Intelligent old-age care service method, system and device based on AI and storage medium

Abstract

The invention provides an AI-based smart old-age care service method, system and device, and a storage medium, and relates to the technical field of smart old-age care. The method comprises the following steps: collecting vital sign data, behavior data and environment data of the elderly, and carrying out standardization processing on past case data to obtain standard case data; performing management fusion on the standard case data, the vital sign data, the behavior data and the environment data to obtain multi-dimensional health data; inputting the multi-dimensional health data into a preset health state evaluation model to obtain health state information; matching the health state information with maintenance knowledge in a preset knowledge base to obtain a plurality of alternative maintenance schemes, and inputting the plurality of alternative maintenance schemes into a preset maintenance effect prediction model to obtain a plurality of maintenance effect prediction results; and taking the alternative maintenance scheme with the highest evaluation score as a target maintenance scheme. Through the alternative scheme, the effect of the pension service is improved.

Description

Technical Field

[0001] This application relates to the technical field of intelligent elderly care, and specifically relates to an AI-based intelligent elderly care service method, system, device, and storage medium. Background Art

[0002] With the continuous aggravation of the aging degree of society, the issue of elderly care has increasingly become the focus of the whole society. The traditional elderly care model mainly relies on manual care. Caregivers need to identify the physical conditions of the elderly, formulate and implement maintenance plans based on experience. However, there are many deficiencies in this manual elderly care service model.

[0003] First of all, it is difficult for caregivers to continuously monitor the vital signs and behavior status of the elderly for 24 hours, which may delay the discovery and intervention of sudden health risks of the elderly. Secondly, caregivers need to manually collect and process the health information of the elderly. The information utilization efficiency is low and it is easy to make mistakes, which is not conducive to comprehensively evaluating the physical and mental health status of the elderly, and the subjective experience differences of different caregivers may lead to inconsistent health assessment criteria. Moreover, caregivers mainly formulate and implement maintenance plans based on their own experience and limited maintenance knowledge, lacking professional knowledge support to formulate more reasonable care plans, resulting in poor care effects. Summary of the Invention

[0004] This application provides an AI-based intelligent elderly care service method, system, device, and storage medium, which can improve the care effect.

[0005] In the first aspect, this application provides an AI-based intelligent elderly care service method, and the method includes: Real-time collection of the vital sign data, behavior data, and environmental data of the elderly through wearable devices worn on the elderly and smart home sensors arranged in the place where the elderly live; Obtain the past medical record data of the elderly, and perform standardized processing on the past medical record data to obtain standard case data; Manage and fuse the standard case data, vital sign data, behavior data, and environmental data to obtain multi-dimensional health data; Input the multi-dimensional health data into a preset health status assessment model to obtain health status information; Match the health status information with the maintenance knowledge in a preset knowledge base to obtain multiple alternative maintenance plans, and input the multiple alternative maintenance plans into a preset maintenance effect prediction model to obtain multiple maintenance effect prediction results; Evaluate the maintenance effect prediction results to obtain the evaluation scores of each alternative maintenance plan, and use the alternative maintenance plan with the highest evaluation score as the target maintenance plan.

[0006] By adopting the above technical solution, by wearing wearable devices on the elderly and arranging smart home sensors in their residences, the vital sign data, behavior data, and environmental data of the elderly can be comprehensively and real-timely collected, the health status and living conditions of the elderly can be perceived in all aspects, providing a rich data basis for subsequent health assessment and maintenance decision-making. At the same time, this method also obtains the past medical record data of the elderly, standardizes it to form standard case data, and then manages and integrates the standard case data with the real-time collected vital sign data, behavior data, and environmental data to obtain multi-dimensional health data. This fusion processing can fully explore the associations between different health data, form a three-dimensional and multi-level representation of the elderly's health status, and provide a more comprehensive and accurate basis for intelligent health status assessment.

[0007] Based on the above-obtained multi-dimensional health data, the present invention inputs it into a preset health status assessment model, and the model automatically discriminates the current health status information of the elderly. Compared with manual assessment, this AI-driven health status assessment method can quickly process a large amount of health data, timely detect the health abnormalities of the elderly, and the assessment results are more objective and accurate, not affected by human factors. Furthermore, this method matches the health status information with the maintenance knowledge in the preset knowledge base, and automatically infers multiple alternative maintenance plans for the current health status of the elderly. This knowledge matching mechanism can make full use of expert knowledge and maintenance practice experience, quickly generate personalized and highly targeted maintenance strategies, and provide strong support for intelligent elderly care decision-making.

