Health risk assessment method and device based on multi-disease data, and terminal

By preprocessing and feature extraction of multi-disease data, and generating decision trees and network models, the problem of inaccurate multi-disease health risk assessment in the existing technology is solved, and personalized health risk assessment for patients with multiple diseases is realized, which improves the Accuracy and targetedness of the assessment.

CN119092110AInactive Publication Date: 2024-12-06SHENZHEN LIKANG TECHNOLOGY CO LTD
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202410992276.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-12-06
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the health risk assessment of multi-disease data, the prediction results are not accurate enough and lack targeted based on multi-factor risk assessment models and based on statistical and machine learning methods.

Method used

By obtaining the initial data set for preprocessing, classification and feature extraction are performed based on cluster analysis algorithm and timing algorithm, target decision tree is generated, and initial network model is constructed, weight value is adjusted to generate target evaluation model, and health risk assessment is carried out for target patients. .

Benefits of technology

The accuracy and pertinence of health risk assessment for patients with multi-disease data has been improved, and health risk assessment standards are formed suitable for various populations.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119092110A_ABST
    Figure CN119092110A_ABST
Patent Text Reader

Abstract

The invention discloses a health risk assessment method and device based on multi-disease data and a terminal. The method comprises the following steps: acquiring an initial data set, and preprocessing the initial data set to obtain a target data set with the same feature value; performing classification and feature extraction on the target data set based on a clustering analysis algorithm and a time sequence algorithm to obtain a target training set and a target support set; generating a target decision tree based on the target training set; constructing an initial network model based on the target decision tree, and adjusting a weight value in the initial network model based on the target support set to obtain a target evaluation model; and evaluating the health risk of the target patient based on the target evaluation model. According to the health risk assessment method based on the multi-disease data provided by the invention, the network model is constructed based on the decision tree, targeted health risk assessment is performed on the patient with the multi-disease data, and the pertinence and accuracy of the model are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of computer-aided diagnosis, and in particular to a health risk assessment method, device and terminal based on multi-disease data. Background Art

[0002] With the development of social economy, people's lifestyles and environmental factors may cause the same patient to suffer from multiple diseases at the same time. Some diseases have little impact on the body and the adverse effects are relatively small. However, some diseases will seriously affect the patient's health if they are suffered from at the same time. For example, if a patient has cardiovascular disease before and is diagnosed with kidney disease, kidney disease may lead to the aggravation of cardiovascular disease. At this time, providing reliable early warning can enable patients to obtain effective coping methods in advance and avoid more serious cardiovascular diseases. Therefore, it is particularly important to assess the health risks of individuals with multiple diseases.

[0003] However, there is currently no health risk assessment method specifically for multi-disease data. When relevant predictions are needed, they can only be made based on existing multi-factor risk assessment models and statistical and machine learning methods. The prediction results are not accurate and targeted.

[0004] Therefore, the prior art still needs to be improved and enhanced. Summary of the invention

[0005] In view of the above-mentioned defects of the prior art, a health risk assessment method, device and terminal based on multi-disease data are provided, aiming to solve the problem that in the prior art, when conducting health risk assessment based on multi-disease data, the health risk assessment can only rely on the existing multi-factor risk assessment model and make predictions based on statistics and machine learning methods, and the prediction results are not accurate and targeted.

[0006] A first aspect of the present invention provides a health risk assessment method based on multi-disease data, comprising:

[0007] Acquire an initial data set, and preprocess the initial data set to obtain a target data set with the same characteristic value;

[0008] Based on a cluster analysis algorithm and a time series algorithm, the target data set is classified and features are extracted to obtain a target training set and a target support set;

[0009] Generate a target decision tree based on the target training set;

[0010] Building an initial network model based on the target decision tree, and adjusting the weight values ​​in the initial network model based on the target support set to obtain a target evaluation model;

[0011] The health risk of the target patient is evaluated based on the target evaluation model.

[0012] In one implementation, preprocessing the initial data set to obtain a target data set with the same characteristic value includes:

[0013] Deduplication and missing value filling are performed on the initial data set to obtain a first data set;

[0014] removing noise from the first data set to obtain a second data set;

[0015] Convert all data in the second data set into a unified format and unit to obtain third data;

[0016] The characteristic values ​​in the third data are scaled to the same range to obtain the target data set.

