Intelligent matching platform for community health resources and home-based care demands

Abstract

The invention discloses an intelligent matching platform for community health resources and home-based care demands, and relates to the technical field of intelligent matching of care services and health resources. The data acquisition module acquires multiple items of data related to community health resources and home-based care demands in real time, performs data preprocessing after acquisition, and transmits the data to the resource and demand analysis and evaluation module; according to the method, comprehensive and real-time multi-source data perception is realized by deploying multiple types of sensors and a multi-source data acquisition interface, accurate quantification and characterization of home-based care requirements and community health resources can be carried out, and a K-Means clustering algorithm and an Apriori association rule mining algorithm are adopted to fully analyze the relationship between the requirements and the resources, so that the accuracy and the reliability of the home-based care requirements and the community health resources are improved. A multi-dimensional resource evaluation index system is established, and an analytic hierarchy process (AHP) is adopted to calculate a resource evaluation index.

Description

Technical Field

[0001] This invention relates to the field of intelligent matching technology for elderly care services and health resources, specifically to an intelligent matching platform for community health resources and home-based elderly care needs. Background Technology

[0002] With the increasing aging of the population and the changes in social development structure and family patterns, "home-based elderly care with community support" has become the mainstream elderly care model. This allows seniors to live in familiar community environments, with personalized elderly care services and support provided jointly by families and communities. However, the health resources provided by communities, such as the number of medical staff, professional rehabilitation therapists, important medical equipment, geographical distribution of medical institutions, and health lectures, are insufficient to meet the growing diversified and personalized needs of home-based elderly care. This leads to an uneven distribution of community health resources, resulting in a supply-demand imbalance, and the scarcity of some resources causes matching imbalances and inefficiencies.

[0003] For example, community health resources are unevenly distributed in space and time. Given the differences in economic development in different regions, there are significant resource gaps in old, remote or newly built communities, with a serious shortage of professional medical staff and equipment. Even in urban communities with relatively abundant resources, there is a structural contradiction between the "tidal" demand that is concentrated in the evening or on holidays and the idle resources during weekdays, which affects the fairness of matching. Most communities still use traditional methods such as manual registration, telephone communication, or simple form statistics to collect user needs and allocate resources. This results in a long cycle from demand reporting to resource response, which cannot meet the high timeliness of emergency health assistance scenarios. At the same time, manual matching relies heavily on the experience and subjective judgment of management personnel, lacks unified and objective quantitative standards, and is easily affected by factors such as incomplete information and personal preferences. It is difficult to achieve optimal resource allocation on a global scale, which not only fails to ensure the fairness and transparency of the matching process and cannot adapt to dynamically changing needs, but also leads to low matching efficiency. Although some communities have introduced information management systems, these systems are often limited in function, only used to record basic user information or service ledgers, resulting in "data silos." These systems generally lack the ability to integrate and process multi-source heterogeneous data, and cannot accurately assess the urgency, complexity, and potential risks of user needs. For example, these systems cannot effectively integrate, deeply correlate, and mine real-time physiological data from IoT health monitoring devices such as smart bracelets and blood pressure monitors, static resource data from community public resource management systems, and historical service records and user feedback data. Logically, resource matching methods typically employ a first-come, first-served approach or simple distance calculation for priority. However, this simplistic matching method struggles to balance fairness and efficiency. For instance, the "first-come, first-served" rule may prevent elderly people with mobility issues or limited access to information from accessing resources for extended periods, exacerbating unfairness. The "nearest distance" rule may ignore the professional suitability of resources and current load conditions, leading to inefficient resource utilization. Therefore, current technologies lack intelligent decision-making capabilities that can dynamically optimize multiple objectives under constraints such as geographical location, service capacity, urgency, and historical fairness. Consequently, they cannot improve matching efficiency while systematically ensuring the fairness and rationality of resource allocation.

[0004] Existing technologies have the following shortcomings: Currently, community health resources applied to the needs of home-based elderly care typically consider only the geographical distribution of medical institutions during the intelligent service matching process. They use GIS technology to integrate the geographical location information of community health resources with the residential location information of elderly people receiving home-based care, or they use big data analysis algorithms to mine potential relationships and patterns in the data on community health resource data and home-based elderly care needs. However, existing technologies do not take into account the regional scarcity and structural imbalance of community health resources, the lack of deep data integration and real-time processing capabilities, and the insufficient intelligent optimization of matching mechanisms. These problems lead to unscientific resource matching, making it difficult to ensure that these scarce resources can be fairly allocated to elderly people in need at home, and failing to balance and maximize the efficiency of resource utilization.

[0005] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0006] The purpose of this invention is to provide an intelligent matching platform for community health resources and home-based elderly care needs. This invention addresses the problems mentioned in the background by setting up a multi-source data acquisition and processing module, deploying multiple types of sensors and configuring multi-source data acquisition interfaces, employing clustering algorithms, association rule mining algorithms, and hierarchical analysis methods in the resource and demand analysis and evaluation module, establishing a multi-objective optimization model and solving it using genetic algorithms in the multi-objective optimization and intelligent matching module, and establishing a multi-channel feedback mechanism, text mining, and regression analysis techniques in the execution feedback improvement module.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an intelligent matching platform for community health resources and home-based elderly care needs, including a multi-source data acquisition and processing module: setting up multiple types of sensors and configuring multi-source data acquisition interfaces to collect multiple data related to community health resources and home-based elderly care needs in real time, performing data preprocessing after collection, and transmitting the data to the resource and demand analysis and evaluation module. Resource and Demand Analysis and Evaluation Module: Receives preprocessed data and performs demand category analysis and demand association analysis on the preprocessed elderly health data using clustering algorithm and association rule mining algorithm, respectively. Different demand categories are identified and demand association analysis results are generated. A multi-dimensional resource evaluation index system is established. The analytic hierarchy process is used to comprehensively evaluate the preprocessed resource status data and generate a resource evaluation index, which is then transmitted to the multi-objective optimization and intelligent matching module. Multi-objective optimization and intelligent matching module: Based on the results of demand correlation analysis and resource assessment index, an intelligent matching model is established using a multi-objective optimization algorithm. Based on the fairness and efficiency indicators of resource matching, the optimal matching solution is found. At the same time, a dynamic adjustment mechanism is set up to optimize and adjust the matching solution in real time, and the final matching result is generated and pushed to the execution feedback improvement module. Execution Feedback Improvement Module: Executes the final matching results, establishes a multi-channel feedback mechanism, collects user feedback information, uses text mining technology to quantify the actual resource demand data of the feedback information, uses regression analysis technology to adjust the parameters of the intelligent matching model, improves the matching algorithm and rules, and updates the intelligent matching model.

