Health management system and method based on multi-source data collection and analysis

Abstract

The present invention discloses a health management system and method based on multi-source data collection and analysis, and relates to the technical field of data analysis. The system of the present invention includes: a data collection and processing module, a user classification and portrait construction module, a knowledge graph construction module, a health management plan matching module and a personalized health management module; the data collection and processing module obtains multi-source historical data of preset users to form a preset user feature data set; the user classification and portrait construction module divides the preset users into different categories and creates corresponding preset user portraits; the knowledge graph construction module constructs a preset user knowledge graph based on the feature data set and user portraits, and establishes a key feature set; the health management plan matching module uses the real-time user input initial data to match the corresponding preset health management plan list; the personalized health management module obtains real-time multi-source data, extracts real-time feature data, and generates a personalized health management plan.

Description

Technical Field

[0001] The present invention relates to the technical field of data analysis, and in particular to a health management system and method based on multi-source data collection and analysis. Background Art

[0002] With the rapid development of information technology, health management has become the focus of modern society. Traditional health management methods often rely on a single medical test method, such as physical examination reports, medical records, etc. These data sources are single and it is difficult to fully reflect the health status of individuals and their dynamic changes. With the rise of technologies such as the Internet of Things, big data, and cloud computing, multi-source data collection and analysis have shown great potential in the field of health management, which can more accurately and comprehensively evaluate the health status of individuals and provide personalized health management solutions.

[0003] However, most existing health management systems use simple statistical analysis methods for data analysis, which makes it difficult to deeply explore the correlation and trends behind the data. At the same time, due to differences in data formats and quality from different sources, existing health management systems are difficult to adapt to the complexity and diversity of multi-source data, resulting in inaccurate analysis results. In addition, the health recommendations corresponding to existing integrated health management systems are relatively general, lacking in specificity and personalization, and often only give recommendations based on a single or limited number of health detection indicators, ignoring individual differences and diversified needs, resulting in poor health management results. Summary of the invention

[0004] The purpose of the present invention is to provide a health management system and method based on multi-source data collection and analysis to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] A health management method based on multi-source data collection and analysis, the method comprising the following steps:

[0007] Step S100. Obtain multi-source historical data of all preset users from the database, perform corresponding screening and merging on the multi-source historical data of each preset user, thereby constructing a preset user feature data set; classify the preset users according to the preset user feature data set, and create a corresponding preset user portrait for each category of preset users;

[0008] Step S200. For each category of preset users, a preset user knowledge graph is constructed based on the preset user feature data set and the preset user portrait; the preset user knowledge graph and the preset user feature data set are combined for analysis, and key features of the preset users of the corresponding category are identified according to the analysis results;

[0009] Step S300. Obtain preset health management plans provided for preset users of different categories in the database, analyze the preset health management plans of each category with the key features of the corresponding preset users, and thus obtain the corresponding relationship between the preset health management plans and the key features of the corresponding preset users;

[0010] Step S400. The real-time user inputs initial data, builds a real-time user portrait based on the initial data, matches the real-time user portrait with the preset user portrait, and outputs a corresponding preset health management plan list based on the matching result; obtains multi-source real-time data of the real-time user, extracts the multi-source real-time data to identify the real-time user's real-time feature data, and analyzes it with the key features of the preset user to generate a personalized health management plan.

[0011] Furthermore, step S100 includes:

[0012] S101. Obtain multi-source historical data of all preset users from the database, where the preset users refer to user groups stored in the database; the multi-source historical data refer to user health-related information collected from multiple different data sources; convert the multi-source historical data of the preset users into the same format, aggregate the multi-source historical data by the preset user ID, extract corresponding feature data from the multi-source historical data of each preset user, and perform standardization on the extracted feature data, and the specific standardization calculation formula is:

[0013] F i,j =(X i,j -μ j ) / σ j ,

[0014] Among them, F i,j represents the jth standardized feature data of preset user i, X i,j represents the jth original feature data of preset user i, μ j and σ j Respectively represent the mean and standard deviation of the j-th feature data;

[0015] S102. For each preset user's standardized feature data, correlation analysis is performed to obtain the linear relationship between the standardized feature data, and the value measuring the linear relationship is defined as the correlation coefficient R; the Pearson correlation coefficient is usually used to measure the linear relationship between two features, and a linear regression model is used to establish the corresponding linear relationship for feature data whose absolute value of the correlation coefficient R is equal to the threshold R0, and the linear regression model is expressed as: F i,k =β0+β1F i,j +ϵ, where β0 represents the intercept and β1 represents the feature F i,jThe coefficient of , ϵ represents the error term; the regression coefficients β0 and β1 are calculated using the least squares method, and the linear regression model is used to i,k Update, and F' i,k =β0+β1F i,j ; Traverse all the standardized feature data of the preset user to construct a preset user feature data set, and one element in the preset user feature data set corresponds to one preset user, and the number of elements in the preset user feature data set is equal to the number of preset users in the database;

[0016] S103. Perform cluster analysis on the preset users according to the preset user feature data set, divide the preset users into K categories, and create corresponding preset user portraits for the preset users in each category. The process of creating the preset user portraits is as follows:

[0017] For each category of preset users, the corresponding feature vector V is extracted from the preset user feature data set, the mean and standard deviation of each feature in the category are calculated based on the feature vector V, and normalized based on the standard deviation to obtain the feature weight W; based on the mean and weight of the clustering category, the preset user portrait U of each category is constructed, and the preset user portrait U is expressed as a weighted feature vector.

