median of similar data. For missing values, if the missing rate is less than 5%, physiological time-series data are supplemented using linear interpolation of adjacent time-series data, and behavioral classification data are supplemented using the mode imputation method based on historical similar scenarios. If the missing rate is higher than 5%, a secondary data collection mechanism is initiated. Finally, format standardization is performed, converting the base sequences of gene data into numerical codes. All heterogeneous data are uniformly mapped to standardized values ​​in the interval [0, 1], resulting in a standardized high-dimensional health dataset, such as... Figure 1 As shown.

[0024] The standardized high-dimensional health dataset is mapped to quantum superposition state data using a data credibility-weighted quantum mapping algorithm, as shown in the formula: ,in For the first The quantum superposition state corresponding to each classical data point These are the standardized classic data values. , These are the global minimum and maximum values ​​for this type of data, respectively. The data credibility weights are determined based on the stability of different data sources. Then, utilizing the properties of quantum entanglement, a quantum dimensionality reduction model is constructed based on the quantum entanglement-enhanced principal component analysis algorithm, with the following formula: ,in For the first The quantum state projection values ​​of a low-dimensional eigenvector. The number of qubits in the original high-dimensional data. For the first The density matrix of qubits, For the first The quantum unitary transformation matrix corresponding to each principal component The quantum entanglement weight coefficients are used to compress the feature space of quantum superposition data, remove redundant features, and extract core low-dimensional feature vectors that are strongly correlated with the health status of working people, covering key dimensions such as physiological status, genetic susceptibility, environmental influence, and behavioral habits.

[0025] The core low-dimensional feature vectors are input into a pre-trained quantum neural network model. This pre-training is based on a federated learning architecture, with a mixed training set containing a public health dataset of over 100,000 users and historical health data from the target working population. Only the model parameter gradients are transmitted during training to protect data privacy. Pre-training consists of two phases: the first phase trains a basic model based on the public dataset, learning the general characteristics and patterns of health data; the second phase fine-tunes the model based on a private dataset of the target population, adapting it to its feature distribution. After pre-training, a quantum state compression algorithm is used to achieve model lightweighting.

[0026] The quantum neural network adopts a three-layer architecture: a quantum entanglement layer, a quantum convolutional layer, and a measurement layer. The quantum entanglement layer constructs multi-order correlation mappings between features by alternating combinations of two-qubit CNOT gates and three-qubit Tooffoli gates. Each layer contains 8-12 quantum gate combination units, and the number of units is adaptively adjusted according to the dimension of the low-dimensional feature vector. The quantum convolutional layer designs quantum convolution kernels and extracts features in the frequency domain through quantum Fourier transform. The kernel size is dynamically set to 2×2 or 3×3 according to the dimension of the feature vector, and the convolution stride is set to 1. Parallel convolution computation is achieved by utilizing the superposition property of quantum states. The model training uses quantum state fidelity as the loss function and adopts the quantum gradient descent algorithm to adjust the quantum gate parameters. The number of training iterations is set to 100-200 times until the fidelity value is higher than 0.95, at which point training stops. The measurement layer converts quantum state data into a classical deep feature matrix through quantum measurement operators and reduces quantum measurement noise by averaging multiple measurements.

[0027] The system incorporates an evaluation indicator system with 5 primary indicators and 20 secondary indicators. Each secondary indicator corresponds to a specific feature dimension in a deep feature matrix. The weights of the primary indicators are determined using the analytic hierarchy process (AHP), and the dynamic weights of the secondary indicators are determined using the entropy weighting method. The scores of the secondary indicators are converted into quantitative values ​​of 0-100 by comparing the feature values ​​with preset thresholds, and then aggregated to obtain the user's overall health status score. The contribution of risk factors is calculated using the elasticity coefficient method, which measures the degree of impact of small changes in each feature dimension on the overall health score. After ranking, the top 5 features with the highest impact are selected as core risk factors, and the contribution ratio of each factor is clearly defined. When the overall health status score falls below a preset warning threshold, or when the contribution of a core risk factor exceeds a safety threshold, targeted warning information and intervention suggestions are generated.