[0008] Optionally, the real-time acquisition steps for the multiple data points are as follows: The IoT device interface is connected to the physiological signs sensor and the home environment sensor through wired or wireless communication protocols, and the health data of the elderly collected in real time by multiple sensors is read according to the set time period. At the same time, the IoT device interface interacts with the service record database, and queries based on the elderly person's ID number as an identity identifier to obtain the corresponding elderly person's historical service data (HS) and historical feedback information. The community resource monitoring interface connects with the management system of the community health service center via HTTP or TCP / IP data interface protocol to obtain resource status data in real time at set time intervals; The health data and resource status data of the elderly are acquired synchronously and in parallel, with the set time period and time interval being consistent, both being ten minutes. Then, the timestamps of the health data and resource status data of the elderly are synchronized and calibrated using a Network Time Protocol (NTP) server to obtain initial multiple data with consistent time sequence and reliability.

[0009] Optionally, the clustering algorithm analyzes the demand categories as follows: Preprocessed elderly health data is selected from the preprocessed data to construct a demand feature vector; The elbow rule is used to determine the optimal number of clusters. K demand feature vectors are randomly selected from the preprocessed elderly health data as the initial cluster centers and labeled as K centroids. Calculate the Euclidean distance from each elderly person's demand feature vector to the K centroids, and assign each elderly person's demand feature vector to the cluster containing the nearest centroid; For each cluster, recalculate the centroid of each cluster. The new centroid is the mean of the eigenvectors required by all members of the cluster. Repeatedly execute the steps of assigning each elderly person's demand feature vector to the centroid and updating the centroid until the change in the centroid's position is less than the set threshold, or the maximum number of iterations is reached; The K-Means clustering algorithm assigns a cluster label to each elderly person and assigns a demand category to each cluster by analyzing the centroid demand vector characteristics of each cluster.

[0010] Optionally, the steps of the association rule mining algorithm to analyze demand associations are as follows: From the preprocessed elderly health data and the K-Means clustering algorithm, different demand categories are divided, and all individual items that meet the minimum support are extracted to form a 1-itemset, which includes abnormal health indicators and service needs of the elderly. Using the Apriori algorithm, candidate 2-itemsets are generated by combining 1-itemsets through join and pruning operations, and the support of the candidate 2-itemsets is calculated. A minimum support threshold is set for comparative analysis. Candidate itemsets whose support is greater than or equal to the set minimum support threshold are considered frequent itemsets, thus filtering out frequent 2-itemsets. This process is repeated recursively to generate candidate (M+1)-itemsets from frequent M-itemsets and to filter out frequent (M+1)-itemsets until no more frequent itemsets can be generated. For each frequent itemset m, generate all possible non-empty subset rules m' that conform to the rule m'->(m-m'), where frequent itemset m represents the set of needs in the health and elderly care scenario, m' represents a subset in frequent itemset m, -> represents the co-occurrence relation symbol that implies the difference set (m-m') is also likely to appear when the subset m' appears, and (m-m') represents the difference operation between the set of needs in frequent itemset m and the subset m', which represents the remaining elements in the set of needs m but not in the subset m'. Calculate the confidence level of each rule, set a minimum confidence threshold, and select rules with a confidence level greater than or equal to the preset minimum confidence threshold as valid association rules. Retain those rules that simultaneously meet the minimum support and minimum confidence thresholds to generate the demand association analysis results.

[0011] Optionally, the steps for establishing the multi-dimensional resource assessment index system are as follows: Identify the types of community health resources to be assessed based on the preprocessed resource status data; An assessment of the overall quality of community health resources is conducted, including an assessment of resource availability, quality, and prioritization. Based on the assessment objectives, quantifiable assessment indicators are identified and decomposed to form a hierarchical structure of assessment objectives. These include availability indicators, which are divided into time matching degree and geographical accessibility indicators; quality indicators, which are divided into professional qualification level, medical equipment integrity rate and historical service user rating indicators; and priority indicators, which are divided into emergency rescue capability and scarcity coefficient indicators. The weight of each evaluation indicator is determined by using the Analytic Hierarchy Process (AHP). Experts are then invited to compare the indicators in the same level pairwise and score them based on their relative importance using the 1-9 scale. A judgment matrix is ​​then constructed for each upper-level indicator.

[0012] Optionally, the steps for generating the resource assessment index are as follows: Based on the AHP judgment matrix, the maximum eigenvalue of the judgment matrix and the eigenvectors of each corresponding evaluation index are solved. The feature vectors are normalized to obtain the weight vectors of each indicator. The consistency ratio is calculated, and a ratio threshold is set for comparative analysis. When the calculated consistency ratio is less than the set ratio threshold, the judgment matrix is ​​considered to have satisfactory consistency, and the weight vectors of each indicator are considered to be effective. Based on the determined weight vectors of each indicator, the preprocessed resource status data is processed by min-max normalization. After the evaluation indicators are quantified, standard values ​​are obtained and unified to the [0,1] interval. Then, the linear weighted summation method is used to generate the resource evaluation index of the comprehensive quality of health resources in each community.

[0013] Optionally, the steps for establishing the intelligent matching model are as follows: Based on the results of demand correlation analysis and resource assessment index, binary code is used to define decision variables for matching home-based elderly care needs with community health resources; A multi-objective function for fairness and efficiency is constructed, including the introduction of the fairness index of Gini coefficient and the efficiency index of resource utilization. Different demand categories and resource assessment indices are divided according to the K-Means clustering algorithm. The Gini coefficient is calculated by the sum of the per capita resource assessment indices allocated to each demand category. Minimizing the Gini coefficient to measure the maximization of fairness in resource allocation is taken as the fairness objective. Resource utilization rate is obtained by calculating the ratio of the sum of all allocated resource evaluation indices to the sum of total available resource evaluation indices, and maximizing resource utilization rate is taken as the efficiency objective. Set constraints, including demand uniqueness constraints, resource capacity constraints, geographical adaptability constraints, and non-negativity constraints for decision variables; Based on the set constraints, the functions are integrated into a multi-objective function to form a multi-objective optimization model.

[0014] Optionally, the steps for finding the optimal matching scheme are as follows: Based on the multi-objective optimization model, a set of initial matching schemes is randomly generated using a genetic algorithm, which serves as chromosomes. Each chromosome consists of decision variables and satisfies the constraints. Let the population size be L. For each chromosome, the fitness value of the matching scheme is calculated using a weighted summation method in conjunction with a multi-objective optimization model. Based on the fitness value of the matching scheme, Y chromosomes are selected from the current population as parents using the tournament selection method; A single-point crossover operation is performed on the selected parent chromosomes to generate offspring chromosomes; To increase population diversity, chromosome mutations are performed on offspring. Add offspring chromosomes to the population, set a fitness threshold, recalculate the fitness of offspring chromosomes and compare it with the set fitness threshold, and eliminate some poor chromosomes based on the fitness value to update and retain the new population. Repeat the above steps of population selection, crossover, mutation, and pruning until the maximum number of iterations is reached or the fitness value no longer improves. Then, output the Pareto optimal solution set and find an optimal matching scheme.