[0017] In one implementation, the classifying and feature extracting the target data set based on a clustering analysis algorithm and a time series algorithm includes:

[0018] Processing first data based on the time series algorithm, where the first data is data with time attributes in the target data set;

[0019] The second data is processed based on the cluster analysis algorithm, where the second data is other data in the target data set except the first data.

[0020] In one implementation, generating a target decision tree based on the target training set includes:

[0021] Recursively splitting the target training set based on a target splitting criterion to form a tree structure, thereby obtaining an initial decision tree;

[0022] Based on the cross-validation technology, training and validating the initial decision tree is performed to obtain a first decision tree;

[0023] The first decision tree is pruned to obtain the target decision tree.

[0024] In one implementation, constructing an initial network model based on the target decision tree includes:

[0025] Converting nodes in the target decision tree into neurons in the initial network model;

[0026] The path rules in the target decision tree are converted into connection weights between neurons in the initial network model.

[0027] In one implementation, the evaluating the health risk of the target patient based on the target evaluation model includes:

[0028] Obtaining target information and disease information of the target patient, wherein the target information includes the target patient's age, height, gender, weight, race, region, drinking, family history, dietary preferences, and exercise frequency, and the disease information includes the diagnosed disease and the corresponding time;

[0029] classifying the target patient based on the target information;

[0030] The health risk of the target patient is evaluated based on the type of the target patient to obtain a target risk level, which includes high, medium and low levels.

[0031] In one implementation, after evaluating the health risk of the target patient based on the target evaluation model, the method further includes:

[0032] Generate a risk assessment report and a health management strategy based on the health risk level of the target patient;

[0033] Dynamically monitor the real-time physiological data of the target patients and predict their health risk levels, so as to provide personalized health management for each of the target patients based on the latest level.

[0034] A second aspect of the present invention provides a health risk assessment device based on multi-disease data, comprising:

[0035] A preprocessing module is used to obtain an initial data set, preprocess the initial data set, and obtain a target data set with the same characteristic value;

[0036] A feature extraction module, used to classify and extract features from the target data set based on a cluster analysis algorithm and a time series algorithm to obtain a target training set and a target support set;

[0037] A decision tree generation module, used to generate a target decision tree based on the target training set;

[0038] A training module, used to construct an initial network model based on the target decision tree, and adjust the weight values ​​in the initial network model based on the target support set to obtain a target evaluation model;

[0039] The prediction module is used to evaluate the health risk of the target patient based on the target evaluation model.

[0040] The third aspect of the present invention provides a terminal, comprising: a processor, a storage medium communicatively connected to the processor, the storage medium being suitable for storing multiple instructions, and the processor being suitable for calling the instructions in the storage medium to execute the steps of implementing any of the above-mentioned health risk assessment methods based on multi-disease data.

[0041] A fourth aspect of the present invention provides a storage medium, wherein the storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of any of the above-mentioned health risk assessment methods based on multi-disease data.

[0042] Beneficial effects: Compared with the prior art, the present invention provides a health risk assessment method, device and terminal based on multi-disease data. In the health risk assessment method based on multi-disease data provided by the present invention, when conducting health risk assessment on a target patient suffering from multiple diseases, an initial data set is obtained, the initial data set is preprocessed to obtain a target data set with the same characteristic values, and then the target data set is classified and feature extracted based on a clustering analysis algorithm and a time series algorithm to obtain a target training set and a target support set, a target decision tree is generated based on the target training set, an initial network model is constructed based on the target decision tree, and the weight values ​​in the initial network model are adjusted based on the target support set to obtain a target evaluation model, and finally, the health risk of the target patient is evaluated based on the target evaluation model. The health risk assessment method based on multi-disease data provided by the present invention can construct the data after classification and feature extraction into a target decision tree through the framework required for quantitative health prediction, and then construct a network model based on the decision tree, and construct different network models based on different data types to obtain a fully intelligent model. This solves the problem that in the prior art, when conducting health risk assessment based on multi-disease data, the prediction can only rely on the existing multi-factor risk assessment model and be based on statistical and machine learning methods, and the prediction results are not accurate enough. Targeted health risk assessment is performed on patients with multi-disease data, the model's targeting and accuracy are improved, and a health risk assessment standard suitable for various populations is formed. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A flowchart of an embodiment of a health risk assessment method based on multi-disease data provided by the present invention;

[0044] Figure 2 A schematic diagram of the structure of a transfer learning model of an embodiment of a health risk assessment method based on multi-disease data provided by the present invention;

[0045] Figure 3 A schematic diagram of the structure of an embodiment of a health risk assessment device based on multi-disease data provided by the present invention;

[0046] Figure 4 A structural principle diagram of an embodiment of a terminal provided by the present invention. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and effect of the present invention clearer and more specific, the present invention is further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0048] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present invention refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0049] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as generally understood by those skilled in the art in the art to which the present invention belongs. It should also be understood that terms such as those defined in general dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless specifically defined as herein.

