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

Abstract

This document provides a health management system and method based on multi-source data collection and analysis. [Solution] The health management system includes a data collection and processing module, a user classification and user profile construction module, a knowledge graph construction module, a health management solution matching module, and a personalized health management module. The data collection and processing module acquires multi-source historical data of users and constructs a user characteristic dataset. The user classification and user profile construction module categorizes users and creates corresponding user profiles. The knowledge graph construction module constructs a knowledge graph and key feature set based on the characteristic dataset and user profiles. The health management solution matching module matches users with a list of health management solutions. The personalized health management module generates personalized health management solutions.

Description

[Technical Field]

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

[0002] With the rapid development of information technology, health management has become a major concern in modern society. Traditional health management methods often rely on a single medical examination method, such as health checkup reports or medical questionnaires. These data sources are monolithic and do not fully reflect an individual's health status or its 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 great potential in the field of health management, making it possible to evaluate an individual's health status more accurately and comprehensively, and to provide personalized health management solutions.

[0003] However, many conventional health management systems utilize simple statistical methods in data analysis, making it difficult to deeply mine the underlying relationships and trends in the data. Furthermore, differences in data format and quality from different sources make it difficult for conventional health management systems to cope with the complexity and diversity of multi-source data, sometimes resulting in insufficient accuracy in analysis results. In addition, the health advice provided by conventional health management systems is often general, lacking specificity and personalization, and frequently based on single or limited health check indicators, ignoring individual differences and diverse needs, which can lead to insufficient health management effectiveness. [Overview of the project] [Problems that the invention aims to solve]

[0004] The present invention aims to provide a health management system and method based on multi-source data collection and analysis in order to solve the problems in the background art described above.

[0005] To solve the above technical problems, the present invention provides the following technical means. [Means for solving the problem]

[0006] The health management method based on multi-source data collection and analysis includes the following steps S100 to S400: In step S100, all multi-source historical data for a given user is retrieved from the database, corresponding selection and integration are performed on the multi-source historical data for each given user to construct a characteristic dataset for the given user, the given user is classified based on the characteristic dataset for the given user, and a corresponding given user profile is created for the given user in each category. In step S200, for a given user in each category, a knowledge graph of that user is constructed based on a given user characteristic dataset and a given user profile. The knowledge graph of the given user and the given user characteristic dataset are then combined and analyzed, and based on the analysis results, the key characteristics of the given user in the corresponding category are identified. In step S300, predetermined health management solutions provided to predetermined users of different categories in the database are retrieved, and the correspondence between predetermined health management solutions and the key characteristics of the corresponding predetermined users is obtained by comparing and analyzing the predetermined health management solutions for each category with the key characteristics of the corresponding predetermined users. In step S400, the real-time user inputs initial data, a real-time user profile is constructed based on the initial data, the real-time user profile is matched with a predetermined user profile, a corresponding predetermined health management solution list is output based on the matching result, multi-source real-time data of the real-time user is acquired, the multi-source real-time data is extracted to identify the real-time feature data of the real-time user, and a personalized health management solution is generated by analyzing it in comparison with the key features of the predetermined user.

[0007] Furthermore, step S100 includes the following steps S101 to S103: In step S101, multi-source history data for all specified users is retrieved from the database. The specified users refer to user groups stored in the database, and the multi-source history data refers to user health-related information collected from multiple different data sources. The multi-source history data for the specified users is then uniformly converted to the same format, the multi-source history data is aggregated by a specified user ID, corresponding feature data is extracted from each specified user's multi-source history data, and the extracted feature data is subjected to standardization. The specific calculation formula for the standardization process is as follows: JPEG2026082605000002.jpg10166 Here, F i,j This is the j-th standardized feature data of a given user i. X i,j This is the j-th original feature data of a given user i. μ j and σ j These represent the mean and standard deviation of the j-th feature data, respectively. In step S102, correlation analysis is performed on the standardized feature data for each predetermined user to obtain a linear relationship between the standardized feature data. The value used to evaluate this linear relationship is defined as the correlation coefficient R. Typically, Pearson's correlation coefficient is used to evaluate the linear relationship between two features. For feature data where the absolute value of the correlation coefficient R is equal to the threshold R0, a linear regression model is used to establish the corresponding linear relationship. The linear regression model is shown in the following equation: JPEG2026082605000003.jpg11166 Here, β0 is the intercept. β1 is characteristic F i,j coefficient ε is the error term The regression coefficients β0 and β1 are calculated using the least squares method, and the feature F is calculated using the following equation with a linear regression model. i,k Update, JPEG2026082605000004.jpg10167 By traversing all standardized feature data of a given user, a feature dataset for that given user is constructed, one element in the feature dataset for that given user corresponds to one given user, and the number of elements in the feature dataset for that given user is equal to the number of given users in the database. In step S103, clustering analysis is performed on a given user based on a given user characteristic dataset, the given user is divided into K categories, and a corresponding given user profile is created for a given user in each category. The process for creating the given user profile is as follows: For each designated user in each category, a corresponding feature vector V is extracted from the designated user feature dataset. Based on the feature vector V, the mean and standard deviation of each feature within the category are calculated. Feature weights W are obtained by normalizing based on the standard deviation. Based on the mean and weight of the clustering categories, a designated user profile U is constructed for each category, and this designated user profile U is shown as a weighted feature vector.