[0018] Furthermore, cluster analysis is performed on the preset users according to the preset user feature data set. The specific cluster analysis process is as follows:

[0019] According to the preset user feature data set, the feature data of the preset user is obtained, and the Euclidean distance is used as the initial distance metric to calculate the distance metric between all preset users, and the distance metric d(a,b)={Σk∈[1,n],(F a,k -F b,k )^2}^(1 / 2), where F a,k and F b,k Respectively represent the standardized values ​​of preset users a and b on the kth feature, and n represents the total number of features; hierarchical clustering is applied as the initial clustering, starting from each preset user, it is regarded as a separate cluster, and clusters whose distance metric d(a,b) is greater than or equal to the threshold d are gradually merged; based on the results of hierarchical clustering, a cluster number K0 of the initial clustering is selected, and K0 clusters are selected from the final clusters of the hierarchical clustering as the initial center points of K-means; the K-means clustering algorithm is executed, and after each iteration, the quality of the cluster is evaluated and the K value is dynamically adjusted in the following way. The specific analysis process is as follows:

[0020] Calculate the cluster variance sum WCSS, and the specific calculation formula of WCSS is:

[0021] WCSS=Σm∈[1,M],Σp∈Cm,Σk∈[1,n],[f(p,k)-μm,k]^2,

[0022] Where M represents the number of clusters, Cm represents the mth cluster, f(p,k) represents the value of the pth sample on the kth feature, and μm,k represents the mean of the mth cluster on the kth feature. If the reduction of WCSS is less than the threshold t in c consecutive iterations, the iteration is stopped. Otherwise, the iteration continues and the K value is adjusted. Based on the final K, the K-means clustering algorithm is re-run to divide the preset users into K categories.

[0023] Further, step S200 includes:

[0024] S201. For each category of preset users, based on the preset user feature data set and the preset user portrait, extract the preset user feature data as entities, extract the relationship between the preset user feature data in combination with the preset user portrait, use the entities as nodes of the knowledge graph, and use the relationship between the entities as edges of the knowledge graph, and combine the mean and weight of the clustering category in S103 to construct a preset user knowledge graph for each category;

[0025] S202. Obtain the preset user knowledge graph for each category, arbitrarily define a node as the initial node, start from the initial node, traverse the preset user knowledge graph, collect all reachable nodes J and paths P, and thus obtain the corresponding path set Pa; for each path in the path set, calculate the corresponding path weight w(path), and the value of w(path) is equal to the weight product of each feature in the path; based on the path weight w(path), define the key indicator G(J_h), and G(J_h)=ΣP∈Pa,[w(path) / |P|]·I(J_h∈P), where |P| represents the path length, I represents the indicator function, when the node J_h is in the path P, then I(J_h∈P)=1, otherwise I(J_h∈P)=0, and h represents the node number; judge the size relationship between the key indicator G(J_h) and the corresponding threshold G0, when G(J_h)>G0, define the corresponding node J_h as a key node, and the feature corresponding to the node J_h is the key feature.

[0026] Furthermore, step S300 includes:

[0027] The health management plans corresponding to different categories of preset users are extracted from the database, and the corresponding health management plans are matched for each category of preset users; according to the key features corresponding to the preset users, a key feature set Tu is constructed for each preset user, and u represents the preset user number; the symmetric difference of the key feature set Tu of the preset users is calculated in turn, and the health management plans corresponding to the preset users are compared, and the symmetric difference result is associated with the comparison result, so as to obtain the corresponding relationship between the preset health management plan and the key features of the corresponding preset users.

[0028] Furthermore, step S400 includes:

[0029] S401. The real-time user inputs initial data, and the initial data is subjected to the same analysis process as multi-source historical data, so as to construct a real-time user portrait U1 based on the initial data; the real-time user portrait U1 is subjected to similarity calculation with the preset user portrait U, and the corresponding preset user category is matched according to the similarity calculation result, and the one with the greatest similarity is taken as the matching result, and a list of preset health management plans corresponding to the preset user category is output according to the matching result;

[0030] S402. According to the list of preset health management plans, the real-time user selects a preset health management plan and implements the preset health management plan, obtains multi-source real-time data of the real-time user at selected time intervals, extracts corresponding real-time feature data based on the multi-source real-time data, and compares the real-time feature data with the key features corresponding to the list of preset health management plans. If the real-time feature data is the same as the key features of other preset health management plans in the list of preset health management plans other than the preset health management plan currently implemented by the real-time user, according to the correspondence between the other preset health management plans and the key features of the preset user, the currently implemented preset health plan is adjusted accordingly according to the other preset health management plans, thereby generating a personalized health management plan.