[0050] The health risk assessment method based on multi-disease data provided by the present invention can be applied to a terminal with computing capabilities. The terminal can execute the health risk assessment method based on multi-disease data provided by the present invention to detect and locate the target position in the image to be processed.

[0051] Embodiment 1

[0052] Due to the influence of lifestyle and environmental factors, the same patient may suffer from multiple diseases at the same time. Suffering from multiple diseases at the same time may significantly increase health risks, and patients may be more susceptible to serious complications and death. For example, if a patient has cardiovascular disease before and is diagnosed with kidney disease, and kidney disease may lead to worsening of cardiovascular disease, providing reliable early warning can enable patients to obtain effective coping methods in advance and avoid more serious cardiovascular disease.

[0053] In response to the above-mentioned defects, this embodiment provides a health risk assessment method based on multi-disease data. Specifically, this embodiment intends to establish a mathematical model and adopts data mining technology to extract valuable data from big data for analysis. Data mining technology mainly includes decision tree algorithm, neural network algorithm, cluster analysis algorithm and time series algorithm. From a large pile of disordered data, cluster analysis algorithm and time series algorithm are used to classify the data, generate data samples required for the decision tree, and test and correct the data samples according to the decision tree algorithm to generate data nodes of the neural network. Weight values ​​are added to the connections of each node, and the weights are continuously adjusted according to the neural network algorithm to achieve the approximation of the output value and the target value. By organizing the algorithm model, algorithm specifications and weight parameters, a health risk assessment standard suitable for people with multiple diseases is formed.

[0054] like Figure 1 As shown, the health risk assessment method based on multi-disease data provided in this embodiment includes the following steps:

[0055] S100 , obtaining an initial data set, and preprocessing the initial data set to obtain a target data set with the same characteristic values.

[0056] The acquisition of the initial data set is to obtain health data of multiple patients with multiple diseases from different data sources, and the health data includes disease information and general information. Specifically, the acquisition includes electronic medical records (EMR), physical examination reports, laboratory test results, imaging data (such as X-rays, CT, MRI, etc.) and diagnostic information issued by doctors as disease information, wherein the disease information includes physical diseases, mental health status and pregnancy status. At the same time, it is also necessary to obtain the patient's general information and the corresponding lifestyle-related data, such as age, height, gender, weight, race, region, drinking, family history, dietary preferences, and exercise frequency. The diversity of data helps to fully understand the patient's health status.

[0057] After the initial data set is acquired, the initial data set is preprocessed to obtain a target data set with the same characteristic values.

[0058] The preprocessing of the initial data set to obtain a target data set with the same characteristic value includes:

[0059] Deduplication and missing value filling are performed on the initial data set to obtain a first data set;

[0060] removing noise from the first data set to obtain a second data set;

[0061] Convert all data in the second data set into a unified format and unit to obtain third data;

[0062] The characteristic values ​​in the third data are scaled to the same range to obtain the target data set.

[0063] Specifically, deduplication of the initial data set includes merging duplicate data of the same patient from different data sources to retain the latest and most comprehensive records.

[0064] The missing value filling includes filling the features with fewer missing values ​​in the initial data with the mean, median or mode; for features with more missing values, considering deleting the feature or using a more complex interpolation method, including but not limited to KNN interpolation and deep learning interpolation methods.

[0065] The noise removal of the first data set includes using statistical methods, including but not limited to box plots and standard deviations, to identify and delete outliers and erroneous data in the first data.

[0066] The converting of all data in the second data set into a unified format and unit includes converting all laboratory results into the standard International System of Units (SI) to ensure the consistency and comparability of the data.

[0067] Scaling the eigenvalues ​​in the third data to the same range includes scaling the values ​​of different features in the third data to the same range using z-score standardization (subtracting the mean and dividing by the standard deviation) or Min-Max standardization (scaling the value to the range of [0, 1]). Among them, z-score standardization scales the eigenvalues ​​to the same range by subtracting the mean and dividing by the standard deviation, while Min-Max standardization scales the eigenvalues ​​to the same range by scaling the values ​​to the range of [0, 1]. Standardization processing helps to improve the convergence speed and stability of the model.