[0008] On this basis, this method also inputs multiple alternative maintenance plans into a preset maintenance effect prediction model, quantitatively estimates the maintenance effects of each plan, and then screens out the target maintenance plan with the highest evaluation score by comparing the effect prediction scores of different plans for implementation. This intelligent decision-making method based on maintenance effect prediction can select the best plan from the alternative plans, maximize the maintenance effect, and improve the quality of elderly care services and the living quality of the elderly.

[0009] In the second aspect of the present application, an AI-based intelligent elderly care service method system is provided, including: A data acquisition module, configured to real-timely collect the vital sign data, behavior data, and environmental data of the elderly through wearable devices worn on the elderly and smart home sensors arranged in the residences of the elderly; A first data processing module, configured to obtain the past medical record data of the elderly and standardize the past medical record data to obtain standard case data; A second data processing module, configured to manage and integrate the standard case data with the vital sign data, behavior data, and environmental data to obtain multi-dimensional health data; A health status assessment module, configured to input multi-dimensional health data into a preset health status assessment model to obtain health status information; A maintenance effect prediction module, configured to match the health status information with maintenance knowledge in a preset knowledge base to obtain multiple alternative maintenance plans, and input the multiple alternative maintenance plans into a preset maintenance effect prediction model to obtain multiple maintenance effect prediction results; A maintenance plan determination module, configured to evaluate the maintenance effect prediction results to obtain an evaluation score for each alternative maintenance plan, and use the alternative maintenance plan with the highest evaluation score as the target maintenance plan.

[0010] In a third aspect of the present application, a computer storage medium is provided. The computer storage medium stores multiple instructions, and the instructions are adapted to be loaded and executed by a processor to perform the above method steps.

[0011] In a fourth aspect of the present application, an electronic device is provided, including a processor, a memory, a user interface, and a network interface. The memory is used to store instructions, the user interface and the network interface are used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device performs the above method.

[0012] In summary, one or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages: 1. In the present application, by wearing a wearable device on the elderly and arranging smart home sensors in their place of residence, vital sign data, behavior data, and environmental data of the elderly can be comprehensively and real-time collected, and the health status and living conditions of the elderly can be perceived in all aspects, providing a rich data basis for subsequent health assessment and maintenance decision-making. At the same time, this method also obtains the past medical record data of the elderly, performs standardized processing on it to form standard case data, and then manages and integrates the standard case data with the real-time collected vital sign data, behavior data, and environmental data to obtain multi-dimensional health data. This fusion processing can fully explore the associations between different health data, form a three-dimensional and multi-level representation of the health status of the elderly, and provide a more comprehensive and accurate basis for intelligent health status assessment.

[0013] 2. The multi-dimensional health data obtained in this application is input into a preset health status assessment model, which automatically determines the current health status information of the elderly. Compared with manual assessment, this AI-driven health status assessment method can quickly process a large amount of health data, timely detect health abnormalities of the elderly, and the assessment results are more objective and accurate, not affected by human factors. Furthermore, this method matches the health status information with the maintenance knowledge in the preset knowledge base, and automatically infers multiple alternative maintenance plans for the current health status of the elderly. This knowledge matching mechanism can make full use of expert knowledge and maintenance practice experience, quickly generate personalized and targeted maintenance strategies, and provide strong support for intelligent elderly care decision-making.

[0014] 3. This application inputs multiple alternative maintenance plans into a preset maintenance effect prediction model to quantitatively estimate the maintenance effect of each plan. Then, by comparing the effect prediction scores of different plans, the target maintenance plan with the highest evaluation score is selected for execution. This intelligent decision-making method based on maintenance effect prediction can select the best plan from alternative plans, maximize the maintenance effect, and improve the quality of elderly care services and the living quality of the elderly. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 It is a schematic flowchart of a method for AI-based intelligent elderly care service provided by an embodiment of this application; Figure 2 It is an architecture diagram of a system for AI-based intelligent elderly care service provided by an embodiment of this application; Figure 3 It is a schematic structural diagram of an electronic device provided by this application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of this specification. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all the embodiments.