[0008] Furthermore, clustering analysis is performed on the specified users based on a predetermined user characteristic dataset. The specific clustering analysis process is as follows: Based on a given user feature dataset, feature data for a given user is obtained, Euclidean distance is used as the initial distance metric, and distance metrics are calculated between pairs of points for all given users, with the distance metric being as follows: JPEG2026082605000005.jpg10167 Here, F a,k and F b,k These are the standardized values ​​of given users a and b in the k-th feature, respectively. n is the total number of features. Hierarchical clustering is applied as the initial clustering, each designated user is treated as a single cluster, clusters whose distance metric d(a,b) is greater than or equal to threshold d are merged stepwise, a cluster count K0 is selected for one initial clustering based on the results of hierarchical clustering, K0 clusters are selected from the final cluster of hierarchical clustering as the initial center points for K-means, the K-means clustering algorithm is executed, and after each iteration, the cluster quality is evaluated and the K value is dynamically adjusted as follows. The specific analysis process is as follows: The intra-cluster sum of squares (WCSS) is calculated, and the specific formula for calculating WCSS is as follows: JPEG2026082605000006.jpg14166 Here, M is the number of clusters. Cm is the mth cluster f(p,k) is the value of the p-th sample in the k-th feature. μm,k is the mean value of the m-th cluster in the k-th feature. In c consecutive iterations, if the decrease in WCSS is less than a threshold t, the iteration is stopped; otherwise, the iteration continues to adjust the K value, and based on the finally determined K, the K-means clustering algorithm is re-executed to divide a given user into K categories.

[0009] Furthermore, step S200 includes the following steps S201 to S202: In step S201, for a given user in each category, the given user characteristic data is extracted as entities based on a given user characteristic dataset and a given user profile, the relationships between the given user characteristic data are extracted by combining the given user profiles, the entities are made into nodes in the knowledge graph, the relationships between entities are made into edges in the knowledge graph, and the knowledge graph for the given user is constructed for each category by combining the mean and weight of the clustering categories in step S103. In step S202, the knowledge graph of a predetermined user in each category is obtained, any node is defined as the initial node, starting from the initial node, the knowledge graph of the predetermined user is traversed, all reachable nodes J and paths P are collected to obtain the corresponding path set Pa. For each path in the path set, the corresponding path weight w(path) is calculated, and the value of w(path) is equal to the product of the weights of each feature in the path. Based on the path weight w(path), the importance index G(J_h) is defined by the following formula: JPEG2026082605000007.jpg11165Here, |P| is the path length I is the indicator function When the node J_h is in the path P, I(J_h∈P)=1; otherwise, I(J_h∈P)=0. h indicates the node number. The magnitude relationship between the importance index G(J_h) and the corresponding threshold G0 is judged. If G(J_h)>G0, the corresponding node J_h is defined as the key node, and the feature corresponding to the node J_h is defined as the key feature.

[0010] Furthermore, step S300 includes extracting health management solutions corresponding to predetermined users in different categories from the database, associating the corresponding health management solutions with the predetermined users in each category, and constructing a key feature set Tu for each predetermined user based on the key features corresponding to the predetermined users, where u indicates the predetermined user number, and sequentially obtaining the symmetric difference for the key feature set Tu of the predetermined user, comparing the corresponding health management solutions of the predetermined user, and associating the result of the symmetric difference with the comparison result to obtain the correspondence between the predetermined health management solution and the key features of the corresponding predetermined user.

[0011] Furthermore, step S400 includes the following steps S401 to S402 In step S401, the real-time user inputs initial data, and an analysis process similar to that for multi-source historical data is performed on the initial data to construct a real-time user profile U1 based on the initial data. The similarity between the real-time user profile U1 and a predetermined user profile U is calculated, and based on the similarity calculation results, it is matched with the corresponding predetermined user category. The one with the highest similarity is selected as the matching result, and based on the matching result, a predetermined list of health management solutions corresponding to the predetermined user category is output. In step S402, based on a predetermined list of health management solutions, the real-time user selects one predetermined health management solution, implements the predetermined health management solution, acquires multi-source real-time data of the real-time user for each selected period, extracts corresponding real-time feature data based on the multi-source real-time data, compares the real-time feature data with the key features corresponding to the predetermined list of health management solutions, and if the real-time feature data is the same as the key features of other predetermined health management solutions in the predetermined list of health management solutions other than the predetermined health management solution currently being implemented by the real-time user, a personalized health management solution is generated by appropriately adjusting the predetermined health management solution currently being implemented according to the other predetermined health management solutions based on the correspondence between the other predetermined health management solutions and the key features of the predetermined user.