[0015] Optionally, the quantification steps for the actual resource demand data are as follows: We collect user feedback through multiple channels, including online questionnaires, telephone follow-ups, and mobile application feedback, including raw text data on user satisfaction, resource utilization, and experience opinions. The collected raw text data is preprocessed to remove special characters / punctuation marks, segment words, and filter stop words, resulting in preprocessed text data. The TF-IDF algorithm, a text mining technique, is used to extract keywords related to resource demand from preprocessed text data and integrate them into a keyword set. The pre-trained sentiment analysis BERT model is used to perform sentiment analysis on each feedback message containing keywords to determine the user's sentiment tendency towards resource demand, including positive, negative, and neutral. Based on the keyword and sentiment analysis results, and by counting the number of times each keyword appears in different sentiment tendencies, the actual resource demand data is quantified.

[0016] Optionally, the parameter adjustment steps of the intelligent matching model are as follows: The fairness and efficiency indicators of the intelligent matching model, including the Gini coefficient and resource utilization rate, recorded after the execution of historical matching schemes, are used as independent variables; The actual resource demand data collected and quantified after the execution of the matching scheme period is used as the dependent variable. Linear regression analysis was used to establish a multiple linear regression model. The least squares method was used to fit the historical dataset to obtain the regression coefficients and error terms. Based on the magnitude and sign of the regression coefficients, analyze the degree and direction of the influence of the independent variables on the dependent variable, and adjust the weight parameters of the intelligent matching model based on the results of the influencing factor analysis. The adjusted weights are then updated into the multi-objective function of the genetic algorithm to calculate the optimal solution for matching demand and resources in the next round.

[0017] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention achieves comprehensive and real-time multi-source data perception by deploying multiple types of sensors and multi-source data acquisition interfaces. It can accurately quantify and characterize the needs of home-based elderly care and community health resources. It also uses K-Means clustering algorithm and Apriori association rule mining algorithm to fully analyze the relationship between needs and resources. By establishing a multi-dimensional resource evaluation index system and using the analytic hierarchy process (AHP) to calculate the resource evaluation index, it not only achieves refined insight and forward-looking prediction of needs, but also realizes the standardization and hierarchical management of resources. By introducing the Gini coefficient and resource utilization rate as objectives to measure the fairness and efficiency of demand and resource matching, and by using a genetic algorithm to find the Pareto optimal solution set, the optimal matching decision is generated, significantly improving the accuracy of resource matching demand, user satisfaction, and resource utilization. By collecting user feedback through multiple channels and using text mining technology to quantify actual resource demand data, the intelligent matching platform can more accurately understand the specific needs of elderly people living at home. By combining regression analysis technology to adjust the parameters of the intelligent matching model and improve the algorithm, community health resources can be more accurately matched with the needs of the elderly. At the same time, accurate matching avoids waste and idleness of resources, enabling community health resources to be used more fully, promoting the long-term effective connection between community health resources and the needs of home-based elderly care, and promoting the sustainable development of community elderly care services. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a block diagram of the intelligent matching platform for community health resources and home-based elderly care needs of the present invention.

[0020] Figure 2 This is a flowchart of the intelligent matching process of the intelligent matching platform of the present invention. Detailed Implementation

[0021] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, they are provided so that the description of this disclosure will be more complete and fully convey the concept of the exemplary embodiments to those skilled in the art.

[0022] Example This invention provides, for example Figure 1-2 The intelligent matching platform for community health resources and home-based elderly care needs shown includes a multi-source data acquisition and processing module: setting up multiple types of sensors and configuring multi-source data acquisition interfaces to collect multiple data related to community health resources and home-based elderly care needs in real time, including elderly health data and resource status data. After collection, the multiple data are preprocessed by data cleaning, normalization and spatiotemporal alignment to obtain preprocessed data and transmit it to the resource and demand analysis and evaluation module. Specifically, the system includes multiple types of sensors, including physiological sensors that monitor the health of the elderly in real time and home environment sensors that are installed in the elderly’s residences. The physiological sensors collect physiological data such as heart rate, blood pressure and blood sugar in real time, while the home environment sensors collect environmental data such as temperature, humidity and air quality in the elderly’s homes in real time. The configuration includes a multi-source data acquisition interface, a community resource monitoring interface, an IoT device interface, and a service record database. The IoT device interface is used to connect to various sensors to collect elderly health data in real time, including the elderly’s physiological data and environmental data. Combined with the service record database, the corresponding elderly’s historical service data and historical feedback information are obtained. The community resource monitoring interface connects to the management system of the community health service center to obtain real-time information on the availability of medical staff, the status of medical equipment, and the geographical distribution of medical institutions. The service record database is used to store historical service data, intelligent matching schemes, and user feedback information. During multiple data collection processes, the IoT device interface and community resource monitoring interface will read relevant data from the service record database as needed. At the same time, new matching schemes generated during the intelligent matching process and user feedback information after execution will also be stored in the service record database in real time.

[0023] Specifically, the real-time data collection steps for multiple data points are as follows: The IoT device interface is connected to the physiological signs sensor and the home environment sensor via wired or wireless communication protocols respectively. The system reads the health data of the elderly collected in real time by multiple sensors according to the set time period. The data includes physiological data and environmental data. The physiological data includes the elderly’s heart rate (HR), blood pressure (BP) (including systolic blood pressure (SBP) and diastolic blood pressure (DBP)), and blood glucose (BG). The environmental data includes the elderly’s home temperature (T), humidity (H), and air quality index (AQI). At the same time, the IoT device interface interacts with the service record database, and queries based on the elderly person's ID number as an identity identifier to obtain the corresponding elderly person's historical service data (HS) and historical feedback information. The community resource monitoring interface connects with the management system of the community health service center via HTTP or TCP / IP data interface protocols, and acquires resource status data in real time at set time intervals, including the presence of medical staff (MS), the status of medical equipment (ED), and the geographical distribution of medical institutions (LD). Among them, the presence of medical staff (MS) includes the number of medical staff and their professional positions; the status of medical equipment (ED) includes the availability, maintenance, and occupancy status of the equipment; and the geographical distribution of medical institutions (LD) is represented by latitude and longitude coordinates. The health data and resource status data of the elderly are acquired synchronously and in parallel, with the set time period and time interval being consistent, both being ten minutes. Then, the timestamps of the health data and resource status data of the elderly are synchronized and calibrated using a Network Time Protocol (NTP) server to obtain initial multiple data with consistent time sequence and reliability.