[0017] In the description of the embodiments of this application, words such as "for example" or "for illustration" are used to indicate examples, illustrations, or explanations. Any embodiment or design solution described as "for example" or "for illustration" in the embodiments of this application should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, using words such as "for example" or "for illustration" is intended to present relevant concepts in a specific manner.

[0018] In the description of the embodiments of the present application, the term "plural" means two or more. For example, plural systems refer to two or more systems, and plural screen terminals refer to two or more screen terminals. In addition, the terms "first" and "second" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of such features. The terms "include", "comprise", "have" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0019] To facilitate the understanding of the method and system provided by the embodiments of the present application, before introducing the embodiments of the present application, the background of the embodiments of the present application will be introduced first.

[0020] With the acceleration of the aging population process, the issue of elderly care has become the focus of the whole society. In the traditional elderly care service model, it mainly relies on caregivers to provide manual care for the elderly. This requires caregivers to accurately judge the physical health status of the elderly and formulate corresponding maintenance plans based on their own experience and implement them. However, this traditional manual elderly care model has many limitations and is difficult to meet the growing demand for intelligent elderly care.

[0021] The primary problem is that caregivers cannot achieve 24-hour uninterrupted monitoring of the vital signs and behavior status of the elderly, which may lead to the delayed discovery and intervention of sudden health risks of the elderly, endangering the life safety of the elderly. Secondly, in the information collection and processing link, caregivers need to manually record and organize various health information of the elderly. This method is not only inefficient but also error-prone, and is not conducive to comprehensively and objectively evaluating the physical and mental health status of the elderly. At the same time, the subjective experiences and judgment criteria of different caregivers vary greatly, which may lead to inconsistent evaluation results of the health status of the same elderly. Moreover, when formulating and implementing maintenance plans, caregivers mainly rely on their own experience and limited maintenance knowledge, lacking necessary professional knowledge support and intelligent assistance means, and it is difficult to customize the optimal personalized care plan for the elderly, resulting in difficult-to-guarantee maintenance effects.

[0022] After the above background introduction, those skilled in the art can understand the problems existing in the prior art. Next, the technical solutions in the embodiments of the present application will be clearly and completely described with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments.

[0023] Based on the above background technology, further, please refer to Figure 1 , Figure 1The figure is a schematic flowchart of a smart elderly care service method based on AI provided by an embodiment of the present application. This system can be implemented relying on a computer program or run as an independent tool-like application. Specifically, in the embodiment of the present application, this method can be applied to a server, but can also be applied to electronic devices such as a server. A smart elderly care service method based on AI includes the following steps: S101, real-time collect the vital sign data, behavior data, and environmental data of the elderly through wearable devices worn on the elderly and smart home sensors arranged in the place where the elderly live; Specifically, wearable devices such as smart bracelets and smart watches are worn on the elderly. Heart rate sensors, temperature sensors, blood pressure sensors, etc. built into the devices are used to collect vital sign data such as the heart rate, temperature, and blood pressure of the elderly, and acceleration sensors, gyroscopes, etc. are used to collect behavior data such as the exercise amount, gait, posture, and falls of the elderly. At the same time, smart home devices equipped with pressure sensors, light sensors, door and window magnetic sensors, such as smart mattresses, smart floors, smart doors and windows, and smart lights, are arranged in the rooms where the elderly live to collect data such as the temperature, humidity, brightness, and safety status of the living environment of the elderly. The data collected by these wearable devices and smart home sensors are uploaded to the smart elderly care cloud platform in real time through the built-in wireless communication module (such as WiFi, Bluetooth, ZigBee, etc.) and aggregated into a health big data resource library for the individual elderly on the platform.

[0024] The health data collected in this way is characterized by comprehensiveness, continuity, and objectivity, and can reflect the physical and mental health status of the elderly in a multi-dimensional and dynamic manner, providing detailed data support for subsequent health assessment, behavior analysis, abnormal warning, and maintenance decision-making. When the elderly have abnormal heart rate, sit still for a long time, get out of bed frequently at night, etc., the system can automatically identify and push warning information, so that the nursing staff and family members can understand the situation in a timely manner and check and intervene in time. At the same time, the accumulated large amount of health data also lays a foundation for elderly care big data analysis, health status modeling, knowledge base construction, etc.