[0031] A health management system based on multi-source data collection and analysis, the system includes: data collection and processing module, user classification and portrait construction module, knowledge graph construction module, health management plan matching module and personalized health management module;

[0032] The data acquisition and processing module obtains multi-source historical data of preset users from multiple databases, uniformly converts the multi-source historical data of different formats and structures, extracts and standardizes the feature data of preset users, and thus forms a feature data set of preset users;

[0033] The user classification and portrait construction module performs correlation analysis on the feature data in the preset user feature data set, calculates the linear relationship between the features, uses a linear regression model to establish the relationship between the features, divides users into different categories through cluster analysis, and creates a corresponding preset user portrait for each category;

[0034] The knowledge graph construction module defines nodes and edges based on the preset user feature data set and preset user portrait, establishes the relationship between entities, and constructs the preset user knowledge graph. By traversing the knowledge graph, it identifies key nodes and key features, thereby establishing a key feature set for each preset user.

[0035] The health management program matching module uses the initial data input by the real-time user to construct a real-time user portrait, matches it with the preset user portrait, thereby obtaining the category corresponding to the real-time user, and outputs a list of preset health management programs corresponding to the preset user category according to the matching results;

[0036] The personalized health management module obtains multi-source real-time data of the real-time user based on the preset health management plan selected by the real-time user, extracts corresponding real-time feature data based on the multi-source real-time data, and compares the real-time feature data with the key features corresponding to the preset health management plan list. If the real-time feature data is the same as the key features of other preset health management plans in the preset health management plan list other than the preset health management plan currently implemented by the real-time user, according to the correspondence between the other preset health management plans and the key features of the preset user, the currently implemented preset health plan is adjusted accordingly according to the other preset health management plans, thereby generating a personalized health management plan.

[0037] Further, the data acquisition and processing module includes a data acquisition unit and a data processing unit;

[0038] The data collection unit collects historical multi-source data of the preset user from the database; the data processing unit uniformly converts the multi-source historical data of different formats and structures, extracts and standardizes the characteristic data of the preset user, thereby forming a characteristic data set of the preset user;

[0039] The user classification and portrait building module includes a user classification unit and a user portrait creation unit;

[0040] The user classification performs cluster analysis on the preset user features in the preset user feature data set, and divides the preset users into several categories; the user portrait creation unit creates a weighted feature vector based on the preset user feature data set for each category of preset users, thereby creating a preset user portrait.

[0041] Furthermore, the knowledge graph construction module includes a knowledge graph construction unit and a key feature identification unit;

[0042] The knowledge graph construction unit constructs a knowledge graph of preset users in each category based on the preset user feature data set and preset user portraits; the key feature identification unit performs association analysis on the knowledge graph, identifies the key features of the preset users in each category, and establishes a key feature set for each preset user.

[0043] The personalized health management module includes a real-time data acquisition unit, a similarity calculation unit, and a plan adjustment unit;

[0044] The real-time data acquisition unit acquires the data input by the user in real time, and performs corresponding analysis to construct a real-time user portrait; the similarity calculation unit calculates the similarity between the real-time user portrait and the preset user portrait, and determines the preset user category corresponding to the real-time user; the plan adjustment unit compares the real-time feature data with the key features corresponding to the preset health management plan list. If the real-time feature data is the same as the key features of other preset health management plans in the preset health management plan list other than the preset health management plan currently implemented by the real-time user, according to the correspondence between the other preset health management plans and the key features of the preset user, the currently implemented preset health plan is adjusted accordingly according to the other preset health management plans, thereby generating a personalized health management plan.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows: through the standardized processing and correlation analysis of multi-source historical data, the complex relationship between the health data of preset users can be revealed, thereby improving the accuracy and reliability of the analysis; using cluster analysis and feature extraction technology, a detailed user portrait can be constructed for each preset user to reflect their unique health status and needs, and this personalized portrait is helpful to formulate a more practical health management plan; by constructing a preset user knowledge graph, the relationship between the preset user feature data and the health management plan can be effectively integrated to help identify key features, thereby providing a scientific basis for health management; real-time user data input and analysis enable the health management plan to be adjusted in a timely manner according to the user's actual situation, ensuring the effectiveness and adaptability of the plan; the system can process information from different data sources, solve the problem of data format and quality differences, and improve the system's adaptability to the complexity and diversity of multi-source data; based on personalized health management plans, it can better meet the diversified needs of users and improve the overall effect and satisfaction of health management. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0047] Figure 1It is a schematic diagram of the health management system module based on multi-source data collection and analysis of the present invention. DETAILED DESCRIPTION

[0048] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0049] See also Figure 1 , the present invention provides a technical solution:

[0050] A health management system based on multi-source data collection and analysis, the system includes: data collection and processing module, user classification and portrait construction module, knowledge graph construction module, health management plan matching module and personalized health management module;

[0051] The data acquisition and processing module obtains multi-source historical data of preset users from multiple databases, uniformly converts the multi-source historical data of different formats and structures, extracts and standardizes the feature data of preset users, and thus forms a feature data set of preset users;

[0052] The user classification and portrait construction module performs correlation analysis on the feature data in the preset user feature data set, calculates the linear relationship between the features, uses a linear regression model to establish the relationship between the features, divides users into different categories through cluster analysis, and creates a corresponding preset user portrait for each category;

[0053] The knowledge graph construction module defines nodes and edges based on the preset user feature data set and preset user portrait, establishes the relationship between entities, and constructs the preset user knowledge graph. By traversing the knowledge graph, it identifies key nodes and key features, thereby establishing a key feature set for each preset user.