[0028] After viewing the health report, users can provide feedback on the implementation of intervention suggestions and changes in their health status through an interactive interface. The model optimization module collects this feedback data as user health status feedback data and dynamically adjusts the formula based on the model parameters. The formula is as follows: ,in For the updated model parameters, These are the current model parameters. For learning rate, The gradient of the loss function. To calculate the feedback data validity coefficient, the feedback data validity coefficient, learning rate, and loss function gradient are incorporated into the calculation. The parameters of the quantum dimensionality reduction model and the quantum neural network model are dynamically adjusted to ensure the model continuously adapts to changes in the user's health status, thereby improving the accuracy of subsequent monitoring and evaluation. Figure 3 As shown.

[0029] Example 2 The multi-source data acquisition module consists of wearable physiological monitoring devices, a gene detection module, a home environment sensor array, and a behavior recording terminal. Data acquisition synchronization is ensured by adhering to a unified standard; all devices are calibrated with the system clock via a network time protocol. The wearable physiological monitoring devices and environmental sensors employ a local edge caching + timed upload mode, with a caching period of 1 minute. Uploaded data includes a timestamp, device identifier, and data verification code, and real-time performance is ensured through 5G / Wi-Fi dual-mode transmission. Non-real-time data such as gene detection data and hospital medical examination records are uploaded in batches with a timestamp completion mechanism, and the completed data is matched with concurrent real-time data. The system verifies the integrity of data transmission using a CRC32 checksum; if verification fails, a re-upload is triggered. After logical consistency verification by the data synchronization verification module, data that passes verification is stored in the dataset. During data preprocessing, an improved 3σ criterion was used to identify outliers. The mean μ and standard deviation σ of numerical data such as blood pressure and blood glucose were calculated. Data exceeding the interval [μ-3.5σ, μ+3.5σ] were marked as outliers. Continuous time-series outliers were smoothed using a local weighted regression correction method, while isolated outliers were replaced with the statistical median of similar data. Regarding missing value handling, when the missing value ratio was less than 5%, physiological time-series data were supplemented using linear interpolation between adjacent time points, and behavioral classification data were supplemented using the mode imputation method based on historical similar scenarios. When the missing value ratio was greater than 5%, a second round of data collection was initiated. Format standardization converted the base sequences of gene data into numerical codes, and all heterogeneous data were uniformly mapped to standardized values ​​in the interval [0, 1], forming a standardized high-dimensional health dataset, such as... Figure 2 As shown.

[0030] A data credibility-weighted quantum mapping algorithm is used to convert standardized high-dimensional health datasets into quantum superposition state data. The data credibility weights are set in accordance with the specific characteristics of data collection from elderly patients. A quantum dimensionality reduction model is constructed based on the quantum entanglement-enhanced principal component analysis algorithm. Utilizing the properties of quantum entanglement, efficient dimensionality reduction is achieved for the quantum superposition state data, focusing on the core dimensions of chronic disease management. Key low-dimensional feature vectors containing crucial information such as blood pressure control level, blood glucose fluctuation amplitude, gene drug sensitivity, environmental adaptability, and medication adherence are extracted. Figure 1 As shown.

[0031] The core low-dimensional feature vectors are input into the pre-trained quantum neural network model. The model pre-training adopts a federated learning architecture, and the mixed training set includes a public health dataset of over 100,000 users and historical health data of the target elderly chronic disease patient group. During the training process, only the model parameter gradients are transmitted to protect patient privacy. The pre-training is divided into two stages: the first stage trains the base model based on the public dataset to learn the general patterns of chronic disease-related health data; the second stage fine-tunes the model based on the private dataset of elderly patients to adapt to the feature distribution of this group. After pre-training, a quantum state compression algorithm is used to achieve lightweight design.