[0068] In this way, the target data set can be obtained by removing duplicates, filling missing values, removing noise, unifying the format, and scaling the eigenvalues ​​of the initial data set.

[0069] S200 , classifying and extracting features from the target data set based on a cluster analysis algorithm and a time series algorithm to obtain a target training set and a target support set.

[0070] The classifying and feature extracting of the target data set based on the clustering analysis algorithm and the time series algorithm includes:

[0071] S210, processing first data based on the time series algorithm, where the first data is data with time attributes in the target data set;

[0072] S220. Process second data based on the cluster analysis algorithm, where the second data is other data in the target data set except the first data.

[0073] Through feature extraction, complex raw data can be converted into input variables with clear meaning, laying the foundation for subsequent model construction.

[0074] Specifically, in this embodiment, the first data is processed based on the time series algorithm, and the first data is the data with time attributes in the target data set. Specifically, the time series algorithm can be used to analyze and process data with time attributes, and commonly used time series analysis algorithms include ARIMA, LSTM, etc. Among them, the ARIMA model is suitable for linear time series data, and the data is modeled by autoregression, difference and moving average to classify the first data; and the LSTM network: is suitable for processing long-term dependent nonlinear time series data, and can capture complex time patterns in the data to classify the first data.

[0075] Then, the second data is processed based on the cluster analysis algorithm, and the second data is other data in the target data set except the first data. Specifically, the cluster analysis algorithm is used to discover the intrinsic structure in the data and group similar data points into a group. Common clustering algorithms include K-means, hierarchical clustering, DBSCAN, etc. Among them, K-means clustering selects the initial cluster center, iteratively updates the attribution and cluster center of each data point until the clustering results converge, to classify the second data; hierarchical clustering is to classify the second data by constructing a tree-like hierarchical structure, gradually aggregating or splitting the data to form clusters of different levels; DBSCAN is a density-based clustering algorithm, which forms clusters by identifying data points with similar density, and is suitable for discovering clusters of arbitrary shapes.

[0076] In this way, the cluster analysis algorithm and the time series algorithm can classify and extract features of different types of data in the target data set to generate the target training set and the target support set. The target training set includes disease development and health risk development under the time series corresponding to multiple groups of patient disease information and general information, and the target support set also includes disease development and health risk development under the time series corresponding to multiple groups of patient disease information and general information, but the data in the target support set is less than that in the target training set.

[0077] S300, generating a target decision tree based on the target training set;

[0078] The generating a target decision tree based on the target training set includes:

[0079] S310, recursively splitting the target training set based on the target splitting criterion to form a tree structure to obtain an initial decision tree;

[0080] S320, performing training verification on the initial decision tree based on a cross-validation technique to obtain a first decision tree;

[0081] S330: Prune the first decision tree to obtain the target decision tree.

[0082] Specifically, according to the target training set obtained by classification, a preliminary decision tree is generated using a decision tree algorithm. In this embodiment, the decision tree algorithms used include but are not limited to CART, ID3, C4.5, etc.

[0083] In this embodiment, the target decision tree is generated based on the target training set. It is necessary to select a suitable splitting criterion such as the Gini coefficient, information gain, gain rate, etc. based on the original data set, and then recursively split the target training set to form a tree structure to obtain the initial decision tree, wherein each node in the initial decision tree represents a decision point, and each leaf node represents a classification result.

[0084] Then, the initial decision tree is trained and verified based on the cross-validation technology to obtain a first decision tree. Specifically, since the initially generated initial decision tree may have overfitting or underfitting problems, correction is required. In this embodiment, the cross-validation technology is used to evaluate the classification accuracy of the initial decision tree, and the optimal decision tree structure is selected through multiple training and verification to obtain the first decision tree.

[0085] Finally, the first decision tree is pruned. Specifically, the complexity of the decision tree is reduced by pre-pruning and post-pruning to prevent overfitting, wherein pre-pruning is to stop splitting in advance during the generation process, and post-pruning is to prune unimportant nodes after generation. Finally, the splitting condition is adjusted according to the data characteristics and distribution, and the splitting condition and threshold of the decision tree are adjusted to obtain the target decision tree. The target decision tree obtained in this way can more accurately reflect the characteristics of the data.

[0086] S400: constructing an initial network model based on the target decision tree, and adjusting the weight values ​​in the initial network model based on the target support set to obtain a target evaluation model.