[0024] Specifically, the preprocessing steps for multiple data points are as follows: The initial multivariate data collected were cleaned, including missing value handling, outlier detection and correction. Specifically, for missing value handling, the mean imputation method was used for continuous elderly health data in the initial multivariate data, and the mode imputation method was used for categorized resource status data in the initial multivariate data. For outlier detection, statistical Z-score normalization or the isolated forest algorithm is used to identify outliers; for outlier correction, median filtering or sliding window mean is used for smoothing outliers that are confirmed to be noise. After cleaning, the multiple data points are normalized and mapped to a unified numerical range [0,1]. For different types of data, the normalization process is performed using Min-Max normalization or Z-score standardization methods. The timestamps of the normalized data are sorted in chronological order and aligned to the same time granularity. For spatial location information, nearest neighbor interpolation or kriging interpolation is used for spatial alignment. The preprocessed data is packaged into a preprocessed data packet and transmitted to the resource and demand analysis and evaluation module for subsequent modules to perform in-depth analysis and matching calculations.

[0025] Resource and Demand Analysis and Evaluation Module: Based on the preprocessed data, the K-Means clustering algorithm is used to analyze the demand categories of the preprocessed elderly health data, classifying different demand categories. The Apriori association rule mining algorithm is used to perform demand association analysis on the preprocessed elderly health data, generating demand association analysis results. A multi-dimensional resource evaluation index system is established. The Analytic Hierarchy Process (AHP) is used to comprehensively evaluate the availability, quality, and priority of the preprocessed resource status data, generating a resource evaluation index, which is then transmitted to the multi-objective optimization and intelligent matching module. Specifically, the steps for clustering algorithms to analyze demand categories are as follows: Preprocessed elderly health data were extracted from the preprocessed data. A demand feature vector was constructed based on physiological and environmental data, where the expression for the demand feature vector is: ,and In the formula, Represented as building The demand feature vector corresponding to each elderly person Represented as the corresponding number Characteristic data of elderly people Represented as the corresponding number Characteristic data of elderly people Represented as the corresponding number Characteristic data of elderly people Indicated as in Heart rate values ​​of elderly people collected at all times. Indicated as in Blood pressure values ​​of elderly people collected at all times. Indicated as in Blood glucose levels of elderly people collected at regular intervals Indicated as in Temperature values ​​collected continuously at home by elderly people Indicated as in Humidity levels collected continuously at the homes of elderly people. Indicated as in Air quality values ​​collected continuously at the homes of elderly people; The elbow rule is used to determine the optimal number of clusters. K demand feature vectors are randomly selected from the preprocessed elderly health data as initial cluster centers, labeled as K centroids. The expression for the sum of squared errors within clusters using the elbow rule is: ,and In the formula, This is expressed as the sum of squared errors for all clusters. Represented as the number of clusters, Represented as the first The required feature vector of a sample point in a cluster, Represented as the first Clusters, Represented as the first The center of mass, It is expressed as the square of the Euclidean distance; Calculate the Euclidean distance from each elderly person's demand feature vector to the K centroids, and assign each elderly person's demand feature vector to the cluster containing the nearest centroid. The formula for calculating the Euclidean distance is as follows: ,and In the formula, This is expressed as the Euclidean distance between the required feature vector of a sample point and its centroid. This is represented as the dimension of the demand feature vector. Represented as the first feature vector within the requirement of the sample point 1 eigenvalue, Represented as the first within the centroid One eigenvalue; For each cluster, recalculate the centroid of each cluster. The new centroid is the mean of the required eigenvectors of all members of the cluster. The expression for updating the centroid is: In the formula, Represented as the new cluster center The Feature values ​​in each dimension Represented as the first The set of data points in each cluster that require feature vectors This represents the number of elements in the set of data points. Represented as the first Demand feature vectors in each cluster Represented as the first The demand feature vector in the cluster is the first Feature values ​​in each dimension; Repeatedly execute the steps of assigning each elderly person's demand feature vector to the centroid and updating the centroid until the change in the centroid's position is less than the set threshold, or the maximum number of iterations is reached; The K-Means clustering algorithm assigns a cluster label to each elderly person. By analyzing the centroid demand vector characteristics of each cluster, it assigns a demand category to each cluster, including cluster A as high-risk monitoring type: characterized by high blood pressure, high blood sugar, and low activity level. Cluster B is the rehabilitation and conditioning type: characterized by the need for regular physical therapy during the postoperative or chronic disease recovery period; Cluster C represents the health maintenance type: characterized by relatively normal health indicators, requiring preventive health guidance.

[0026] Specifically, the steps involved in analyzing requirement associations using association rule mining algorithms are as follows: From the preprocessed elderly health data and the K-Means clustering algorithm, different demand categories are divided and all individual items that meet the minimum support are extracted to form a 1-itemset, which includes abnormal health indicators and service needs of the elderly. Among them, abnormal health indicators include arrhythmia, hypertension, hyperglycemia, sudden changes in temperature and humidity, and low air quality. Service needs include heart rate monitoring, hypertension monitoring, diabetes medication guidance, fall risk intervention, indoor ventilation reminders, dynamic temperature and humidity control during seasonal changes, stroke hemiplegic limb care / emotional support, and medication guidance for various chronic diseases. Using the Apriori algorithm, candidate 2-itemsets are generated by combining 1-itemsets through join and pruning operations. The support of the candidate 2-itemsets is then calculated, where the formula for calculating the support is: In the formula, This represents the support for generating candidate 2-itemsets by combining 1-itemsets. This represents the probability that a 1-itemset and a candidate 2-itemset both occur simultaneously. A minimum support threshold is set for comparative analysis. Candidate itemsets whose support is greater than or equal to the set minimum support threshold are considered frequent itemsets, thus filtering out frequent 2-itemsets. The expression for a frequent 2-itemset is: In the formula, This indicates setting a minimum support threshold; This process is repeated recursively to generate candidate (M+1)-itemsets from frequent M-itemsets and to filter out frequent (M+1)-itemsets until no more frequent itemsets can be generated. For each frequent itemset m, generate all possible non-empty subset rules m' that conform to the rule m'->(m-m'). Here, frequent itemset m represents the set of needs in the health and elderly care scenario, m' represents a subset of frequent itemset m, -> represents the co-occurrence relation symbol implying that the difference set (m-m') is also likely to appear simultaneously when the subset m' appears, and (m-m') represents the difference operation between the set of needs in frequent itemset m and the subset m', representing the remaining elements in the set of needs m but not in the subset m'. For example, using the Apriori algorithm, find a frequent itemset m that includes hypertension medication guidance, exercise control, mood regulation, and dietary counseling. From this frequent itemset m, multiple candidate association rules can be generated. For instance, if the subset m' includes hypertension medication guidance, then (m-m') includes exercise control, mood regulation, and dietary counseling; or if the subset m' includes hypertension medication guidance and exercise control, then (m-m')... If a subset m' includes regulating mood swings and dietary counseling, or if a subset m' includes guidance on hypertension medication, controlling exercise levels, and regulating mood swings, then (m-m') includes dietary counseling. Based on the formation of the m'->(m-m') rule, the confidence level of each rule is calculated. Calculate the confidence score for each rule, where the formula for calculating the confidence score is: In the formula, Represented as frequent 2-itemsets Confidence level of the rule This is represented as the conditional probability that a rule containing 1-itemsets also contains frequent 2-itemsets; A minimum confidence threshold is set, and rules with a confidence level greater than or equal to this threshold are considered valid association rules. The expression for a valid association rule is: In the formula, This indicates setting a minimum confidence threshold; Retain those rules that simultaneously meet the minimum support and minimum confidence thresholds to generate demand correlation analysis results.