[0032] The quantum neural network adopts a three-layer architecture: the quantum entanglement layer constructs multi-order correlation mappings between features by alternating combinations of two-qubit CNOT gates and three-qubit Tooffoli gates, with each layer containing 8-12 quantum gate combination units, adaptively adjusted according to the dimension of the low-dimensional feature vector; the quantum convolution layer designs quantum convolution kernels, extracts feature frequency domain information through quantum Fourier transform, and dynamically sets the kernel size to 2×2 or 3×3 with a convolution stride of 1, achieving parallel convolution computation by leveraging the superposition property of quantum states; the model training uses quantum state fidelity as the loss function, employs the quantum gradient descent algorithm to adjust the quantum gate parameters, iterates 100-200 times until the fidelity is higher than 0.95, and then stops training; the measurement layer converts quantum state data into a classical deep feature matrix through quantum measurement operators, and uses multiple measurements and averaging to reduce noise.

[0033] The system incorporates an assessment indicator system for elderly patients with chronic diseases, comprising 5 primary indicators and 20 secondary indicators. Each secondary indicator corresponds to a specific dimension of a deep feature matrix. The weights of the primary indicators are determined using the analytic hierarchy process (AHP), and the dynamic weights of the secondary indicators are determined using the entropy weight method. Secondary indicator scores are converted into quantifiable values ​​of 0-100 points by comparing feature values ​​with preset thresholds, and these scores are aggregated to obtain a total health status score. The contribution of risk factors is calculated using the elasticity coefficient method, selecting the top 5 core risk factors with the highest impact and clarifying the contribution ratio of each factor. When the total health score falls below a warning threshold, or the contribution of core risk factors exceeds the standard, targeted warning information and intervention suggestions are generated and simultaneously transmitted to associated medical institutions and family terminals.

[0034] The model optimization module collects user health status feedback data and dynamically adjusts the formula based on the model parameters. It incorporates the feedback data validity coefficient, learning rate, and loss function gradient into the calculation, dynamically adjusting the parameters of the quantum dimensionality reduction model and the quantum neural network model. This allows the model to continuously adapt to changes in the elderly patient's condition, declining physical function, and adjustments in their lifestyle. Figure 3 As shown, this improves the accuracy and adaptability of health monitoring and risk warning.

[0035] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A health status monitoring method based on multi-dimensional assessment using big data, characterized in that, The specific steps of this method are as follows: S1. Multi-source health data acquisition and preprocessing: Acquire users' physiological indicator data, gene data, environmental parameter data and behavioral data through multi-source data acquisition equipment, and obtain standardized high-dimensional health datasets through cleaning, outlier correction and format standardization. S2, Quantum Coding and Dimensionality Reduction of High-Dimensional Data: The standardized high-dimensional health dataset is mapped to quantum superposition state data through a quantum bit coding algorithm. A quantum dimensionality reduction model is constructed using the quantum entanglement property. The feature space of the quantum superposition state data is compressed to extract the core low-dimensional feature vector. S3. Quantum Neural Network Feature Mining: Input low-dimensional feature vectors into a pre-trained quantum neural network model to mine the hidden nonlinear correlation features in the feature vectors and output a deep feature matrix. S4. Health Status Assessment and Risk Warning: Calculate the user's health status score and the contribution of each risk factor based on the deep feature matrix. When the warning conditions are met, generate targeted warning information and intervention suggestions. S5. Model Optimization and Dynamic Update: Based on user health status feedback data, dynamically adjust the parameters of the quantum dimensionality reduction model and the quantum neural network model.

2. The health status monitoring method based on big data multi-dimensional assessment according to claim 1, characterized in that, The synchronization guarantee process for multi-source data acquisition in S1 is as follows: First, all data acquisition devices are calibrated with the system clock using the Network Time Protocol (NTP). Second, for wearable devices and environmental sensor real-time acquisition devices, a local edge caching + timed upload method is adopted, with a caching period of 1 minute. Uploads include timestamps, device identifiers, and data verification codes, and 5G / Wi-Fi dual-mode transmission ensures data real-time performance. For non-real-time data from gene testing devices and medical examination records, a batch upload + timestamp completion mechanism is used. Accurate timestamps are completed based on the data's corresponding acquisition period, and the data is matched with concurrent real-time data using a data association algorithm. Then, when the system receives data, it verifies the data transmission integrity using a CRC32 checksum. If the verification fails, the device is triggered to re-upload. Finally, a data synchronization verification module is constructed to perform logical consistency verification on multi-source data from the same time period. Data that passes verification is stored in a standardized dataset, while data that fails verification is marked as suspicious data and stored separately.