[0054] The health management program matching module uses the initial data input by the real-time user to construct a real-time user portrait, matches it with the preset user portrait, thereby obtaining the category corresponding to the real-time user, and outputs a list of preset health management programs corresponding to the preset user category according to the matching results;

[0055] The personalized health management module obtains multi-source real-time data of the real-time user based on the preset health management plan selected by the real-time user, extracts corresponding real-time feature data based on the multi-source real-time data, and compares the real-time feature data with the key features corresponding to the preset health management plan list. If the real-time feature data is the same as the key features of other preset health management plans in the preset health management plan list other than the preset health management plan currently implemented by the real-time user, according to the correspondence between the other preset health management plans and the key features of the preset user, the currently implemented preset health plan is adjusted accordingly according to the other preset health management plans, thereby generating a personalized health management plan.

[0056] The data acquisition and processing module includes a data acquisition unit and a data processing unit;

[0057] The data collection unit collects historical multi-source data of the preset user from the database; the data processing unit uniformly converts the multi-source historical data of different formats and structures, extracts and standardizes the characteristic data of the preset user, thereby forming a characteristic data set of the preset user;

[0058] The user classification and portrait building module includes a user classification unit and a user portrait creation unit;

[0059] The user classification performs cluster analysis on the preset user features in the preset user feature data set, and divides the preset users into several categories; the user portrait creation unit creates a weighted feature vector based on the preset user feature data set for each category of preset users, thereby creating a preset user portrait.

[0060] The knowledge graph construction module includes a knowledge graph construction unit and a key feature identification unit;

[0061] The knowledge graph construction unit constructs a knowledge graph of preset users in each category based on the preset user feature data set and preset user portraits; the key feature identification unit performs association analysis on the knowledge graph, identifies the key features of the preset users in each category, and establishes a key feature set for each preset user.

[0062] The personalized health management module includes a real-time data acquisition unit, a similarity calculation unit, and a plan adjustment unit;

[0063] The real-time data acquisition unit acquires the data input by the user in real time, and performs corresponding analysis to construct a real-time user portrait; the similarity calculation unit calculates the similarity between the real-time user portrait and the preset user portrait, and determines the preset user category corresponding to the real-time user; the plan adjustment unit compares the real-time feature data with the key features corresponding to the preset health management plan list. If the real-time feature data is the same as the key features of other preset health management plans in the preset health management plan list other than the preset health management plan currently implemented by the real-time user, according to the correspondence between the other preset health management plans and the key features of the preset user, the currently implemented preset health plan is adjusted accordingly according to the other preset health management plans, thereby generating a personalized health management plan.

[0064] A health management method based on multi-source data collection and analysis, the method comprising the following steps:

[0065] Step S100. Obtain multi-source historical data of all preset users from the database, perform corresponding screening and merging on the multi-source historical data of each preset user, thereby constructing a preset user feature data set; classify the preset users according to the preset user feature data set, and create a corresponding preset user portrait for each category of preset users;

[0066] Step S200. For each category of preset users, a preset user knowledge graph is constructed based on the preset user feature data set and the preset user portrait; the preset user knowledge graph and the preset user feature data set are combined for analysis, and key features of the preset users of the corresponding category are identified according to the analysis results;

[0067] Step S300. Obtain preset health management plans provided for preset users of different categories in the database, analyze the preset health management plans of each category with the key features of the corresponding preset users, and thus obtain the corresponding relationship between the preset health management plans and the key features of the corresponding preset users;

[0068] Step S400. The real-time user inputs initial data, builds a real-time user portrait based on the initial data, matches the real-time user portrait with the preset user portrait, and outputs a corresponding preset health management plan list based on the matching result; obtains multi-source real-time data of the real-time user, extracts the multi-source real-time data to identify the real-time user's real-time feature data, and analyzes it with the key features of the preset user to generate a personalized health management plan.

[0069] Step S100 includes:

[0070] S101. Obtain multi-source historical data of all preset users from the database, where the preset users refer to user groups stored in the database; the multi-source historical data refer to user health-related information collected from multiple different data sources; convert the multi-source historical data of the preset users into the same format, aggregate the multi-source historical data by the preset user ID, extract corresponding feature data from the multi-source historical data of each preset user, and perform standardization on the extracted feature data, and the specific standardization calculation formula is:

[0071] F i,j =(X i,j -μ j ) / σ j ,

[0072] Among them, F i,j represents the jth standardized feature data of preset user i, X i,j represents the jth original feature data of preset user i, μ j and σ j Respectively represent the mean and standard deviation of the j-th feature data;

[0073] In this embodiment, in the process of feature extraction and constructing feature data sets, different types of features can be calculated and processed accordingly. The following is a specific calculation method for physiological features, behavioral features, and health record features.