[0012] A health management system based on multi-source data collection and analysis, the system including a data collection and processing module, a user classification and user profile construction module, a knowledge graph construction module, a health management solution matching module, and a personalized health management module. The data collection and processing module obtains multi-source historical data of a predetermined user from multiple databases, uniformly converts multi-source historical data with different formats and structures, extracts and standardizes the feature data of the predetermined user, and constructs a feature dataset of the predetermined user. The user classification and user profile construction module performs a correlation analysis on the feature data in the feature dataset of the predetermined user, calculates the linear relationship between features, establishes the relationship between features using a linear regression model, divides users into different categories through clustering analysis, and creates corresponding predetermined user profiles for each category. The knowledge graph construction module defines nodes and edges based on the feature dataset of the predetermined user and the predetermined user profile, establishes the relationship between entities, constructs the knowledge graph of the predetermined user, and traverses the knowledge graph to identify key nodes and key features, thereby constructing a key feature set for each predetermined user. The health management solution matching module constructs a real-time user profile using the initial data input by the real-time user, matches it with the predetermined user profile, obtains the category corresponding to the real-time user, and outputs a list of predetermined health management solutions corresponding to the predetermined user category based on the matching result. The personalized health management module acquires multi-source real-time data of a real-time user based on a predetermined health management solution selected by the real-time user, extracts corresponding real-time feature data based on the multi-source real-time data, compares the real-time feature data with key features corresponding to a predetermined list of health management solutions, and if the real-time feature data is the same as the key features of other predetermined health management solutions in the predetermined list of health management solutions other than the predetermined health management solution currently being used by the real-time user, it generates a personalized health management solution by appropriately adjusting the predetermined health management solution currently being used according to the other predetermined health management solutions, based on the correspondence between the other predetermined health management solutions and the predetermined user's key features.

[0013] Furthermore, the data acquisition and processing module includes a data acquisition unit and a data processing unit. The data collection unit collects multi-source historical data for a given user from a database, and the data processing unit unifies the multi-source historical data in different formats and structures, extracts characteristic data for the given user, and standardizes it to construct a characteristic dataset for the given user. The user classification / user profile construction module includes a user classification unit and a user profile creation unit. The user classification unit performs clustering analysis on predetermined user features within a predetermined user feature dataset to divide the predetermined users into multiple categories. The user profile creation unit then creates a predetermined user profile for each predetermined user in each category by creating weighted feature vectors based on the predetermined user feature dataset.

[0014] Furthermore, the knowledge graph construction module includes a knowledge graph construction unit and a key feature identification unit. The knowledge graph construction unit constructs a knowledge graph for a given user in each category based on a given user characteristic dataset and a given user profile. The key feature identification unit performs relationship analysis on the knowledge graph to identify the key features of a given user in each category and constructs a key feature set for each given user.

[0015] The personalized health management module includes a real-time data collection unit, a similarity calculation unit, and a solution adjustment unit. The real-time data collection unit acquires data entered by the user in real time, performs corresponding analysis to construct a real-time user profile, the similarity calculation unit calculates the similarity between the real-time user profile and a predetermined user profile, and determines a predetermined user category corresponding to the real-time user, and the solution adjustment unit compares the real-time feature data with key features corresponding to a predetermined list of health management solutions, and if the real-time feature data is the same as the key features of other predetermined health management solutions in the predetermined list of health management solutions other than the predetermined health management solution currently being used by the real-time user, the solution adjustment unit generates a personalized health management solution by appropriately adjusting the predetermined health management solution currently being used according to the other predetermined health management solutions based on the correspondence between the other predetermined health management solutions and the key features of the predetermined user. [Effects of the Invention]

[0016] Compared to conventional technologies, the beneficial effects of the present invention include: the ability to reveal complex relationships between health data of a given user through standardization processing and correlation analysis of multi-source historical data, thereby improving the accuracy and reliability of the analysis; the ability to construct a detailed user profile for each given user through clustering analysis and feature extraction techniques, reflecting their unique health status and needs; such personalized user profiles contribute to the formulation of more practical health management solutions; the ability to effectively integrate the relationship between a given user's characteristic data and health management solutions by constructing a knowledge graph for a given user, helping to identify key features and providing a scientific basis for health management; the ability to timely adjust health management solutions according to the user's actual situation through real-time input and analysis of user data, ensuring the effectiveness and adaptability of the solutions; the ability of the system to process information from different data sources, solving the problem of differences in data format and quality, improving the system's ability to adapt to the complexity and diversity of multi-source data, and better meeting the diverse needs of users based on personalized health management solutions, thereby improving the overall effectiveness and satisfaction of health management. [Brief explanation of the drawing]

[0017] The drawings included herein, which constitute part of this specification, are intended to be interpreted together with the embodiments of the invention and do not limit the invention.

[0018] [Figure 1] This is a block diagram of the health management system based on multi-source data collection and analysis according to the present invention. [Modes for carrying out the invention]

[0019] In the following, the technical solutions in embodiments of the present invention will be clearly and completely described with reference to the drawings of the embodiments. It should be noted that the embodiments described are only some embodiments of the present invention, not all embodiments.

[0020] All other embodiments that can be obtained by those skilled in the art without creative work based on the embodiments of the present invention are all within the scope of the protection of the present invention.

[0021] As shown in Figure 1, the present invention provides the following technical means.