3. The health status monitoring method based on big data multi-dimensional assessment according to claim 1, characterized in that, The specific implementation process of data preprocessing in S1 is as follows: First, an improved 3σ criterion is used to identify outliers in the original data. For physiological indicator data and environmental parameter data, which are numerical data, the mean μ and standard deviation σ are calculated. Data exceeding the interval [μ-3.5σ, μ+3.5σ] are marked as outliers. For continuous time-series outliers, a correction method based on local weighted regression is used for smoothing. For isolated outliers, the statistical median of the same type of data is used for replacement. Second, for missing values, if the missing ratio is less than 5%, an adaptation method is selected according to the data type: for physiological time-series data, linear interpolation based on adjacent time-series data is used for supplementation, and for behavioral classification data, mode imputation based on historical similar scenarios is used for supplementation. If the missing ratio is greater than 5%, the secondary acquisition mechanism of the data acquisition device is activated, and the target device is triggered to re-acquire the corresponding time period data through the device wake-up command. Finally, format standardization processing is performed to convert the base sequence of the gene data into numerical codes and map all heterogeneous data to standardized values ​​in the interval [0, 1].

4. The health status monitoring method based on big data multi-dimensional assessment according to claim 1, characterized in that, The quantum bit encoding algorithm in S2 adopts a data reliability-weighted quantum mapping algorithm, and the formula is: ,in For the first The quantum superposition state corresponding to each classical data point These are the standardized classic data values. , These are the global minimum and maximum values ​​for this type of data, respectively. This represents the data credibility weight.

5. The health status monitoring method based on big data multi-dimensional assessment according to claim 1, characterized in that, The quantum dimensionality reduction model in S2 is constructed based on the quantum entanglement-enhanced principal component analysis algorithm, and the formula is as follows: ,in For the first The quantum state projection values ​​of a low-dimensional eigenvector. The number of qubits in the original high-dimensional data. For the first The density matrix of qubits, For the first The quantum unitary transformation matrix corresponding to each principal component For quantum entanglement weighting coefficients.

6. The health status monitoring method based on big data multi-dimensional assessment according to claim 1, characterized in that, The pre-training process of the quantum neural network model in S3 is as follows: First, a hybrid training set of a large-scale public health dataset and a private health dataset is constructed. The public dataset contains multi-source health data from over 100,000 users, while the private dataset contains historical health data of the target user group. Second, a federated learning architecture is used for pre-training. The public dataset is deployed on a cloud server, and the private dataset is deployed on a local edge node. Collaborative training of the model is achieved through parameter federation and sharing. During training, only the gradients of the model parameters are transmitted, without leaking the original data. Then, the pre-training is divided into two stages: the first stage trains a basic model based on the public dataset to learn the general characteristics and patterns of health data; the second stage fine-tunes the basic model based on the private dataset to adapt to the feature distribution of the target user group. Finally, after pre-training is completed, the model is lightweighted using a quantum state compression algorithm, while data privacy and security are ensured through the federated learning architecture.