[0074] Physiological characteristics:

[0075] Weight, unit: kilogram (kg); Calculation: directly extract the user's weight record from the dataset.

[0076] Height, unit: centimeters (cm); Calculation: directly extract the user's height record from the dataset.

[0077] Body Mass Index (BMI), calculation: Calculate BMI by extracting weight and height.

[0078] Blood Pressure, unit: mmHg, calculation: usually recorded as systolic / diastolic, such as 120 / 80 mmHg, extracted from the dataset.

[0079] Behavioral characteristics:

[0080] Exercise Frequency, unit: times per week, calculation: extract the number of exercises in a week from the user's records.

[0081] Dietary Habits,Description: Record the user's dietary preferences (such as vegetarian, low sugar, high protein, etc.) and weekly food intake,Calculation: Different dietary habits can be quantified in a coded way, for example:

[0082] Vegetarian: 1,

[0083] Meat consumption: 0.

[0084] Health Records:

[0085] Medical History, Description: Record the diseases the user has suffered from (such as high blood pressure, diabetes, etc.); Calculation: A binary variable can be used to indicate whether a certain disease exists:

[0086] Hypertension: 1 (yes), 0 (no),

[0087] Diabetes: 1 (yes), 0 (no);

[0088] Allergy History, Description: Records the user's allergies to certain substances (such as pollen, seafood, etc.); Calculation: Binary variables can also be used to indicate whether a certain allergy exists:

[0089] Pollen allergy: 1 (yes), 0 (no),

[0090] Seafood allergy: 1 (yes), 0 (no).

[0091] Feature dataset construction:

[0092] After calculating the above features, these features can be integrated into a feature dataset.

[0093] S102. For each preset user's standardized feature data, correlation analysis is performed to obtain the linear relationship between the standardized feature data, and the value measuring the linear relationship is defined as the correlation coefficient R; the Pearson correlation coefficient is usually used to measure the linear relationship between two features, and a linear regression model is used to establish the corresponding linear relationship for feature data whose absolute value of the correlation coefficient R is equal to the threshold R0, and the linear regression model is expressed as: F i,k =β0+β1F i,j +ϵ, where β0 represents the intercept and β1 represents the feature F i,j The coefficient of , ϵ represents the error term; the regression coefficients β0 and β1 are calculated using the least squares method, and the linear regression model is used to i,k Update, and F' i,k =β0+β1F i,j; Traverse all the standardized feature data of the preset user to construct a preset user feature data set, and one element in the preset user feature data set corresponds to one preset user, and the number of elements in the preset user feature data set is equal to the number of preset users in the database;

[0094] S103. Perform cluster analysis on the preset users according to the preset user feature data set, divide the preset users into K categories, and create corresponding preset user portraits for the preset users in each category. The process of creating the preset user portraits is as follows:

[0095] For each category of preset users, the corresponding feature vector V is extracted from the preset user feature data set, the mean and standard deviation of each feature in the category are calculated based on the feature vector V, and normalized based on the standard deviation to obtain the feature weight W; based on the mean and weight of the clustering category, the preset user portrait U of each category is constructed, and the preset user portrait U is expressed as a weighted feature vector.

[0096] In this embodiment, the following characteristic data are assumed:

[0097] Characteristic data of Category 1 users:

[0098] User 1: [0.8,3.5,5] [0.8,3.5,5],

[0099] User 2: [0.9,4,4][0.9,4,4],

[0100] User 3: [0.7,3,6][0.7,3,6];

[0101] Mean calculation:

[0102] Feature 1:

[0103] μ1,1=(0.8+0.9+0.7) / 3=0.8,

[0104] Feature 2:

[0105] μ1,2=(3.5+4+3) / 3=3.5,

[0106] Feature 3:

[0107] μ1,3=(5+4+6) / 3=5;

[0108] Standard deviation calculation:

[0109] Feature 1: σ1,1≈0.13; Feature 2: σ1,2≈0.29; Feature 3: σ1,3≈0.38;

[0110] Feature weight calculation:

[0111] W1,j=σ1,j / (σ1,1+σ1,2+σ1,3)

[0112] User portrait construction:

[0113] U1=[W1,1·μ1,1,W1,2·μ1,2,W1,3·μ1,3].

[0114] Perform cluster analysis on preset users based on the preset user feature data set. The specific cluster analysis process is as follows:

[0115] According to the preset user feature data set, the feature data of the preset user is obtained, and the Euclidean distance is used as the initial distance metric to calculate the distance metric between all preset users, and the distance metric d(a,b)={Σk∈[1,n],(F a,k -F b,k )^2}^(1 / 2), where F a,k and F b,k Respectively represent the standardized values ​​of preset users a and b on the kth feature, and n represents the total number of features; hierarchical clustering is applied as the initial clustering, starting from each preset user, it is regarded as a separate cluster, and clusters whose distance metric d(a,b) is greater than or equal to the threshold d are gradually merged; based on the results of hierarchical clustering, a cluster number K0 of the initial clustering is selected, and K0 clusters are selected from the final clusters of the hierarchical clustering as the initial center points of K-means; the K-means clustering algorithm is executed, and after each iteration, the quality of the cluster is evaluated and the K value is dynamically adjusted in the following way. The specific analysis process is as follows:

[0116] Calculate the cluster variance sum WCSS, and the specific calculation formula of WCSS is:

[0117] WCSS=Σm∈[1,M],Σp∈Cm,Σk∈[1,n],[f(p,k)-μm,k]^2,

[0118] Where M represents the number of clusters, Cm represents the mth cluster, f(p,k) represents the value of the pth sample on the kth feature, and μm,k represents the mean of the mth cluster on the kth feature. If the reduction of WCSS is less than the threshold t in c consecutive iterations, the iteration is stopped. Otherwise, the iteration continues and the K value is adjusted. Based on the final K, the K-means clustering algorithm is re-run to divide the preset users into K categories.