[0025] For example, the smart bracelet worn by the elderly Xiao Wang detects that his heart rate has suddenly risen to 120 beats per minute, which is significantly higher than the normal value range. The system immediately sends a warning text message to the on-duty caregiver. The caregiver then goes to visit and finds that the elderly is emotional, so he soothes the emotion and feeds antihypertensive drugs, continuously monitors the changes in vital signs until the heart rate returns to normal. Nowadays, most elderly people have problems with sleep quality. By deploying pressure sensors under the beds of the elderly, the sleep status of the elderly can be monitored and analyzed in real time, such as the sleep duration, light and deep sleep time, and the number of times of getting up at night, so as to evaluate the sleep quality of the elderly and take sleep assistance measures to improve sleep when necessary.

[0026] S102. Obtain the previous medical record data of the elderly, and perform standardization processing on the previous medical record data to obtain standard medical record data; Specifically, first, by docking with the information system of medical institutions or accessing the personal health records after being authorized by the elderly, obtain the original medical data such as outpatient medical records, inpatient medical records, test reports, and imaging reports of the elderly during their visits to various hospitals. Then, use natural language processing technology to perform word segmentation, entity recognition, semantic analysis, etc. on the medical record text, and extract key information elements such as structured diagnoses, surgeries, and medications. Next, use standardized corpus resources such as medical vocabularies, disease classifications, and drug dictionaries to map the extracted information elements into standardized medical term expressions, removing the differences of synonyms and heteronyms. Finally, generate a standardized medical record data set for the individual elderly according to the unified data format specification.

[0027] Through the above processing, the unstructured and heterogeneous original medical record data can be converted into structured, standardized, and semantically consistent standard medical record data, which is convenient for computer programs to parse, store, retrieve, and utilize. The medical big data analysis platform can automatically generate the health records and comprehensive health status reports of the elderly based on the standard medical record data, and match them with the knowledge base to conduct risk assessment and grading early warning for the common chronic diseases of the elderly. For example, it is found from Lao Zhang's standard medical record data that he has a 20-year history of hypertension, was hospitalized for cerebral infarction 5 years ago, and his blood lipid was found to be severely elevated during a physical examination. The platform automatically determines that Lao Zhang is a high-risk group for stroke, gives an early warning prompt, and recommends relevant prevention and treatment measures.

[0028] S103. Manage and integrate the standard medical record data with the vital sign data, behavior data, and environmental data to obtain multi-dimensional health data; Specifically, first, according to the unified data format specification and field standard, perform cleaning, mapping, and normalization processing on multi-source heterogeneous data such as standard medical record data, vital sign data, behavior data, and environmental data to eliminate problems such as redundancy, ambiguity, and inconsistency between data from different sources. Then, associate and match various types of data according to the unique identifier of the elderly individual (such as ID card number, medical insurance card number, etc.), and organize and store the matched data according to the idea of "person-centered" to form a health data set with the elderly individual as the basic unit. On this basis, use technologies such as time series databases and graph databases to construct a data model and link map connecting multi-dimensional data, depicting the semantic connections between various types of data. Finally, through the combination of intelligent algorithms and expert experience, develop a series of cross-data-dimensional composite health assessment indicators to form an application scenario for multi-dimensional health data.

[0029] Based on the above embodiments, as an optional embodiment, managing and integrating the standard medical record data with the vital sign data, behavior data, and environmental data to obtain multi-dimensional health data includes: S201, conduct quality assessment on standard case data, vital sign data, behavior data, and environmental data to obtain the quality assessment scores of each piece of data; Specifically, construct targeted quality assessment index systems for the quality characteristics of standard case data, vital sign data, behavior data, and environmental data respectively. For example, for standard case data, focus on inspecting structural integrity, content consistency, diagnosis and treatment standardization, etc.; for vital sign data, focus on inspecting collection accuracy, reporting timeliness, sequence integrity, etc. Secondly, based on technologies such as association rules and data profiling, realize the automated assessment of quality indicators. For example, check the logical consistency of diagnostic information and test results through cross-reference, and analyze the proportion of outliers in vital sign data using statistical distribution. On this basis, determine the quality indicator weights based on the analytic hierarchy process, calculate the comprehensive quality assessment scores of each piece of data through weighted average, and set the quality scoring grades and threshold standards. Thirdly, for data of different quality levels, match corresponding quality improvement paths. For data with quality scores lower than the preset threshold, conduct verification and correction through technologies such as data cleaning and data repair, or directly filter and eliminate them to prevent the backward transmission of quality defects. For high-quality data that meets the standards, enter the subsequent management and integration link to provide support for applications such as health care knowledge discovery and auxiliary decision-making. Finally, embed the entire process of quality assessment into the health care big data platform to form an automated and normalized quality "physical examination" mechanism, and use the quality assessment report to continuously guide and improve business processes such as medical data collection, storage, and exchange, so as to achieve the closed-loop improvement of data quality.