[0087] The constructing an initial network model based on the target decision tree comprises:

[0088] Converting nodes in the target decision tree into neurons in the initial network model;

[0089] The path rules in the target decision tree are converted into connection weights between neurons in the initial network model.

[0090] Specifically, by parsing the target decision tree, a series of if-then rules can be extracted, which define the path from the root node to the leaf node, and these rules can be directly converted into the weights and thresholds of the multi-layer perceptron model in the neural network. Each node in the target decision tree can correspond to a neuron in the neural network, and each rule in the target decision tree can be converted into the connection weights between neurons in the neural network.

[0091] In this embodiment, the constructing of the initial network model based on the target decision tree further includes:

[0092] Constructing the initial network model based on the initial data set type;

[0093] When the initial data set is an ultra-health parameter data set, that is, when the initial data set is large in size and has high data integrity, constructing the initial network model is to construct a deep learning neural network containing K hidden layers.

[0094] At this point, the structure of the initial network model is as follows:

[0095]

[0096] Where X = H (0) is the original super-healthy input matrix, H (k) ,k=1,2,L,K is the k-th layer depth representation of X, W (k) is the weight of the kth layer of the deep neural network. In this embodiment, a non-iterative method is used to determine W (k) The specific calculation can be expressed as

[0097]

[0098] Where V (k) is a random value matrix.

[0099] In this embodiment, in order to solve the problem of super-healthy parameters of mixed-value attributes, it is necessary to convert discrete-value attributes into continuous-value attributes. Specifically, the discrete-value attributes are firstly encoded using a one-hot method, and then a deep neural network is constructed to convert the one-hot encoding. In the process of training the neural network, the loss function uses a combination of mean square error and information volume to ensure that the continuous-value attributes after conversion have the lowest error while minimizing the loss of information volume.

[0100] For the non-convex logistic regression loss function, a complementary constrained optimization problem is constructed as follows:

[0101]

[0102] The non-negative vertical constraint set has the advantages of equivalent optimization, good initialization, strong convergence, greedy monotonicity, and modular optimization. It is proposed to use a smooth approximation method based on minimizing entropy function to solve the above complementary constraint optimization problem.

[0103] When the initial data set is a missing health parameter data set, the initial network model includes a plurality of base layers and a plurality of target layers;

[0104] The multiple target layers are trained based on the multiple base layers to obtain the initial network target evaluation model.

[0105] Specifically, in this embodiment, transfer learning technology is used as a key technology for data filling to provide high-precision data for feature extraction and later machine learning. In the process of human brain recognition of samples, basic judgments are initially made, and then special classifications are performed. The deep neural network adopts the same mechanism, by identifying and recording labeled samples, training the weights of each layer of the neural network, and then identifying unclassified objects. Since the preliminary judgment process for different samples is similar, this part of the neural network (base layer) is reusable.

[0106] The transfer learning process is as follows Figure 2 As shown in the figure: First, the first deep neural network is trained with a labeled data set, and several layers are selected as the base layer Layer L1, and the remaining layers are the target layer Layer L2; then, for a new unlabeled or incompletely labeled data set, a new deep neural network is established, Layer L1 is migrated to the new network as the base layer, and the new target layer is trained to improve the neural network. Finally, multiple attempts are made to select different numbers of layers for Layer L1, determine a more reasonable number of base layers, and establish the initial network model of adaptive transfer learning.

[0107] When the initial data set is a random health parameter data set, the initial network model is a multi-hidden layer feature subspace model or a nested hybrid regularization model.

[0108] Specifically, the research on feature grouping has not been able to effectively mine the statistical correlation within and between health big data feature groups. The data set may contain random health parameter data, such as whether there is a Gaussian mixture distribution, whether there is a positive or negative correlation between groups, etc. Therefore, in this embodiment, the initial network model can be constructed as a multi-hidden layer feature subspace model, and a linear mapping is established through the feature subspace hidden layer method to convert high-dimensional data into The groupable features Projected into a low-dimensional space. Through this framework, the original data information is deconstructed, so that the feature grouping is optimized and merged using a discretized biological evolution algorithm, and the weights are optimized using an analytical method. A high-precision feature analysis is obtained through a low amount of calculation, and subsequent machine learning is performed; or the initial network model is constructed as a nested hybrid regular model. In different types of ultra-high-dimensional data, the weights of features also have different characteristics. For example, the importance of features in some data is not much different, while in some data there are a large number of redundant variables and redundant features. Regularization methods provide an effective way to solve the above problems. In this way, the interactive relationship between the two learning processes of feature grouping and clustering is used to nest the hybrid regular model into the unsupervised clustering process, and the data is clustered and the features are learned at the same time.