[0119] Step S200 includes:

[0120] S201. For each category of preset users, based on the preset user feature data set and the preset user portrait, extract the preset user feature data as entities, extract the relationship between the preset user feature data in combination with the preset user portrait, use the entities as nodes of the knowledge graph, and use the relationship between the entities as edges of the knowledge graph, and combine the mean and weight of the clustering category in S103 to construct a preset user knowledge graph for each category;

[0121] S202. Obtain the preset user knowledge graph for each category, arbitrarily define a node as the initial node, start from the initial node, traverse the preset user knowledge graph, collect all reachable nodes J and paths P, and thus obtain the corresponding path set Pa; for each path in the path set, calculate the corresponding path weight w(path), and the value of w(path) is equal to the weight product of each feature in the path; based on the path weight w(path), define the key indicator G(J_h), and G(J_h)=ΣP∈Pa,[w(path) / |P|]·I(J_h∈P), where |P| represents the path length, I represents the indicator function, when the node J_h is in the path P, then I(J_h∈P)=1, otherwise I(J_h∈P)=0, and h represents the node number; judge the size relationship between the key indicator G(J_h) and the corresponding threshold G0, when G(J_h)>G0, define the corresponding node J_h as a key node, and the feature corresponding to the node J_h is the key feature.

[0122] Step S300 includes:

[0123] The health management plans corresponding to different categories of preset users are extracted from the database, and the corresponding health management plans are matched for each category of preset users; according to the key features corresponding to the preset users, a key feature set Tu is constructed for each preset user, and u represents the preset user number; the symmetric difference of the key feature set Tu of the preset users is calculated in turn, and the health management plans corresponding to the preset users are compared, and the symmetric difference result is associated with the comparison result, so as to obtain the corresponding relationship between the preset health management plan and the key features of the corresponding preset users.

[0124] Step S400 includes:

[0125] S401. The real-time user inputs initial data, and the initial data is subjected to the same analysis process as multi-source historical data, so as to construct a real-time user portrait U1 based on the initial data; the real-time user portrait U1 is subjected to similarity calculation with the preset user portrait U, and the corresponding preset user category is matched according to the similarity calculation result, and the one with the greatest similarity is taken as the matching result, and a list of preset health management plans corresponding to the preset user category is output according to the matching result;

[0126] S402. According to the list of preset health management plans, the real-time user selects a preset health management plan and implements the preset health management plan, obtains multi-source real-time data of the real-time user at selected time intervals, extracts corresponding real-time feature data based on the multi-source real-time data, and compares the real-time feature data with the key features corresponding to the list of preset health management plans. If the real-time feature data is the same as the key features of other preset health management plans in the list of preset health management plans other than the preset health management plan currently implemented by the real-time user, according to the correspondence between the other preset health management plans and the key features of the preset user, the currently implemented preset health plan is adjusted accordingly according to the other preset health management plans, thereby generating a personalized health management plan.