[0030] S202, construct a semantic mapping knowledge base of health data based on medical concepts, and fuse the quality assessment scores of each piece of data with the semantic mapping knowledge base of health data to obtain a multi-dimensional data fusion model; Specifically, a semantic mapping knowledge base for health data is constructed based on a medical ontology. Taking clinical medical concepts as the primitives of knowledge representation, the semantic relationships between concepts are characterized using ontology description languages to form a hierarchical, structured, and extensible concept system. Secondly, based on technologies such as natural language processing and data mining, semantic annotation is performed on standard case data, vital sign data, behavioral data, and environmental data. Medical concepts contained in the data are extracted through methods such as named entity recognition and word sense disambiguation, and the internal relationships between concepts are explored using methods such as similarity calculation and association rule mining to form a semantic index of multi-source data. Thirdly, the quality scores of various types of data are embedded into the semantic mapping knowledge base to form a multi-dimensional data fusion model that integrates semantic connotations and quality attributes. By setting quality tags in the knowledge base, the quality scores are associated with the corresponding concepts to achieve the semantic representation of data quality. At the same time, methods such as link analysis and social network analysis are used to characterize the conduction mechanism of quality scores in the concept network, that is, the "driving effect" of high-quality concepts on low-quality concepts, and a quality fusion model oriented by semantics is constructed. Finally, based on the fusion model, panoramic semantic quality management of health data is realized, providing accurate and efficient semantic retrieval services for data applications. When a user retrieves a certain medical concept, the system can not only return multi-source data related to the concept but also present the quality attributes of these data to assist the user in selecting reliable data. At the same time, the fusion model can real-time associate quality problem data with improvement paths, forming a closed-loop process of quality problem tracing, improvement measure formulation, and improvement effect evaluation to continuously improve the quality of data fusion.

[0031] S203, Input the standard case data, vital sign data, behavioral data, and environmental data into the multi-dimensional data fusion model to obtain multi-dimensional health data.

[0032] Specifically, according to the characteristics of standard case data, vital sign data, behavioral data, and environmental data, corresponding data acquisition interfaces are designed to complete the real-time aggregation of multi-source data. For example, the medical information of patients is collected through the data interface of the hospital information system, and the vital sign monitoring data of residents is uploaded using wearable devices. Secondly, the collected multi-source data is input into the semantic mapping module. Through techniques such as named entity recognition and word sense disambiguation, the medical concepts contained in the data are extracted, and the concepts are mapped to the standard vocabulary of the semantic knowledge base to achieve the standardization of the semantic representation of multi-source data. For example, the "hypertension" in the medical record is mapped to the concept of "essential hypertension" in the knowledge base. Thirdly, according to the quality scores embedded in the fusion model, the quality calculation is carried out on the mapped concept data to obtain the fusion quality scores of each concept. A quality score threshold is set. For concepts below the threshold, trace back to the corresponding original data and perform quality improvement processes such as data cleaning, repair, and filtering. For concepts that meet the quality requirements, they are included in the fusion result set to generate a high-quality and semantically rich health knowledge graph. Finally, through technologies such as graph databases and semantic retrieval interfaces, the knowledge graph provides convenient and efficient data services for multi-scenario applications such as medical treatment, healthcare, scientific research, and management. Users can intuitively explore the semantic connections between concepts through the graph. The system can intelligently recommend relevant concepts according to semantic similarity and display the original data details behind the concepts in an interactive visual analysis manner, facilitating users to follow the data, trace the origin, and gain insights into health facts.