[0109] S500: Evaluate the health risk of the target patient based on the target evaluation model.

[0110] The health risk assessment of the target patient based on the target assessment model includes:

[0111] Obtaining target information and disease information of the target patient, wherein the target information includes the target patient's age, height, gender, weight, race, region, drinking, family history, dietary preferences, and exercise frequency, and the disease information includes the diagnosed disease and the corresponding time;

[0112] classifying the target patient based on the target information;

[0113] The health risk of the target patient is evaluated based on the type of the target patient to obtain a target risk level, which includes high, medium and low levels.

[0114] In this embodiment, the target information and disease information of the target patient are target data for evaluation. Assume that the current data set is D old , whose probability density function is p old The existing health risk assessment model is L old , the health risk assessment model based on the newly added data training is Then the relative data set The health risk assessment model can be expressed as:

[0115]

[0116] The operator symbol represents the integration of different evaluation models, represents the weight of the base model. The key here is the weight For the vertically increasing data, the data set is first obtained by the kernel density estimation method. The probability density function of By measuring With D old The similarity between For horizontally increasing data, the weight is determined by measuring the amount of information in the data set. The value of data set D old and The amount of information can be calculated by substituting it into the entropy.

[0117] Specifically, analyze the health data characteristics of different groups of people and develop feature grouping standards. Feature grouping refers to grouping the features in the data according to certain rules for more effective analysis and evaluation. Then, use the deep neural network model to analyze each feature group and calculate the health risk weight of each group. The health risk weight refers to the contribution of each feature group to the overall health risk. By calculating the health risk weight, the impact of different features on health risks can be evaluated. Finally, based on the analysis results, formulate health risk assessment standards suitable for different groups of people to ensure the wide applicability and accuracy of the assessment standards. Health risk assessment standards refer to a set of systematic evaluation standards, through which individual health risks can be comprehensively evaluated.

[0118] In this embodiment, after evaluating the health risk of the target patient based on the target evaluation model, the method further includes:

[0119] Generate a risk assessment report and a health management strategy based on the health risk level of the target patient;

[0120] Dynamically monitor the real-time physiological data of the target patients and predict their health risk levels, so as to provide personalized health management for each of the target patients based on the latest level.

[0121] Specifically, for high-risk patients, we can focus on implementing health management strategies, dynamically monitor their real-time physiological data and predict their health risk levels, so as to provide personalized health management for each target patient based on the latest level.

[0122] In this way, through the quantitative framework of health prediction: the diversity of health data sources and multi-format types cause difficulties in quantitative research, optimize the data transmission terminal compatible with multiple communication protocols and data protocols, improve the integrity and standardization of data, and be compatible with models, algorithms and instruments of structured and unstructured data. Build a fully intelligent model for health hyperparameters: convert the health prediction target into a non-convex optimization problem. Establish a deep learning model for accurate prediction under ultra-high dimensional data volume. Use transfer learning to improve the efficiency of the algorithm, so that the prediction mechanism for individuals has low cost, high speed and high precision. Furthermore, a chronic disease risk platform can also be constructed based on the health risk assessment method based on multi-disease data provided in this embodiment: a distributed architecture design is used to build a cloud platform to realize an incremental database and prediction system. On the one hand, the existing intelligent big data is used to provide health warnings to individuals, and on the other hand, individual dynamic health data is collected to improve internal algorithms and systems. It has high stability and high scalability.

[0123] In summary, this embodiment provides a health risk assessment method based on multi-disease data. When performing health risk assessment on a target patient suffering from multiple diseases, an initial data set is obtained, the initial data set is preprocessed to obtain a target data set with the same characteristic values, and then the target data set is classified and feature extracted based on a clustering analysis algorithm and a time series algorithm to obtain a target training set and a target support set, a target decision tree is generated based on the target training set, an initial network model is constructed based on the target decision tree, and the weight values ​​in the initial network model are adjusted based on the target support set to obtain a target evaluation model. Finally, the health risk of the target patient is evaluated based on the target evaluation model. The health risk assessment method based on multi-disease data provided in this embodiment can construct the data after classification and feature extraction into a target decision tree through a framework required for quantitative health prediction, and then construct a network model based on the decision tree, and construct different network models based on different data types to obtain a fully intelligent model. This solves the problem that in the prior art, when conducting health risk assessment based on multi-disease data, the prediction can only rely on the existing multi-factor risk assessment model and be based on statistical and machine learning methods, and the prediction results are not accurate enough. Targeted health risk assessment is conducted on patients with multi-disease data, the model's targetedness and accuracy are improved, and a health risk assessment standard suitable for various populations is formed.