[0022] A health management system based on multi-source data collection and analysis, the system including a data collection and processing module, a user classification and user profile construction module, a knowledge graph construction module, a health management solution matching module, and a personalized health management module. The data collection and processing module acquires multi-source historical data for a given user from multiple databases, unifies the multi-source historical data in different formats and structures, extracts characteristic data for the given user, and standardizes it to construct a characteristic dataset for the given user. The user classification and user profile construction module performs correlation analysis on feature data within a given user feature dataset, calculates linear relationships between features, establishes relationships between features using a linear regression model, divides users into different categories using clustering analysis, and creates a corresponding predetermined user profile for each category. The knowledge graph construction module defines nodes and edges based on a given user feature dataset and a given user profile, establishes relationships between entities, constructs a knowledge graph for a given user, and builds a key feature set for each given user by traversing the knowledge graph and identifying key nodes and key features. The health management solution matching module constructs a real-time user profile using initial data entered by the real-time user, matches it with a predetermined user profile to obtain a category corresponding to the real-time user, and outputs a predetermined list of health management solutions corresponding to the predetermined user category based on the matching results. The personalized health management module acquires multi-source real-time data of a real-time user based on a predetermined health management solution selected by the real-time user, extracts corresponding real-time feature data based on the multi-source real-time data, compares the real-time feature data with key features corresponding to a predetermined list of health management solutions, and if the real-time feature data is the same as the key features of other predetermined health management solutions in the predetermined list of health management solutions other than the predetermined health management solution currently being used by the real-time user, it generates a personalized health management solution by appropriately adjusting the predetermined health management solution currently being used according to the other predetermined health management solutions, based on the correspondence between the other predetermined health management solutions and the predetermined user's key features.

[0023] The data collection and processing module includes a data collection unit and a data processing unit. The data collection unit collects multi-source historical data for a given user from a database, and the data processing unit unifies the multi-source historical data in different formats and structures, extracts characteristic data for the given user, and standardizes it to construct a characteristic dataset for the given user. The user classification / user profile construction module includes a user classification unit and a user profile creation unit. The user classification unit performs clustering analysis on predetermined user features within a predetermined user feature dataset to divide the predetermined users into multiple categories. The user profile creation unit then creates a predetermined user profile for each predetermined user in each category by creating weighted feature vectors based on the predetermined user feature dataset.

[0024] The knowledge graph construction module includes a knowledge graph construction unit and a key feature identification unit. The knowledge graph construction unit constructs a knowledge graph for a given user in each category based on a given user characteristic dataset and a given user profile. The key feature identification unit performs relationship analysis on the knowledge graph to identify the key features of a given user in each category and constructs a key feature set for each given user.

[0025] The personalized health management module includes a real-time data collection unit, a similarity calculation unit, and a solution adjustment unit. The real-time data collection unit acquires data entered by the user in real time, performs corresponding analysis to construct a real-time user profile, the similarity calculation unit calculates the similarity between the real-time user profile and a predetermined user profile, and determines a predetermined user category corresponding to the real-time user, and the solution adjustment unit compares the real-time feature data with key features corresponding to a predetermined list of health management solutions, and if the real-time feature data is the same as the key features of other predetermined health management solutions in the predetermined list of health management solutions other than the predetermined health management solution currently being used by the real-time user, the solution adjustment unit generates a personalized health management solution by appropriately adjusting the predetermined health management solution currently being used according to the other predetermined health management solutions based on the correspondence between the other predetermined health management solutions and the key features of the predetermined user.

[0026] The health management method based on multi-source data collection and analysis includes the following steps S100 to S400: In step S100, all multi-source historical data for a given user is retrieved from the database, corresponding selection and integration are performed on the multi-source historical data for each given user to construct a characteristic dataset for the given user, the given user is classified based on the characteristic dataset for the given user, and a corresponding given user profile is created for the given user in each category. In step S200, for a given user in each category, a knowledge graph of that user is constructed based on a given user characteristic dataset and a given user profile. The knowledge graph of the given user and the given user characteristic dataset are then combined and analyzed, and based on the analysis results, the key characteristics of the given user in the corresponding category are identified. In step S300, predetermined health management solutions provided to predetermined users of different categories in the database are retrieved, and the correspondence between predetermined health management solutions and the key characteristics of the corresponding predetermined users is obtained by comparing and analyzing the predetermined health management solutions for each category with the key characteristics of the corresponding predetermined users. In step S400, the real-time user inputs initial data, a real-time user profile is constructed based on the initial data, the real-time user profile is matched with a predetermined user profile, a corresponding predetermined health management solution list is output based on the matching result, multi-source real-time data of the real-time user is acquired, the multi-source real-time data is extracted to identify the real-time feature data of the real-time user, and a personalized health management solution is generated by analyzing it in comparison with the key features of the predetermined user.

[0027] Step S100 includes the following steps S101 to S103: In step S101, multi-source history data for all specified users is retrieved from the database. The specified users refer to user groups stored in the database, and the multi-source history data refers to user health-related information collected from multiple different data sources. The multi-source history data for the specified users is then uniformly converted to the same format, the multi-source history data is aggregated by a specified user ID, corresponding feature data is extracted from each specified user's multi-source history data, and the extracted feature data is subjected to standardization. The specific calculation formula for the standardization process is as follows: JPEG2026082605000008.jpg10166 Here, F i,j is the j-th standardized feature data of a predetermined user i X i,j is the j-th original feature data of a predetermined user i μ j and σ j are the mean value and standard deviation of the j-th feature data, respectively In this embodiment, in the process of feature extraction and construction of the feature dataset, corresponding calculations and processes can be performed for different types of features.

[0028] The following are specific calculation methods for physiological features, behavioral features, and health record features.

[0029] Physiological features: Weight, unit: kilogram (kg), calculation: directly extract the user's weight record from the dataset.