[0033] S204, a data acquisition module, is used to collect the vital sign data, behavioral data, and environmental data of the elderly in real time through wearable devices worn on the elderly and smart home sensors arranged in the place where the elderly live; Specifically, based on various indicators that significantly affect the health status of the elderly, wearable devices and smart home devices with compatible functions and reliable performance are selected. For example, intelligent bracelets integrated with functions such as electrocardiogram, blood pressure, blood oxygen, body temperature, and steps, as well as home sensors integrated with environmental monitoring functions such as formaldehyde and PM2.5. Secondly, considering factors such as the acceptance degree and convenience of elderly users, the wearing scheme of wearable devices and the layout scheme of home sensors are optimized to maximize the user experience while ensuring the quality of the collected data. For example, designing a "one-button" bracelet wearing process that conforms to the usage habits of the elderly, and selecting miniature home sensors with strong concealment and convenient installation. Thirdly, a data transmission "pipeline" between wearable devices and smart homes is established. Through communication protocols such as low-power Bluetooth and WIFI, multi-source data such as the vital signs, exercise steps, sleep duration, and home temperature and humidity of the elderly are collected and transmitted back to the intelligent elderly care service platform in real time. Finally, preprocessing such as cleaning, repairing, and desensitization is carried out on the collected multi-source data. Combining with the semantic mapping knowledge base and the multi-dimensional fusion model, the dynamic shaping of the "holographic portrait" of the elderly's health is realized, providing high-quality data assets for various health service applications. For example, the vital signs such as the heart rate and blood pressure of the elderly are monitored in real time. Once abnormal fluctuations occur, the platform automatically issues a warning to notify the children and community medical staff to provide first aid to the elderly in a timely manner. Another example is to comprehensively analyze data such as the diet behavior, exercise habits, and indoor air quality of the elderly, and use intelligent algorithms to generate personalized health risk assessment reports to provide accurate health improvement plans for the elderly.

[0034] S104. Input the multi-dimensional health data into a preset health status assessment model to obtain health status information. Specifically, first, machine learning algorithms such as random forest, support vector machine, and neural network are used. Taking multi-dimensional health data as sample input and the comprehensive health status scored by experts as labels, the health status assessment model is trained and optimized. The input features of the model include data indicators in multiple dimensions such as the chronic medical history, physical examination report, daily vital signs, exercise volume, sleep quality, emotional state, and living environment of the elderly. Through feature engineering and parameter tuning, the model fully learns the key evaluation factors. In the application stage, input the multi-dimensional health data of the elderly to be evaluated into the pre-trained evaluation model, and the model can automatically determine the comprehensive health status of the elderly according to the data features, forming structured health status information including the total health score, sub-item scores, and risk warning levels. The health status information can intuitively present the comprehensive performance of the elderly in aspects such as physiological function, self-care ability, cognitive ability, and social participation, and clearly point out the main health risk factors.

[0035] Based on the above embodiments, as an optional embodiment, before inputting the multi-dimensional health data into a preset health status assessment model to obtain health status information, it further includes: S301. Obtain historical sample data, classify the historical sample data into health levels based on expert knowledge to obtain the health levels of each historical sample data, and construct a training dataset based on the historical sample data and the health levels of each historical sample data; Specifically, first, adopt a combination of manual collection and automatic collection to widely collect health data from multiple sources and of various types, such as the historical health records, physical examination reports, and life logs of community elderly people, covering the physiological, psychological, behavioral and other health information of the elderly in the past period, and form a preliminary historical sample database. Then, invite medical experts to screen and clean the historical sample data, remove the incomplete, inconsistent, and incorrect dirty data, and convert the unstructured data into a structured form. On this basis, authoritative experts in the fields of clinical, nursing, psychology, nutrition, etc., make a comprehensive judgment on the health status of each historical sample data according to past diagnosis and treatment experience and industry guidelines, and refer to general standards such as geriatric syndromes and ADL scores, and assign health level labels (such as excellent, good, general, poor, extremely poor), and finally form a training dataset with authoritative annotations. The scale of the dataset should reach thousands or tens of thousands of cases to ensure the sufficiency of model training. Continuously accumulate high-quality training data through a combination of human and machine to provide data support for the continuous iteration and upgrade of the health status assessment model.