[0127] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

Claims

1. A health management method based on multi-source data collection and analysis, characterized in that: The method comprises the following steps: Step S100. Obtain multi-source historical data of all preset users from the database, perform corresponding screening and merging on the multi-source historical data of each preset user, thereby constructing a preset user feature data set; classify the preset users according to the preset user feature data set, and create a corresponding preset user portrait for each category of preset users; The step S100 includes: S101. Obtain multi-source historical data of all preset users from the database, where the preset users refer to user groups stored in the database; the multi-source historical data refer to user health-related information collected from multiple different data sources; convert the multi-source historical data of the preset users into the same format, aggregate the multi-source historical data by the preset user ID, extract corresponding feature data from the multi-source historical data of each preset user, and perform standardization on the extracted feature data, and the specific standardization calculation formula is: F i,j =(X i,j -m j ) / s j , Among them, F i,j represents the jth standardized feature data of preset user i, X i,j represents the jth original feature data of preset user i, μ j and σ j Respectively represent the mean and standard deviation of the j-th feature data; S102. For each preset user's standardized feature data, correlation analysis is performed to obtain a linear relationship between the standardized feature data, and the value measuring the linear relationship is defined as a correlation coefficient R; for feature data whose absolute value of the correlation coefficient R is equal to the threshold R0, a linear regression model is used to establish a corresponding linear relationship, and the linear regression model is expressed as: F i,k =β0+β1F i,j +ϵ, where F i,k represents the kth standardized feature data of the preset user i, β0 represents the intercept, and β1 represents the feature F i,j The coefficient of , ϵ represents the error term; the regression coefficients β0 and β1 are calculated using the least squares method, and the linear regression model is used to i,k Update, and F' i,k =β0+β1F i,j , where F' i,k Represents the kth standardized feature data of the preset user i after being updated by the linear regression model; traverses all the standardized feature data of the preset user to construct a preset user feature data set, and one element in the preset user feature data set corresponds to one preset user, and the number of elements in the preset user feature data set is equal to the number of preset users in the database; S103. Perform cluster analysis on the preset users according to the preset user feature data set, divide the preset users into K categories, and create corresponding preset user portraits for the preset users in each category. The process of creating the preset user portraits is as follows: For each category of preset users, the corresponding feature vector V is extracted from the preset user feature data set, the mean and standard deviation of each feature in the category are calculated based on the feature vector V, and normalized based on the standard deviation to obtain the feature weight W; based on the mean and weight of the clustered categories, the preset user portrait U of each category is constructed, and the preset user portrait U is expressed as a weighted feature vector; Step S200. For each category of preset users, a preset user knowledge graph is constructed based on the preset user feature data set and the preset user portrait; the preset user knowledge graph and the preset user feature data set are combined for analysis, and key features of the preset users of the corresponding category are identified according to the analysis results; The step S200 comprises: S201. For each category of preset users, based on the preset user feature data set and the preset user portrait, extract the preset user feature data as entities, extract the relationship between the preset user feature data in combination with the preset user portrait, use the entities as nodes of the knowledge graph, and use the relationship between the entities as edges of the knowledge graph, and combine the mean and weight of the clustering category in S103 to construct a preset user knowledge graph for each category; S202. Obtain the preset user knowledge graph of each category, arbitrarily define a node as the initial node, start from the initial node, traverse the preset user knowledge graph, collect all reachable nodes J and paths P, and thus obtain the corresponding path set Pa; for each path in the path set, calculate the corresponding path weight w(path), and the value of w(path) is equal to the weight product of each feature in the path; based on the path weight w(path), define the key indicator G(J_h), and G(J_h)=ΣP∈Pa,[w(path) / |P|]·I(J_h∈P), where |P| represents the path length, I represents the indicator function, when the node J_h is in the path P, then I(J_h∈P)=1, otherwise I(J_h∈P)=0, and h represents the node number; judge the size relationship between the key indicator G(J_h) and the corresponding threshold G0, when G(J_h)>G0, define the corresponding node J_h as a key node, and the feature corresponding to the node J_h is a key feature; Step S300. Obtain preset health management plans provided for preset users of different categories in the database, analyze the preset health management plans of each category with the key features of the corresponding preset users, and thus obtain the corresponding relationship between the preset health management plans and the key features of the corresponding preset users; The step S300 includes: Extract health management plans corresponding to preset users of different categories from the database, and correspond the corresponding health management plans to each category of preset users; construct a key feature set Tu for each preset user according to the key features corresponding to the preset users, and u represents the preset user number; calculate the symmetric difference of the key feature set Tu of the preset users in turn, compare the health management plans corresponding to the preset users, and associate the symmetric difference result with the comparison result, so as to obtain the corresponding relationship between the preset health management plan and the key features of the corresponding preset users; Step S400. The real-time user inputs initial data, builds a real-time user portrait based on the initial data, matches the real-time user portrait with the preset user portrait, and outputs a corresponding preset health management plan list based on the matching result; obtains multi-source real-time data of the real-time user, extracts the multi-source real-time data to identify the real-time user's real-time feature data, and analyzes it with the key features of the preset user to generate a personalized health management plan.

2. The health management method based on multi-source data collection and analysis according to claim 1 is characterized in that: The cluster analysis is performed on the preset users according to the preset user feature data set. The specific cluster analysis process is as follows: According to the preset user feature data set, the feature data of the preset user is obtained, and the Euclidean distance is used as the initial distance metric to calculate the distance metric between all preset users, and the distance metric d(a,b)={Σk∈[1,n],(F a,k -F b,k )^2}^(1 / 2), where F a,k and F b,k Respectively represent the standardized values ​​of preset users a and b on the kth feature, and n represents the total number of features; hierarchical clustering is applied as the initial clustering, starting from each preset user, it is regarded as a separate cluster, and clusters whose distance metric d(a,b) is greater than or equal to the threshold d are gradually merged; based on the results of hierarchical clustering, a cluster number K0 of the initial clustering is selected, and K0 clusters are selected from the final clusters of the hierarchical clustering as the initial center points of K-means; the K-means clustering algorithm is executed, and after each iteration, the quality of the cluster is evaluated and the K value is dynamically adjusted in the following way. The specific analysis process is as follows: Calculate the cluster variance sum WCSS, and the specific calculation formula of WCSS is: WCSS=Σm∈[1,M],Σp∈Cm,Σk∈[1,n],[f(p,k)-μm,k]^2, Where M represents the number of clusters, Cm represents the mth cluster, f(p,k) represents the value of the pth sample on the kth feature, and μm,k represents the mean of the mth cluster on the kth feature. If the reduction of WCSS is less than the threshold t in c consecutive iterations, the iteration is stopped. Otherwise, the iteration continues and the K value is adjusted. Based on the final K, the K-means clustering algorithm is re-run to divide the preset users into K categories.