7. The health status monitoring method based on big data multi-dimensional assessment according to claim 1, characterized in that, The specific implementation process of the quantum neural network model in S3 is as follows: First, a three-layer network architecture of quantum entanglement layer, quantum convolution layer, and measurement layer is constructed. The quantum entanglement layer uses alternating combinations of two-qubit CNOT gates and three-qubit Tooffoli gates to construct multi-order correlation mappings between features. Each layer contains 8-12 quantum gate combination units, and the number of units is adaptively adjusted according to the dimension of the low-dimensional feature vector. Second, the quantum convolution layer adopts a quantum convolution kernel design to achieve frequency domain extraction of features through quantum Fourier transform. The size of the convolution kernel is dynamically set to 2 according to the dimension of the feature vector. The convolutional stride is set to 1, using the quantum state superposition property to achieve parallel convolutional computation. During model training, quantum state fidelity is used as the loss function, that is, the fidelity value between the model output quantum state and the labeled quantum state is calculated, and the quantum gate parameters are adjusted using the quantum gradient descent algorithm. The number of training iterations is set to 100-200 times, and training stops when the fidelity value is higher than 0.

95. Finally, the measurement layer converts the quantum state data into a classical deep feature matrix through quantum measurement operators. During the measurement process, multiple measurements are taken and averaged to reduce quantum measurement noise.

8. A health status monitoring method based on big data multi-dimensional assessment according to claim 1, characterized in that, The specific calculation process of the health status score in S4 is as follows: First, an evaluation index system containing 5 primary indicators and 20 secondary indicators is constructed, with each secondary indicator corresponding to a specific feature dimension in the deep feature matrix; second, the weight of the primary indicators is determined by the analytic hierarchy process, and the dynamic weight of the secondary indicators is determined by the entropy weight method; then, the score of each indicator is calculated: the secondary indicator score is converted into a quantitative value of 0-100 points by comparing the feature value with a preset threshold; finally, the contribution of risk factors is calculated using the elasticity coefficient method, that is, the degree of influence of small changes in each feature dimension on the total health score is calculated, and the top 5 features with the highest degree of influence are selected as core risk factors after ranking, thus clarifying the contribution ratio of each factor.

9. A health status monitoring method based on big data multi-dimensional assessment according to claim 1, characterized in that, The specific formula for dynamically adjusting the model parameters in S5 is as follows: ,in For the updated model parameters, These are the current model parameters. For learning rate, The gradient of the loss function. This is the validity coefficient of the feedback data.

10. A health status monitoring system based on big data multi-dimensional assessment, applicable to the health status monitoring method based on big data multi-dimensional assessment as described in any one of claims 1-9, characterized in that, The system includes: a multi-source data acquisition module, a data preprocessing module, a quantum computing processing module, a health assessment and early warning module, a model optimization module, and a data storage and interaction module. The multi-source data acquisition module consists of a wearable physiological monitoring device, a gene detection module, an environmental sensor array, and a behavior recording terminal, which simultaneously collects multi-dimensional health-related data. The data preprocessing module is connected to the multi-source data acquisition module to clean, correct outliers, and standardize the format of the raw acquired data, and output a standardized high-dimensional health dataset. The quantum computing processing module includes a quantum encoding unit, a quantum dimensionality reduction unit, and a quantum neural network unit. The quantum encoding unit is used to map classical high-dimensional data into quantum superposition state data. The quantum dimensionality reduction unit is used to efficiently reduce the dimensionality of quantum superposition state data by utilizing the properties of quantum entanglement and extract core low-dimensional feature vectors. The quantum neural network unit is used to mine nonlinear correlation features in low-dimensional feature vectors and output a deep feature matrix. The health assessment and early warning module is connected to the quantum computing processing module. It has a built-in health status assessment index system and early warning threshold. It calculates the health status score based on the deep feature matrix, analyzes the contribution of risk factors, and generates early warning information and intervention suggestions. The model optimization module is connected to the health assessment and early warning module and the quantum computing processing module respectively. Based on the user's health status feedback data, it dynamically adjusts the parameters of the quantum dimensionality reduction model and the quantum neural network model to achieve adaptive optimization of the model. The data storage and interaction module stores standardized high-dimensional health datasets, deep feature matrices, health assessment results, early warning information, and model parameters. It also provides a user interaction interface, allowing users to view health reports, early warning information, and intervention suggestions, and supports data interaction with medical institutions and health management platforms.

Images