[0030] Height, unit: cm, calculation: directly extract the user's height record from the dataset.

[0031] Body Mass Index (BMI) is calculated based on the extracted weight and height.

[0032] Blood Pressure, unit: millimeter of mercury (mmHg), calculation: usually recorded as systolic blood pressure / diastolic blood pressure, for example, 120 / 80 mmHg, and extracted from the dataset.

[0033] Behavioral features: Exercise Frequency, unit: number of times per week, calculation: extract the number of weekly exercise times from the user's record.

[0034] Dietary Habits: Description: Records the user's dietary preferences (e.g., vegetarian, low-sugar, high-protein) and weekly food intake. Calculation: Different dietary habits can be quantified using a coding method, for example, Plain food: 1, Meat eating: 0.

[0035] Health record: Medical History: Description: Records the diseases the user has suffered from (e.g., hypertension, diabetes, etc.). Calculation: A binary variable can indicate the presence or absence of a certain disease, for example, Hypertension: 1 (present), 0 (absent) Diabetes: 1 (present), 0 (absent).

[0036] Allergy History: Description: Records the user's allergy status to certain substances (e.g., pollen, seafood, etc.). Calculation: Similarly, it can indicate the presence or absence of a certain allergy using binary variables, for example, Pollen allergy: 1 (Yes), 0 (No) Seafood allergies: 1 (Yes), 0 (No).

[0037] Feature dataset construction: After calculating the above features, these features can be integrated into a single feature dataset.

[0038] In step S102, correlation analysis is performed on the standardized feature data for each predetermined user to obtain a linear relationship between the standardized feature data. The value used to evaluate this linear relationship is defined as the correlation coefficient R. Typically, Pearson's correlation coefficient is used to evaluate the linear relationship between two features. For feature data where the absolute value of the correlation coefficient R is equal to the threshold R0, a linear regression model is used to establish the corresponding linear relationship. The linear regression model is given by the following equation: JPEG2026082605000009.jpg11166 Here, β0 is the intercept. β1 is characteristic F i,j coefficient ε is the error term The regression coefficients β0 and β1 are calculated using the least squares method, and the feature F is calculated using the following equation with a linear regression model. i,k Update, JPEG2026082605000010.jpg10167 By traversing all standardized feature data of a given user, a feature dataset for that given user is constructed, one element in the feature dataset for that given user corresponds to one given user, and the number of elements in the feature dataset for that given user is equal to the number of given users in the database. In step S103, clustering analysis is performed on a given user based on a given user characteristic dataset, the given user is divided into K categories, and a corresponding given user profile is created for a given user in each category. The process for creating the given user profile is as follows: For each designated user in each category, a corresponding feature vector V is extracted from the designated user feature dataset. Based on the feature vector V, the mean and standard deviation of each feature within the category are calculated. Feature weights W are obtained by normalizing based on the standard deviation. Based on the mean and weight of the clustering categories, a designated user profile U is constructed for each category, and this designated user profile U is shown as a weighted feature vector.

[0039] In this embodiment, the following characteristic data is assumed.

[0040] User characteristic data for Category 1: User 1: [0.8,3.5,5][0.8,3.5,5] User 2: [0.9,4,4][0.9,4,4] User 3: [0.7,3,6][0.7,3,6] Calculating the average: Feature 1: μ1,1 = (0.8 + 0.9 + 0.7) / 3 = 0.8 Feature 2: μ1,2 = (3.5 + 4 + 3) / 3 = 3.5 Feature 3: μ1,3 = (5 + 4 + 6) / 3 = 5 Standard deviation calculation: Feature 1: σ1,1 ≈ 0.13, Feature 2: σ1,2 ≈ 0.29, Feature 3: σ1,3 ≈ 0.38 Feature weight calculation: W1,j=σ1,j / (σ1,1+σ1,2+σ1,3) User profile development: U1=[W1,1·μ1,1,W1,2·μ1,2,W1,3·μ1,3].

[0041] Based on a specified user characteristic dataset, clustering analysis is performed on the specified users. The specific clustering analysis process is as follows: Based on a given user feature dataset, feature data for a given user is obtained, Euclidean distance is used as the initial distance metric, and distance metrics are calculated between pairs of points for all given users, with the distance metric being as follows: JPEG2026082605000011.jpg10167 Here, F a,k and F b,k These are the standardized values ​​of given users a and b in the k-th feature, respectively. n is the total number of features. Hierarchical clustering is applied as the initial clustering, each designated user is treated as a single cluster, clusters whose distance metric d(a,b) is greater than or equal to threshold d are merged stepwise, a cluster count K0 is selected for one initial clustering based on the results of hierarchical clustering, K0 clusters are selected from the final cluster of hierarchical clustering as the initial center points for K-means, the K-means clustering algorithm is executed, and after each iteration, the cluster quality is evaluated and the K value is dynamically adjusted as follows. The specific analysis process is as follows: The intra-cluster sum of squares (WCSS) is calculated, and the specific formula for calculating WCSS is as follows: JPEG2026082605000012.jpg14166 Here, M is the number of clusters. Cm is the mth cluster f(p,k) is the value of the p-th sample in the k-th feature. μm,k is the mean value of the m-th cluster in the k-th feature. In c consecutive iterations, if the decrease in WCSS is less than a threshold t, the iteration is stopped; otherwise, the iteration continues to adjust the K value, and based on the finally determined K, the K-means clustering algorithm is re-executed to divide a given user into K categories.