[0027] Specifically, the steps for establishing a multi-dimensional resource assessment indicator system are as follows: Based on the preprocessed resource status data, the types of community health resources to be assessed are identified, including healthcare worker availability (MS), medical equipment status (ED), and geographical distribution of healthcare facilities (LD). The overall quality of community health resources is assessed, including the availability, quality, and priority of resources. Availability is the assessment of how easily resources can be accessed in terms of time and space. Quality is the technical level and reliability of resources in providing health services to older users. Prioritization is the urgency and strategic importance that should be considered when allocating resources; Based on the assessment objectives, quantifiable assessment indicators are identified and decomposed, forming a hierarchical structure of assessment objectives. Availability is divided into time matching degree and geographical accessibility indicators. Among them, time matching degree is the degree of overlap between the available service time of the assessment resources and the peak demand period. Geographic accessibility is the inverse of the average distance between a resource's location and the people it serves; Quality is divided into professional qualification level, medical equipment integrity rate and historical service user rating indicators. Among them, professional qualification level refers to the professional title and years of practice of medical staff. Medical equipment availability rate is a percentage of medical equipment that is in normal working order. Historical service user ratings are the average satisfaction levels of users who have received past services. Priority is divided into emergency rescue capability and scarcity coefficient indicators. Emergency rescue capability is used to determine whether an emergency rescue mission can be handled. The scarcity coefficient is used to determine the degree of scarcity of similar resources within a community; The Analytic Hierarchy Process (AHP) is used to determine the weight of each evaluation indicator. Experts are then invited to conduct pairwise comparisons of indicators within the same level, scoring them using a 1-9 scale to assess their relative importance. A judgment matrix is ​​constructed for each higher-level indicator, where the expression for the judgment matrix is: ,and In the formula, Represented as an AHP judgment matrix, This is represented as the first value assigned using the 1-9 scale. The evaluation indicator is relative to the first The importance of each evaluation indicator This is expressed as the order of the judgment matrix. , These represent the number of rows and columns of the judgment matrix, respectively. This represents the total number of quantitative evaluation indicators.

[0028] Specifically, the steps for generating the resource assessment index are as follows: Based on the AHP judgment matrix, the largest eigenvalue of the judgment matrix and the corresponding eigenvectors of each evaluation index are obtained. The formula for calculating the eigenvectors is as follows: ,and In the formula, Represented as The feature vector of each evaluation indicator Represented as the largest eigenvalue, This represents the weight of the first evaluation indicator. Represented as the first The weight of each evaluation indicator, Represented as the vector transpose symbol; The feature vectors are normalized to obtain the weight vectors of each indicator. The consistency ratio is calculated, and a threshold is set for comparative analysis. When the calculated consistency ratio is less than the set threshold, the judgment matrix is ​​considered to have satisfactory consistency, and the weight vectors of each indicator are considered valid. Otherwise, the judgment matrix needs to be adjusted. The formula for calculating the consistency ratio is as follows: ,and In the formula, This is represented as a consistency indicator. Expressed as the consistency ratio, This is represented as a random consistency index that can be obtained by looking up a table. This is expressed as the order of the judgment matrix. This represents setting a ratio threshold, and ; Based on the determined weight vectors of each indicator, the preprocessed resource status data is subjected to min-max normalization. After quantifying the evaluation indicators, standardized values ​​are obtained within the [0,1] interval. Then, a linear weighted summation method is used to generate a resource evaluation index for the comprehensive quality of health resources in each community. The expression for the resource evaluation index is as follows: ,and In the formula, Represented as the first A resource assessment index for the overall quality of health resources in a community. Represented as the first The weight of each evaluation indicator, Represented as resources in the first place Standard values ​​for each evaluation indicator.

[0029] Multi-objective optimization and intelligent matching module: Based on the results of demand correlation analysis and resource assessment index, an intelligent matching model is established using a multi-objective optimization genetic algorithm. Based on the fairness and efficiency of resource matching, the fairness index of the Gini coefficient and the efficiency index of resource utilization are introduced. The optimal matching solution is continuously searched through iterative searches. At the same time, a dynamic adjustment mechanism is set up to continuously monitor the latest changes in the needs of the elderly and the use of resources, optimize and adjust the matching solution in real time, generate the final matching result, and push it to the execution feedback improvement module. Specifically, the steps for building the intelligent matching model are as follows: Based on the demand correlation analysis results and resource assessment index, binary code is used to define decision variables for matching home-based elderly care needs with community health resources. The expression for defining the decision variables is as follows: ,and ,as well as In the formula, Indicated as the first Community health resources are allocated to the first Decision variables for individual home-based elderly care needs Indicated as the first Community health resources are not allocated to the first Decision variables for individual home-based elderly care needs Represented as a collection of community health resources, The number of resources represented by the community health resource set. Indicated as the first community health resource, Represented as the first Community health resources This represents a collection of needs for home-based elderly care. This represents the number of demands for home-based elderly care. This is indicated as the first home-based elderly care need. Represented as the first Individual home-based elderly care needs These are indices representing the number of resources and the number of demands, respectively. A multi-objective function for fairness and efficiency is constructed, including the introduction of the Gini coefficient as a fairness indicator and resource utilization efficiency as an efficiency indicator. Fairness is reflected in the equitable distribution of resources among different elderly groups. Different demand categories and resource assessment indices are identified using the K-Means clustering algorithm. The Gini coefficient is calculated by summing the per capita resource assessment indices allocated to each demand category. Minimizing the Gini coefficient, which measures the maximization of fairness in resource allocation, is taken as the fairness objective. The formula for calculating the Gini coefficient is as follows: ,and In the formula, The Gini coefficient, representing a measure of fairness in resource allocation, Represented as the first The amount of community health resources available to individuals with home-based elderly care needs Represented as the first The amount of community health resources available to individuals with home-based elderly care needs This is expressed as the average amount of community health resources. This represents the quantity of demand for home-based elderly care, and ; Efficiency is reflected in the full utilization of resources and the avoidance of idleness. Resource utilization rate is calculated by dividing the sum of the evaluation indices of all allocated resources by the sum of the evaluation indices of the total available resources. Maximizing resource utilization rate is taken as the efficiency objective. The formula for calculating resource utilization rate is as follows: In the formula, This is expressed as resource utilization rate, which measures the efficiency of resource allocation. Represented as the first The actual usage of community health resources Represented as the first The availability of health resources in each community; The expression for the multi-objective function is: In the formula, Represented as a multi-objective function, This is expressed as minimizing the Gini coefficient. This is expressed as maximizing resource utilization. The constraints are set, including the uniqueness of demand, the resource capacity, the regional suitability, and the non-negativity of decision variables. Among them, the uniqueness of demand means that for each home-based elderly care demand, at most one community health resource service can be allocated. Resource capacity constraints set an upper limit on the number of home-based elderly care needs that each community health resource can serve within the same time period. The geographical adaptation constraint is to select the nearest and best medical institutions when medical personnel provide home-based elderly care services, taking into account the geographical distribution of medical institutions. The non-negativity constraint of decision variables is that any home-based elderly care needs must be matched with certain community health resources. The expression for the constraint is In the formula, This indicates that constraints are set. This is represented as a uniqueness constraint on demand. This is represented as a resource capacity constraint. This is represented as a region adaptation constraint. This is represented as a non-negativity constraint on the decision variable; Based on the established constraints, these constraints are integrated into a multi-objective function to form a multi-objective optimization model. The expression for this multi-objective optimization model is as follows: In the formula, This can be expressed as an expression for a multi-objective optimization model. This represents the set constraints that a multi-objective function needs to satisfy.