[0124] It should be understood that, although the various steps in the flowcharts given in the accompanying drawings of the present invention are displayed in sequence according to the indications of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear description in this article, the execution of the steps in the present invention is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps of the present invention may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0125] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0126] Embodiment 2

[0127] Based on the above embodiments, the present invention also provides a health risk assessment device based on multi-disease data, and its functional module diagram is as follows: Figure 3 As shown, the health risk assessment device based on multi-disease data includes:

[0128] A preprocessing module, used to obtain an initial data set, preprocess the initial data set, and obtain a target data set with the same characteristic value, as described in the first embodiment;

[0129] A feature extraction module, used to classify and extract features from the target data set based on a cluster analysis algorithm and a time series algorithm to obtain a target training set and a target support set, as specifically described in Example 1;

[0130] A decision tree generation module, used to generate a target decision tree based on the target training set, as described in Embodiment 1;

[0131] A training module, used to construct an initial network model based on the target decision tree, and adjust the weight values ​​in the initial network model based on the target support set to obtain a target evaluation model, as described in Example 1;

[0132] The prediction module is used to evaluate the health risk of the target patient based on the target evaluation model, as described in Example 1.

[0133] Embodiment 3

[0134] Based on the health risk assessment method based on multi-disease data described in the first embodiment, the present invention also provides a terminal, whose principle block diagram can be as follows: Figure 4 The terminal includes a memory 10 and a processor 20. The memory 10 stores a health risk assessment program based on multi-disease data. When the processor 10 executes the computer program, at least the following steps can be implemented:

[0135] Acquire an initial data set, and preprocess the initial data set to obtain a target data set with the same characteristic value;

[0136] Based on a cluster analysis algorithm and a time series algorithm, the target data set is classified and features are extracted to obtain a target training set and a target support set;

[0137] Generate a target decision tree based on the target training set;

[0138] Building an initial network model based on the target decision tree, and adjusting the weight values ​​in the initial network model based on the target support set to obtain a target evaluation model;

[0139] The health risk of the target patient is evaluated based on the target evaluation model.

[0140] The preprocessing of the initial data set to obtain a target data set with the same characteristic value includes:

[0141] Deduplication and missing value filling are performed on the initial data set to obtain a first data set;

[0142] removing noise from the first data set to obtain a second data set;

[0143] Convert all data in the second data set into a unified format and unit to obtain third data;

[0144] The characteristic values ​​in the third data are scaled to the same range to obtain the target data set.

[0145] The method of classifying and extracting features of the target data set based on a cluster analysis algorithm and a time series algorithm includes:

[0146] Processing first data based on the time series algorithm, where the first data is data with time attributes in the target data set;

[0147] The second data is processed based on the cluster analysis algorithm, where the second data is other data in the target data set except the first data.

[0148] Wherein, generating a target decision tree based on the target training set includes:

[0149] Recursively splitting the target training set based on a target splitting criterion to form a tree structure, thereby obtaining an initial decision tree;

[0150] Based on the cross-validation technology, training and validating the initial decision tree is performed to obtain a first decision tree;

[0151] The first decision tree is pruned to obtain the target decision tree.

[0152] Wherein, constructing an initial network model based on the target decision tree includes:

[0153] Converting nodes in the target decision tree into neurons in the initial network model;

[0154] The path rules in the target decision tree are converted into connection weights between neurons in the initial network model.

[0155] Wherein, the health risk assessment of the target patient based on the target assessment model includes:

[0156] Obtaining target information and disease information of the target patient, wherein the target information includes the target patient's age, height, gender, weight, race, region, drinking, family history, dietary preferences, and exercise frequency, and the disease information includes the diagnosed disease and the corresponding time;

[0157] classifying the target patient based on the target information;

[0158] The health risk of the target patient is evaluated based on the type of the target patient to obtain a target risk level, which includes high, medium and low levels.

[0159] Wherein, after evaluating the health risk of the target patient based on the target evaluation model, the method further includes:

[0160] Generate a risk assessment report and a health management strategy based on the health risk level of the target patient;

[0161] Dynamically monitor the real-time physiological data of the target patients and predict their health risk levels, so as to provide personalized health management for each of the target patients based on the latest level.