[0042] Step S200 includes the following steps S201 to S202: In step S201, for a given user in each category, the given user characteristic data is extracted as entities based on a given user characteristic dataset and a given user profile, the relationships between the given user characteristic data are extracted by combining the given user profiles, the entities are made into nodes in the knowledge graph, the relationships between entities are made into edges in the knowledge graph, and the knowledge graph for the given user is constructed for each category by combining the mean and weight of the clustering categories in step S103. In step S202, the knowledge graph of a given user in each category is obtained, one node is defined as the initial node, and starting from the initial node, the knowledge graph of the given user is traversed, all reachable nodes J and paths P are collected, and the corresponding path set Pa is obtained. For each path in the path set, the corresponding path weight w(path) is calculated, and the value of w(path) is equal to the product of the weights of each feature in the path. Based on the path weight w(path), the importance index G(J_h) is defined by the following formula: JPEG2026082605000013.jpg11165 Here, |P| is the path length I is the indicator function If node J_h is on path P, I(J_h∈P)=1; otherwise, I(J_h∈P)=0, where h is the node number. The relationship between the importance index G(J_h) and the corresponding threshold G0 is determined. If G(J_h)>G0, the corresponding node J_h is defined as a key node, and the feature corresponding to node J_h is defined as a key feature.

[0043] Step S300 is, The process includes the steps of: extracting health management solutions corresponding to specific users in different categories from a database; associating the corresponding health management solutions with specific users in each category; and constructing a key feature set Tu for each specific user based on the key features corresponding to the specific user, wherein u indicates a specific user number; and sequentially calculating symmetric differences for the key feature set Tu of the specific user, comparing the corresponding health management solutions for the specific user, and relating the symmetric difference results with the comparison results to obtain a correspondence between the specific health management solution and the key features of the corresponding specific user.

[0044] Step S400 includes the following steps S401 to S402: In step S401, the real-time user inputs initial data, and an analysis process similar to that for multi-source historical data is performed on the initial data to construct a real-time user profile U1 based on the initial data. The similarity between the real-time user profile U1 and a predetermined user profile U is calculated, and based on the similarity calculation results, it is matched with the corresponding predetermined user category. The one with the highest similarity is selected as the matching result, and based on the matching result, a predetermined list of health management solutions corresponding to the predetermined user category is output. In step S402, based on a predetermined list of health management solutions, the real-time user selects one predetermined health management solution, implements the predetermined health management solution, acquires multi-source real-time data of the real-time user for each selected period, extracts corresponding real-time feature data based on the multi-source real-time data, compares the real-time feature data with the key features corresponding to the predetermined list of health management solutions, and if the real-time feature data is the same as the key features of other predetermined health management solutions in the predetermined list of health management solutions other than the predetermined health management solution currently being implemented by the real-time user, a personalized health management solution is generated by appropriately adjusting the predetermined health management solution currently being implemented according to the other predetermined health management solutions based on the correspondence between the other predetermined health management solutions and the key features of the predetermined user.

[0045] To those skilled in the art, it will be apparent that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be realized in other specific forms without departing from the spirit or fundamental features of the invention. Therefore, in all respects, the embodiments should be considered illustrative and non-limiting, and the scope of protection of the invention is limited by the appended claims, not limited to the above description; thus, all variations within the meaning and scope of equivalent requirements in the claims are intended to be included in the invention. Any reference numeral in the claims should not be considered limiting to the relevant claims.