[0030] Specifically, the steps for finding the optimal matching scheme are as follows: Based on a multi-objective optimization model, a set of initial matching schemes is randomly generated using a genetic algorithm, serving as chromosomes. Each chromosome consists of decision variables. The population consists of individuals and satisfies the constraints. Let the population size be L. For each chromosome, using a multi-objective optimization model, the fitness value of the matching scheme is calculated using a weighted summation method. The formula for calculating the fitness value is as follows: ,and In the formula, This is represented as the fitness value of the matching scheme. This represents the weighting coefficients corresponding to the minimum Gini coefficient. This is represented as the weighting coefficient corresponding to maximizing resource utilization; Based on the fitness value of the matching scheme, Y chromosomes are selected from the current population as parents using the tournament selection method; A single-point crossover operation is performed on the selected parent chromosomes to generate offspring chromosomes; To increase population diversity, the chromosomes of offspring are modified. Add offspring chromosomes to the population, set a fitness threshold, recalculate the fitness of offspring chromosomes and compare it with the set fitness threshold, and eliminate some poor chromosomes based on the fitness value to update and retain the new population. Repeat the above steps of population selection, crossover, mutation, and pruning until the maximum number of iterations is reached or the fitness value no longer improves. Then, output the Pareto optimal solution set and find an optimal matching scheme.

[0031] Specifically, the steps for setting up a dynamic adjustment mechanism for real-time optimization are as follows: Continuous data monitoring, real-time collection of the latest data on changes in the needs of the elderly and the use of resources, to obtain the latest health data and resource status data of the elderly; Based on the latest health data of the elderly and the latest resource status data, the current matching scheme is re-evaluated, and the fairness and efficiency indicators of the new data are calculated. A Gini coefficient threshold and a resource utilization rate threshold are set. If the difference between the fairness index of the newly calculated data and the minimum Gini coefficient calculated in the previous time step exceeds the set Gini coefficient threshold, or if the difference between the efficiency index of the newly calculated data and the maximum resource utilization rate calculated in the previous time step exceeds the set resource utilization rate threshold, then the optimal matching scheme found in the previous time step is considered to need adjustment. The expression for the judgment condition requiring adjustment of the matching scheme is as follows: ,or In the formula, This is expressed as the Gini coefficient for calculating new data. This is expressed as the resource utilization rate for calculating new data. This is expressed as the Gini coefficient threshold. This is expressed as a resource utilization threshold. The genetic algorithm is still used, and multi-objective optimization is performed again based on the new data to find a new optimal matching scheme that fits the current state. The latest optimal matching scheme is updated to the final matching result and pushed to the execution feedback improvement module for task distribution.

[0032] The execution feedback improvement module executes the final matching results, establishes a multi-channel feedback mechanism, including online questionnaires, telephone follow-ups, and mobile terminal application feedback, collects feedback information on user satisfaction, resource utilization, and experience opinions, uses text mining technology to extract keywords and perform sentiment analysis on the feedback information to quantify actual resource demand data, uses regression analysis technology to evaluate the relationship between the fairness and efficiency indicators of the intelligent matching model and the quantified actual resource demand data in the feedback information, adjusts the parameters of the intelligent matching model through regression analysis of influencing factors, improves the matching algorithm and rules, and updates the intelligent matching model.

[0033] Specifically, the steps for quantifying actual resource demand data are as follows: We collect user feedback through multiple channels, including online questionnaires, telephone follow-ups, and mobile application feedback, including raw text data on user satisfaction, resource utilization, and experience opinions. The collected raw text data is preprocessed to remove special characters / punctuation marks, segment words, and filter stop words, resulting in preprocessed text data. The TF-IDF algorithm, a text mining technique, is used to extract keywords related to resource demand from preprocessed text data and integrate them into a keyword set. The expression for keyword extraction using the TF-IDF algorithm is as follows: ,and In the formula, This is represented as a set of extracted keywords. The frequency of a keyword is expressed as the number of times it appears in the text. Represented as inverse document frequency, Represented as keywords, Represented as a document, Represented as a collection of documents, This represents the total number of texts in the text collection. This is expressed as the number of texts containing the keyword; The pre-trained sentiment analysis BERT model is used to perform sentiment analysis on each feedback message containing keywords to determine the user's sentiment tendency towards resource demand, including positive, negative, and neutral. Based on keyword and sentiment analysis results, and by counting the frequency of each keyword across different sentiment levels, the actual resource demand data is quantified. The expression for quantifying the actual resource demand data is as follows: ,and In the formula, This is represented as quantified actual resource demand data converted into demand values ​​for resource categories. Represented as resource category, Represented as all categories that contain mappings to resource categories Keywords A collection of feedback documents, Represented as a document The set of keywords extracted from it. Represented as mapping to resource categories All keywords The set, Represented as the intersection symbol, Represented as keywords In the document and document collection The TF-IDF value below, Represented as a document The emotional score, and , Represented as a sentiment weighting function, Represented as a collection of feedback documents The number of Chinese documents.

[0034] Specifically, the parameter adjustment steps for the intelligent matching model are as follows: The fairness and efficiency indicators of the intelligent matching model, including the Gini coefficient and resource utilization rate, recorded after the execution of historical matching schemes, are used as independent variables; The actual resource demand data collected and quantified after the execution of the matching scheme period is used as the dependent variable. Linear regression analysis was employed to establish a multiple linear regression model. The historical dataset was then fitted using the least squares method to obtain the regression coefficients and error terms. The expression for the multiple linear regression model is as follows: In the formula, Represented as a vector of dependent variables based on actual resource demand data. Represented as a vector of independent variables consisting of the Gini coefficient and resource utilization rate. It is represented as a vector of regression coefficients, consisting of the intercept term, the regression coefficient of the Gini coefficient, and the regression coefficient of resource utilization rate. Represented as a random error vector; The expression for estimating regression coefficients using the least squares method is: In the formula, This is represented as an estimate of the regression coefficient vector. Represented as the transpose symbol, This is represented as a matrix inversion operation; Based on the magnitude and sign of the regression coefficients, the degree and direction of the influence of the independent variables on the dependent variable are analyzed. Based on the results of the influencing factor analysis, the weight parameters of the intelligent matching model are adjusted. The adjusted weight expression is as follows: ,and In the formula, This is expressed as the normalized weight of the Gini coefficient in the new multi-objective function. This is represented as the normalized weight of resource utilization in the new multi-objective function. This represents the regression coefficient estimate of the Gini coefficient obtained from the regression model. This is expressed as the regression coefficient estimate of resource utilization obtained from the regression model; The adjusted weights are then updated into the multi-objective function of the genetic algorithm to calculate the optimal solution for matching demand and resources in the next round.