[0162] Embodiment 4

[0163] The present invention also provides a storage medium storing one or more programs, which can be executed by one or more processors to implement the steps of the health risk assessment method based on multi-disease data described in the above embodiment.

[0164] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A health risk assessment method based on multi-disease data, characterized in that: include: Acquire an initial data set, and preprocess the initial data set to obtain a target data set with the same characteristic value; Based on a cluster analysis algorithm and a time series algorithm, the target data set is classified and features are extracted to obtain a target training set and a target support set; Generate a target decision tree based on the target training set; Building an initial network model based on the target decision tree, and adjusting the weight values ​​in the initial network model based on the target support set to obtain a target evaluation model; The health risk of the target patient is evaluated based on the target evaluation model.

2. The health risk assessment method based on multi-disease data according to claim 1, characterized in that: The preprocessing of the initial data set to obtain a target data set with the same characteristic value includes: Deduplication and missing value filling are performed on the initial data set to obtain a first data set; removing noise from the first data set to obtain a second data set; Convert all data in the second data set into a unified format and unit to obtain third data; The characteristic values ​​in the third data are scaled to the same range to obtain the target data set.

3. The health risk assessment method based on multi-disease data according to claim 2, characterized in that: The classifying and feature extracting of the target data set based on the clustering analysis algorithm and the time series algorithm includes: Processing first data based on the time series algorithm, where the first data is data with time attributes in the target data set; The second data is processed based on the cluster analysis algorithm, where the second data is other data in the target data set except the first data.

4. The health risk assessment method based on multi-disease data according to claim 1, characterized in that: The generating a target decision tree based on the target training set includes: Recursively splitting the target training set based on a target splitting criterion to form a tree structure, thereby obtaining an initial decision tree; Based on the cross-validation technology, training and validating the initial decision tree is performed to obtain a first decision tree; The first decision tree is pruned to obtain the target decision tree.

5. The health risk assessment method based on multi-disease data according to claim 4, characterized in that: The constructing an initial network model based on the target decision tree comprises: Converting nodes in the target decision tree into neurons in the initial network model; The path rules in the target decision tree are converted into connection weights between neurons in the initial network model.

6. The health risk assessment method based on multi-disease data according to claim 1, characterized in that: The health risk assessment of the target patient based on the target assessment model includes: Obtaining target information and disease information of the target patient, wherein the target information includes the target patient's age, height, gender, weight, race, region, drinking, family history, dietary preferences, and exercise frequency, and the disease information includes the diagnosed disease and the corresponding time; classifying the target patient based on the target information; The health risk of the target patient is evaluated based on the type of the target patient to obtain a target risk level, which includes high, medium and low levels.

7. The health risk assessment method based on multi-disease data according to claim 1, characterized in that: After evaluating the health risk of the target patient based on the target evaluation model, the method further includes: Generate a risk assessment report and a health management strategy based on the health risk level of the target patient; Dynamically monitor the real-time physiological data of the target patients and predict their health risk levels, so as to provide personalized health management for each of the target patients based on the latest level.

8. A health risk assessment device based on multi-disease data, characterized in that: The device comprises: A preprocessing module is used to obtain an initial data set, preprocess the initial data set, and obtain a target data set with the same characteristic value; A feature extraction module, used to classify and extract features from the target data set based on a cluster analysis algorithm and a time series algorithm to obtain a target training set and a target support set; A decision tree generation module, used to generate a target decision tree based on the target training set; A training module, used to construct an initial network model based on the target decision tree, and adjust the weight values ​​in the initial network model based on the target support set to obtain a target evaluation model; The prediction module is used to evaluate the health risk of the target patient based on the target evaluation model.

9. A terminal, characterized in that: The terminal includes: a processor, a storage medium communicatively connected to the processor, the storage medium being suitable for storing a plurality of instructions, and the processor being suitable for calling the instructions in the storage medium to execute the steps of the health risk assessment method based on multi-disease data as described in any one of claims 1 to 7 above.

10. A storage medium, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the steps of the health risk assessment method based on multi-disease data as described in any one of claims 1-7.

Citation Information

Cited By

  • Heart disease prediction system based on AI

    CN119480113A

  • Data management system and method based on NLP

    CN120336540A

  • Lung cancer drug resistance risk assessment system based on multi-modal data and lightweight network

    CN121709271A