Claims

1. A health management method based on multi-source data collection and analysis, the method comprising the following steps S100 to S400: In step S100, all multi-source history data for a given user is obtained from the database, corresponding selection and integration are performed on the multi-source history data for each given user to construct a characteristic dataset for the given user, the given user is classified based on the characteristic dataset for the given user, and a corresponding given user profile is created for the given user in each category. The aforementioned step S100 includes the following steps S101 to S103: In step S101, multi-source history data for all specified users is obtained from the database. The specified users refer to user groups stored in the database, and the multi-source history data refers to user health-related information collected from multiple different data sources. The multi-source history data for the specified users is uniformly converted into the same format, the multi-source history data is grouped by a specified user ID, corresponding feature data is extracted from the multi-source history data for each specified user, and standardization processing is performed on the extracted feature data. The specific calculation formula for the standardization processing is as follows: Here, F ｉ，ｊ This is the j-th standardized feature data of a given user i. X ｉ，ｊ This is the j-th original feature data of a given user i. μ ｊ and σ ｊ These represent the mean and standard deviation of the j-th feature data, respectively. In step S102, correlation analysis is performed on the standardized feature data for each predetermined user to obtain linear relationships between the standardized feature data. A value for evaluating these linear relationships is defined as the correlation coefficient R. For feature data where the absolute value of the correlation coefficient R is equal to the threshold R0, a corresponding linear relationship is established using a linear regression model. The linear regression model is given by the following equation: Here, β0 is the intercept β1 is characteristic F ｉ，ｊ coefficient ε is the error term The regression coefficients β0 and β1 are calculated using the least squares method, and the feature F is calculated using the following equation with a linear regression model. ｉ，ｋ Update, Here, F' ｉ，ｋ This shows the k-th standardized feature data of a given user i, updated by a linear regression model. A feature dataset for a given user is constructed by traversing all the standardized feature data of that given user. Each element in the feature dataset for a given user corresponds to one given user, and the number of elements in the feature dataset for a given user is equal to the number of given users in the database. In step S103, clustering analysis is performed on a predetermined user based on a predetermined user characteristic dataset, the predetermined user is divided into K categories, and a corresponding predetermined user profile is created for each predetermined user in each category. The process for creating the predetermined user profile is as follows: For each category of a given user, a corresponding feature vector V is extracted from the given user feature dataset. Based on the feature vector V, the mean and standard deviation of each feature within the category are calculated. Feature weights W are obtained by normalizing based on the standard deviation. Based on the mean and weight of the clustering categories, a given user profile U is constructed for each category, and this given user profile U is shown as a weighted feature vector. In step S200, for a given user in each category, a knowledge graph of the given user is constructed based on a given user characteristic dataset and a given user profile. The knowledge graph of the given user and the given user characteristic dataset are then combined and analyzed. Based on the analysis results, the key characteristics of the given user in the corresponding category are identified. Step S200 includes the following steps S201 to S202: In step S201, for a predetermined user in each category, the predetermined user characteristic data is extracted as entities based on a predetermined user characteristic dataset and a predetermined user profile, the relationships between the predetermined user characteristic data are extracted by combining the predetermined user profiles, the entities are made into nodes in the knowledge graph, the relationships between entities are made into edges in the knowledge graph, and the knowledge graph for the predetermined user is constructed for each category by combining the mean and weight of the clustering categories in step S103. In step S202, the knowledge graph of a predetermined user in each category is obtained, one node is defined as the initial node, the knowledge graph of the predetermined user is traversed starting from the initial node, all reachable nodes J and paths P are collected, the corresponding path set Pa is obtained, the corresponding path weight w(path) is calculated for each path in the path set, and the value of w(path) is equal to the product of the weights of each feature in the path, and based on the path weight w(path), the importance index G(J_h) is defined by the following formula: Here, |P| is the path length. I is the indicator function If node J_h is on path P, then I(J_h∈P) = 1; otherwise, I(J_h∈P) = 0, where h is the node number. The relationship between the importance index G(J_h) and the corresponding threshold G0 is determined. If G(J_h) > G0, the corresponding node J_h is defined as a key node, and the feature corresponding to node J_h is defined as a key feature. In step S300, predetermined health management solutions provided to predetermined users of different categories in the database are retrieved, and the correspondence between predetermined health management solutions and the key characteristics of the corresponding predetermined users is obtained by comparing and analyzing the predetermined health management solutions for each category with the key characteristics of the corresponding predetermined users. The aforementioned step S300 is, The steps include: extracting health management solutions corresponding to predetermined users of different categories from a database, associating the corresponding health management solutions with predetermined users in each category; constructing a key feature set Tu for each predetermined user based on the key features corresponding to the predetermined user, wherein u indicates a predetermined user number; and sequentially calculating symmetric differences for the key feature set Tu of the predetermined user, comparing the corresponding health management solutions for the predetermined user, and relating the symmetric difference results with the comparison results to obtain a correspondence relationship between the predetermined health management solution and the key features of the corresponding predetermined user. Step S400 is characterized by the following steps: inputting initial data from a real-time user; constructing a real-time user profile based on the initial data; matching the real-time user profile with a predetermined user profile; outputting a corresponding predetermined health management solution list based on the matching result; acquiring multi-source real-time data from the real-time user; extracting the multi-source real-time data to identify the real-time user's real-time characteristic data; and generating a personalized health management solution by analyzing it in comparison with the predetermined user's key characteristics.

2. Based on a specified user characteristic dataset, clustering analysis is performed on the specified users. The specific clustering analysis process is as follows: Based on a given user feature dataset, feature data for a given user is obtained, Euclidean distance is used as the initial distance metric, and distance metrics are calculated between pairs of points for all given users, with the distance metric being as follows: Here, F ａ，ｋ and F ｂ，ｋ are the normalized values of a predetermined users a and b in the k-th feature, respectively n is the total number of features. Hierarchical clustering is applied as the initial clustering, each designated user is treated as a separate cluster, clusters whose distance metric d(a,b) is greater than or equal to threshold d are merged in stages, a cluster count K0 is selected for one initial clustering based on the results of hierarchical clustering, K0 clusters are selected from the final cluster of hierarchical clustering as the initial center point of K-means, the K-means clustering algorithm is executed, and after each iteration, the cluster quality is evaluated and the K value is dynamically adjusted as follows. The specific analysis process is as follows: The cluster sum of squares (WCSS) is calculated, and the specific formula for calculating WCSS is as follows: Here, M is the number of clusters. Cm is the m-th cluster f(p, k) is the value of the p-th sample in the k-th feature. μm,k is the mean value of the m-th cluster in the k-th feature. A health management method based on multi-source data collection and analysis according to claim 1, characterized in that, in c consecutive iterations, if the decrease in WCSS is less than a threshold t, the iteration is stopped; otherwise, the iteration is continued to adjust the K value, and based on the finally determined K, the K-means clustering algorithm is re-executed to divide a given user into K categories.