[0035] Specifically, this describes a matching method for an intelligent matching platform that connects community health resources with the needs of home-based elderly care, comprising the following steps: Multiple types of sensors are set up and multi-source data acquisition interfaces are configured to collect multiple data related to community health resources and home-based elderly care needs in real time, including elderly health data and resource status data. After collection, multiple data are preprocessed by data cleaning, normalization and spatiotemporal alignment. Based on the preprocessed data, the K-Means clustering algorithm and the Apriori association rule mining algorithm were used to perform demand category analysis and demand association analysis on the preprocessed elderly health data, respectively. Different demand categories were identified and demand association analysis results were generated. A multi-dimensional resource assessment index system was established. The analytic hierarchy process (AHP) was used to comprehensively evaluate the availability, quality and priority of the preprocessed resource status data and generate a resource assessment index. Based on the results of demand correlation analysis and resource assessment index, a smart matching model is established using a multi-objective optimization genetic algorithm. Based on the fairness and efficiency of resource matching, the fairness index of the Gini coefficient and the efficiency index of resource utilization are introduced. The model continuously iterates and searches for the optimal matching solution. At the same time, a dynamic adjustment mechanism is set up to continuously monitor the latest changes in the needs of the elderly and the use of resources, and optimize and adjust the matching solution in real time to generate the final matching result. The system utilizes a task dispatcher to push the final matching results and a status tracker to monitor task execution progress. A multi-channel feedback mechanism is established, including online surveys, telephone follow-ups, and mobile application feedback, to collect user satisfaction, resource utilization, and experience feedback. Text mining techniques are used to extract keywords and perform sentiment analysis on the feedback information to quantify actual resource demand data. Regression analysis is used to evaluate the relationship between the fairness and efficiency indicators of the intelligent matching model and the quantified actual resource demand data in the feedback information. By analyzing influencing factors through regression analysis, the parameters of the intelligent matching model are adjusted, and the matching algorithm and rules are improved and updated.

[0036] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0037] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0038] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0039] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0040] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An intelligent matching platform for community health resources and home-based elderly care needs, characterized in that, It includes a multi-source data acquisition and processing module: setting up multiple types of sensors and configuring multi-source data acquisition interfaces to collect multiple data related to community health resources and home-based elderly care needs in real time; after collection, the data is preprocessed and transmitted to the resource and demand analysis and evaluation module. Resource and Demand Analysis and Evaluation Module: Receives preprocessed data and performs demand category analysis and demand association analysis on the preprocessed elderly health data using clustering algorithm and association rule mining algorithm, respectively. Different demand categories are identified and demand association analysis results are generated. A multi-dimensional resource evaluation index system is established. The analytic hierarchy process is used to comprehensively evaluate the preprocessed resource status data and generate a resource evaluation index, which is then transmitted to the multi-objective optimization and intelligent matching module. Multi-objective optimization and intelligent matching module: Based on the results of demand correlation analysis and resource assessment index, an intelligent matching model is established using a multi-objective optimization algorithm. Based on the fairness and efficiency indicators of resource matching, the optimal matching solution is found. At the same time, a dynamic adjustment mechanism is set up to optimize and adjust the matching solution in real time, and the final matching result is generated and pushed to the execution feedback improvement module. Execution Feedback Improvement Module: Executes the final matching results, establishes a multi-channel feedback mechanism, collects user feedback information, uses text mining technology to quantify the actual resource demand data of the feedback information, uses regression analysis technology to adjust the parameters of the intelligent matching model, improves the matching algorithm and rules, and updates the intelligent matching model.

2. The intelligent matching platform for community health resources and home-based elderly care needs according to claim 1, characterized in that, The real-time acquisition steps for the aforementioned multiple data points are as follows: The IoT device interface is connected to the physiological signs sensor and the home environment sensor through wired or wireless communication protocols, and the health data of the elderly collected in real time by multiple sensors is read according to the set time period. At the same time, the IoT device interface interacts with the service record database, and queries based on the elderly person's ID number as an identity identifier to obtain the corresponding elderly person's historical service data and historical feedback information. The community resource monitoring interface connects with the management system of the community health service center via HTTP or TCP / IP data interface protocol to obtain resource status data in real time at set time intervals; The health data and resource status data of the elderly are acquired synchronously and in parallel, with the set time period and time interval being consistent, both being ten minutes. Then, the timestamps of the health data and resource status data of the elderly are synchronized and calibrated using a Network Time Protocol (NTP) server to obtain initial multiple data with consistent time sequence and reliability.

3. The intelligent matching platform for community health resources and home-based elderly care needs according to claim 2, characterized in that, The steps of the clustering algorithm to analyze demand categories are as follows: Preprocessed elderly health data is selected from the preprocessed data to construct a demand feature vector; The elbow rule is used to determine the optimal number of clusters. K demand feature vectors are randomly selected from the preprocessed elderly health data as the initial cluster centers and labeled as K centroids. Calculate the Euclidean distance from each elderly person's demand feature vector to the K centroids, and assign each elderly person's demand feature vector to the cluster containing the nearest centroid; For each cluster, recalculate the centroid of each cluster. The new centroid is the mean of the eigenvectors required by all members of the cluster. Repeatedly execute the steps of assigning each elderly person's demand feature vector to the centroid and updating the centroid until the change in the centroid's position is less than the set threshold, or the maximum number of iterations is reached; The K-Means clustering algorithm assigns a cluster label to each elderly person and assigns a demand category to each cluster by analyzing the centroid demand vector characteristics of each cluster.