3. The health management method based on multi-source data collection and analysis according to claim 1 is characterized in that: The step S400 includes: S401. The real-time user inputs initial data, and the initial data is subjected to the same analysis process as multi-source historical data, so as to construct a real-time user portrait U1 based on the initial data; the real-time user portrait U1 is subjected to similarity calculation with the preset user portrait U, and the corresponding preset user category is matched according to the similarity calculation result, and the one with the greatest similarity is taken as the matching result, and a list of preset health management plans corresponding to the preset user category is output according to the matching result; S402. According to the list of preset health management plans, the real-time user selects a preset health management plan and implements the preset health management plan, obtains multi-source real-time data of the real-time user at selected time intervals, extracts corresponding real-time feature data based on the multi-source real-time data, and compares the real-time feature data with the key features corresponding to the list of preset health management plans. If the real-time feature data is the same as the key features of other preset health management plans in the list of preset health management plans other than the preset health management plan currently implemented by the real-time user, according to the correspondence between the other preset health management plans and the key features of the preset user, the currently implemented preset health plan is adjusted accordingly according to the other preset health management plans, thereby generating a personalized health management plan.

4. A health management system based on multi-source data collection and analysis, applied to a health management method based on multi-source data collection and analysis as claimed in any one of claims 1 to 3, characterized in that: The system includes: a data collection and processing module, a user classification and portrait construction module, a knowledge graph construction module, a health management plan matching module and a personalized health management module; The data acquisition and processing module obtains multi-source historical data of preset users from multiple databases, uniformly converts the multi-source historical data of different formats and structures, extracts and standardizes the feature data of the preset users, thereby forming a feature data set of the preset users; The user classification and portrait construction module performs correlation analysis on the feature data in the preset user feature data set, calculates the linear relationship between the features, uses a linear regression model to establish the relationship between the features, divides the users into different categories through cluster analysis, and creates a corresponding preset user portrait for each category; The knowledge graph construction module defines nodes and edges based on a preset user feature data set and a preset user portrait, establishes relationships between entities, and constructs a preset user knowledge graph. By traversing the knowledge graph, key nodes and key features are identified, thereby establishing a key feature set for each preset user. The health management program matching module uses the initial data input by the real-time user to construct a real-time user portrait, matches it with the preset user portrait, thereby obtaining the category corresponding to the real-time user, and outputs a list of preset health management programs corresponding to the preset user category according to the matching result; The personalized health management module obtains multi-source real-time data of the real-time user based on the preset health management plan selected by the real-time user, extracts corresponding real-time feature data based on the multi-source real-time data, and compares the real-time feature data with the key features corresponding to the preset health management plan list. If the real-time feature data is the same as the key features of other preset health management plans in the preset health management plan list other than the preset health management plan currently implemented by the real-time user, according to the correspondence between the other preset health management plans and the key features of the preset user, the currently implemented preset health plan is adjusted accordingly according to the other preset health management plans, thereby generating a personalized health management plan.

5. The health management system based on multi-source data collection and analysis according to claim 4 is characterized in that: The data acquisition and processing module includes a data acquisition unit and a data processing unit; The data collection unit collects historical multi-source data of a preset user from a database; the data processing unit uniformly converts the multi-source historical data of different formats and structures, extracts and standardizes the feature data of the preset user, thereby forming a feature data set of the preset user; The user classification and portrait building module includes a user classification unit and a user portrait creation unit; The user classification performs cluster analysis on preset user features in the preset user feature data set to divide the preset users into several categories; the user portrait creation unit creates a weighted feature vector based on the preset user feature data set for each category of preset users, thereby creating a preset user portrait.

6. The health management system based on multi-source data collection and analysis according to claim 4 is characterized in that: The knowledge graph construction module includes a knowledge graph construction unit and a key feature identification unit; The knowledge graph construction unit constructs a preset user knowledge graph for each category based on a preset user feature data set and a preset user portrait; the key feature identification unit performs association analysis on the knowledge graph, identifies the key features of the preset users in each category, and establishes a key feature set for each preset user.

7. The health management system based on multi-source data collection and analysis according to claim 4 is characterized in that: The personalized health management module includes a real-time data acquisition unit, a similarity calculation unit and a plan adjustment unit; The real-time data acquisition unit acquires the data input by the user in real time and performs corresponding analysis to construct a real-time user portrait; The similarity calculation unit calculates the similarity between the real-time user portrait and the preset user portrait, and determines the preset user category corresponding to the real-time user; The plan adjustment unit compares the real-time feature data with the key features corresponding to the preset health management plan list. If the real-time feature data is the same as the key features of other preset health management plans in the preset health management plan list other than the preset health management plan currently implemented by the real-time user, according to the correspondence between the other preset health management plans and the key features of the preset user, the currently implemented preset health plan is adjusted accordingly according to the other preset health management plans, thereby generating a personalized health management plan.

Images