3. The aforementioned step S400 includes steps S401 to S402, In step S401, the real-time user inputs initial data, and the same analysis process as for multi-source historical data is performed on the initial data to construct a real-time user profile U1 based on the initial data. The similarity between the real-time user profile U1 and a predetermined user profile U is calculated, and based on the similarity calculation results, it is matched with the corresponding predetermined user category. The one with the highest similarity is selected as the matching result, and based on the matching result, a predetermined list of health management solutions corresponding to the predetermined user category is output. In step S402, a real-time user selects one predetermined health management solution based on a predetermined health management solution list, implements the predetermined health management solution, acquires multi-source real-time data of the real-time user for each selected period, extracts corresponding real-time feature data based on the multi-source real-time data, compares the real-time feature data with key features corresponding to the predetermined health management solution list, and if the real-time feature data is the same as the key features of other predetermined health management solutions in the predetermined health management solution list other than the predetermined health management solution currently being implemented by the real-time user, a personalized health management solution is generated by appropriately adjusting the predetermined health management solution currently being implemented according to the other predetermined health management solutions based on the correspondence between the other predetermined health management solutions and the key features of the predetermined user. This is the health management method based on multi-source data collection and analysis according to claim 1.

4. A health management system based on multi-source data collection and analysis, applicable to a health management method based on multi-source data collection and analysis according to any one of claims 1 to 3, wherein the system includes a data collection and processing module, a user classification and user profile construction module, a knowledge graph construction module, a health management solution matching module, and a personalized health management module. The aforementioned data collection and processing module acquires multi-source historical data of a specified user from multiple databases, uniformly converts multi-source historical data in different formats and structures, extracts characteristic data of a specified user, and standardizes it to construct a characteristic dataset of a specified user. The user classification and user profile construction module performs correlation analysis on feature data within a predetermined user feature dataset, calculates linear relationships between features, establishes relationships between features using a linear regression model, divides users into different categories using clustering analysis, and creates a corresponding predetermined user profile for each category. The knowledge graph construction module defines nodes and edges based on a predetermined user feature dataset and a predetermined user profile, establishes relationships between entities, constructs a knowledge graph for a predetermined user, and constructs a key feature set for each predetermined user by traversing the knowledge graph and identifying key nodes and key features. The aforementioned health management solution matching module constructs a real-time user profile using initial data entered by the real-time user, and by matching it with a predetermined user profile, obtains a category corresponding to the real-time user, and based on the matching result, outputs a predetermined list of health management solutions corresponding to the predetermined user category. The personalized health management module is characterized by acquiring multi-source real-time data of a real-time user based on a predetermined health management solution selected by the real-time user, extracting corresponding real-time feature data based on the multi-source real-time data, comparing the real-time feature data with key features corresponding to a predetermined list of health management solutions, and, if the real-time feature data is the same as the key features of other predetermined health management solutions in the predetermined list of health management solutions other than the predetermined health management solution currently being implemented by the real-time user, appropriately adjusting the predetermined health management solution currently being implemented according to the other predetermined health management solutions based on the correspondence between the other predetermined health management solutions and the predetermined user's key features, thereby generating a personalized health management solution.

5. The aforementioned data acquisition and processing module includes a data acquisition unit and a data processing unit. The data collection unit collects multi-source historical data of a predetermined user from a database, and the data processing unit unifies the multi-source historical data in different formats and structures, extracts characteristic data of the predetermined user, and standardizes it to construct a characteristic dataset of the predetermined user. The aforementioned user classification and user profile construction module includes a user classification unit and a user profile creation unit. The health management system based on multi-source data collection and analysis according to claim 4, characterized in that the user classification unit performs clustering analysis on predetermined user features in a predetermined user feature dataset to divide predetermined users into multiple categories, and the user profile creation unit creates a predetermined user profile by creating weighted feature vectors for predetermined users in each category based on the predetermined user feature dataset.

6. The aforementioned knowledge graph construction module includes a knowledge graph construction unit and a key feature identification unit, The health management system based on multi-source data collection and analysis according to claim 4, characterized in that the knowledge graph construction unit constructs a knowledge graph for a predetermined user in each category based on a predetermined user characteristic dataset and a predetermined user profile, and the key feature identification unit performs relationship analysis in the knowledge graph to identify key features of a predetermined user in each category and constructs a key feature set for each predetermined user.

7. The personalized health management module includes a real-time data collection unit, a similarity calculation unit, and a solution adjustment unit. A health management system based on multi-source data collection and analysis according to claim 4, characterized in that the real-time data collection unit acquires data entered by the user in real time, performs corresponding analysis to construct a real-time user profile, the similarity calculation unit calculates the similarity between the real-time user profile and a predetermined user profile, and determines a predetermined user category corresponding to the real-time user, and the solution adjustment unit compares real-time feature data with key features corresponding to a predetermined list of health management solutions, and if the real-time feature data is the same as the key features of other predetermined health management solutions in the predetermined list of health management solutions other than the predetermined health management solution currently being implemented by the real-time user, the system generates a personalized health management solution by appropriately adjusting the predetermined health management solution currently being implemented according to the other predetermined health management solutions based on the correspondence between the other predetermined health management solutions and the key features of the predetermined user.

Images