4. The intelligent matching platform for community health resources and home-based elderly care needs according to claim 3, characterized in that, The steps of the association rule mining algorithm for analyzing demand associations are as follows: From the preprocessed elderly health data and the K-Means clustering algorithm, different demand categories are divided, and all individual items that meet the minimum support are extracted to form a 1-itemset, which includes abnormal health indicators and service needs of the elderly. Using the Apriori algorithm, candidate 2-itemsets are generated by combining 1-itemsets through join and pruning operations, and the support of the candidate 2-itemsets is calculated. A minimum support threshold is set for comparative analysis. Candidate itemsets whose support is greater than or equal to the set minimum support threshold are considered frequent itemsets, thus filtering out frequent 2-itemsets. This process is repeated recursively to generate candidate M+1-itemsets from frequent M-itemsets and to filter out frequent M+1-itemsets until no more frequent itemsets can be generated. For each frequent itemset m, generate all possible non-empty subset rules m' that conform to the rule m'->m-m'. Here, frequent itemset m represents the set of needs in the health and elderly care scenario, m' represents a subset in frequent itemset m, -> represents the co-occurrence relation symbol that implies the simultaneous occurrence of the difference m-m' when the subset m' appears, and m-m' represents the difference operation between the set of needs in frequent itemset m and the subset m', which represents the remaining elements in the set of needs m but not in the subset m'. Calculate the confidence level of each rule, set a minimum confidence threshold, and select rules with a confidence level greater than or equal to the preset minimum confidence threshold as valid association rules. Retain those rules that simultaneously meet the minimum support and minimum confidence thresholds to generate the demand association analysis results.

5. The intelligent matching platform for community health resources and home-based elderly care needs according to claim 4, characterized in that, The steps for establishing the multi-dimensional resource assessment indicator system are as follows: Identify the types of community health resources to be assessed based on the preprocessed resource status data; An assessment of the overall quality of community health resources is conducted, including an evaluation of resource availability, quality, and prioritization. Based on the assessment objectives, quantifiable assessment indicators are identified and decomposed to form a hierarchical structure of assessment objectives. These include availability indicators, which are divided into time matching degree and geographical accessibility indicators; quality indicators, which are divided into professional qualification level, medical equipment integrity rate and historical service user rating indicators; and priority indicators, which are divided into emergency rescue capability and scarcity coefficient indicators. The weight of each evaluation indicator is determined by using the Analytic Hierarchy Process (AHP). Experts are then invited to compare the indicators in the same level pairwise and score them based on their relative importance using the 1-9 scale. A judgment matrix is ​​then constructed for each upper-level indicator.

6. The intelligent matching platform for community health resources and home-based elderly care needs according to claim 5, characterized in that, The steps for generating the resource assessment index are as follows: Based on the AHP judgment matrix, the maximum eigenvalue of the judgment matrix and the eigenvectors of each corresponding evaluation index are solved. The feature vectors are normalized to obtain the weight vectors of each indicator. The consistency ratio is calculated, and a ratio threshold is set for comparative analysis. When the calculated consistency ratio is less than the set ratio threshold, the judgment matrix is ​​considered to have satisfactory consistency, and the weight vectors of each indicator are considered to be effective. Based on the determined weight vectors of each indicator, the preprocessed resource status data is processed by min-max normalization. After the evaluation indicators are quantified, standard values ​​are obtained and unified to the [0,1] interval. Then, the linear weighted summation method is used to generate the resource evaluation index of the comprehensive quality of health resources in each community.

7. The intelligent matching platform for community health resources and home-based elderly care needs according to claim 6, characterized in that, The steps for establishing the intelligent matching model are as follows: Based on the results of demand correlation analysis and resource assessment index, binary code is used to define decision variables for matching home-based elderly care needs with community health resources; A multi-objective function for fairness and efficiency is constructed, including the introduction of the fairness index of Gini coefficient and the efficiency index of resource utilization. Different demand categories and resource assessment indices are divided according to the K-Means clustering algorithm. The Gini coefficient is calculated by the sum of the per capita resource assessment indices allocated to each demand category. Minimizing the Gini coefficient to measure the maximization of fairness in resource allocation is taken as the fairness objective. Resource utilization rate is obtained by calculating the ratio of the sum of all allocated resource evaluation indices to the sum of total available resource evaluation indices, and maximizing resource utilization rate is taken as the efficiency objective. Set constraints, including demand uniqueness constraints, resource capacity constraints, geographical adaptability constraints, and non-negativity constraints for decision variables; Based on the set constraints, the functions are integrated into a multi-objective function to form a multi-objective optimization model.

8. The intelligent matching platform for community health resources and home-based elderly care needs according to claim 7, characterized in that, The steps for finding the optimal matching scheme are as follows: Based on the multi-objective optimization model, a set of initial matching schemes is randomly generated using a genetic algorithm, which serves as chromosomes. Each chromosome consists of decision variables and satisfies the constraints. Let the population size be L. For each chromosome, the fitness value of the matching scheme is calculated using a weighted summation method in conjunction with a multi-objective optimization model. Based on the fitness value of the matching scheme, Y chromosomes are selected from the current population as parents using the tournament selection method; A single-point crossover operation is performed on the selected parent chromosomes to generate offspring chromosomes; To increase population diversity, the chromosomes of offspring are modified. Add offspring chromosomes to the population, set a fitness threshold, recalculate the fitness of offspring chromosomes and compare it with the set fitness threshold, and eliminate some poor chromosomes based on the fitness value to update and retain the new population. Repeat the above steps of population selection, crossover, mutation, and pruning until the maximum number of iterations is reached or the fitness value no longer improves. Then, output the Pareto optimal solution set and find an optimal matching scheme.

9. The intelligent matching platform for community health resources and home-based elderly care needs according to claim 8, characterized in that, The steps for quantifying the actual resource demand data are as follows: We collect user feedback through multiple channels, including online questionnaires, telephone follow-ups, and mobile application feedback, including raw text data on user satisfaction, resource utilization, and experience opinions. The collected raw text data is preprocessed to remove special characters / punctuation marks, segment words, and filter stop words, resulting in preprocessed text data. The TF-IDF algorithm, a text mining technique, is used to extract keywords related to resource demand from preprocessed text data and integrate them into a keyword set. The pre-trained sentiment analysis BERT model is used to perform sentiment analysis on each feedback message containing keywords to determine the user's sentiment tendency towards resource demand, including positive, negative, and neutral. Based on the keyword and sentiment analysis results, and by counting the number of times each keyword appears in different sentiment tendencies, the actual resource demand data is quantified.

10. The intelligent matching platform for community health resources and home-based elderly care needs according to claim 9, characterized in that, The parameter adjustment steps for the intelligent matching model are as follows: The fairness and efficiency indicators of the intelligent matching model, including the Gini coefficient and resource utilization rate, recorded after the execution of historical matching schemes, are used as independent variables; The actual resource demand data collected and quantified after the corresponding matching scheme execution period is used as the dependent variable; Linear regression analysis was used to establish a multiple linear regression model. The least squares method was used to fit the historical dataset to obtain the regression coefficients and error terms. Based on the magnitude and sign of the regression coefficients, analyze the degree and direction of the influence of the independent variables on the dependent variable, and adjust the weight parameters of the intelligent matching model based on the results of the influencing factor analysis. The adjusted weights are then updated into the multi-objective function of the genetic algorithm to calculate the optimal solution for matching demand and resources in the next round.

Images