[0036] S302. Extract features from the historical sample data to obtain sample feature data, and use a feature screening algorithm to screen the sample feature data to obtain a feature subset dataset; Specifically, first, for different types of historical sample data, adopt corresponding feature extraction methods to automatically extract the most valuable features for discriminating the health status from the original data. For structured physical examination report data, statistical analysis, data mining and other technologies can be used to extract statistical features such as the mean, variance, and peak value of physiological indicators such as blood pressure, heart rate, and blood sugar, and extract features such as the frequency and time interval of medication and medical record visits. For unstructured medical image data, image processing, pattern recognition and other technologies can be used to extract imaging omics features such as morphology, texture, and gray histogram. For time series wearable device monitoring data, time-frequency analysis and signal processing technologies are used to extract the frequency domain and time domain features of electrocardiogram and accelerometer signals, and characterize the periodicity and stability of physiological rhythms. For text-based self-reports, questionnaires and other data, natural language processing technologies are used to extract semantic features such as sentiment tendency and theme word distribution. Through all-round and multi-angle feature extraction, the effective information contained in the original health data is condensed and refined to the greatest extent, and the historical sample data is converted into sample feature data with regular structure and concentrated information.

[0037] After obtaining the sample feature data, further use feature selection algorithms such as filter, wrapper, and embedded to automatically evaluate the contribution of each feature to the discrimination of the health status, and accordingly select the most discriminative feature subset. Commonly used feature selection algorithms include variance selection method, chi-square test, mutual information method, LASSO regression, decision tree, etc. Through feature selection, the noise data and redundant information in the sample features are removed, the dimension of the learning task is reduced, and the curse of dimensionality problem is alleviated. The finally obtained feature subset can best reflect the internal mechanism of the health status of the elderly, contains rich health influencing factors, and can be used for the training of the subsequent evaluation model. To further optimize the feature subset, optimization strategies such as ensemble learning and adaptive weighting can be adopted, the results of multiple feature selection algorithms are integrated, the advantages and disadvantages of each algorithm are balanced, and a feature combination with better stability and robustness is selected to further improve the feature quality.

[0038] S303. Construct a preliminary health status evaluation model based on the feature sub-data, and train the preliminary health status evaluation model based on the training data set. When the evaluation accuracy rate reaches the preset accuracy threshold, obtain the preset health status evaluation model.

[0039] Specifically, first select a suitable machine learning algorithm (such as logistic regression, support vector machine, random forest, neural network, etc.) according to the application scenario and task type of the health status evaluation, and build a preliminary health status evaluation model framework based on the feature sub-data set. After determining the model framework, use methods such as cross-validation to divide the training data set with health level labels into a training set and a validation set. Use the training set samples to train the preliminary model, and through multiple rounds of iteration, optimize the internal weight parameters of the model so that it can fully learn the key features and discrimination rules of the health status evaluation from the data. After each round of iteration, use the validation set samples to evaluate the performance of the trained model, and comprehensively examine the classification and discrimination ability of the model using evaluation indicators such as accuracy rate, precision rate, recall rate, and F1 value. When the evaluation accuracy rate and other indicators of the model on the validation set reach the preset threshold (such as 90%), it can be determined that the current model has reached the optimal state and has a stable and reliable health evaluation ability, and it is officially determined as the health status evaluation model of the present invention. To further improve the model performance, hyperparameter optimization strategies such as grid search and random search can be adopted to automatically find the optimal model parameter combination. At the same time, strategies such as L1, L2 regularization, and Dropout are introduced to prevent the model from overfitting and improve its generalization ability.

[0040] S105. Match the health status information with the maintenance knowledge in the preset knowledge base to obtain multiple alternative maintenance plans, and input the multiple alternative maintenance plans into the preset maintenance effect prediction model to obtain multiple maintenance effect prediction results; Based on the above embodiment, as an optional embodiment, the health status information is matched with the maintenance knowledge in the preset knowledge base to obtain multiple alternative maintenance plans, including: S401, obtain maintenance knowledge and build domain ontology based on maintenance knowledge; Specifically, we first collect authoritative literature in the field of elderly care, including textbooks, guidelines, cases, and literature, to form a preliminary knowledge base for nursing care. The knowledge base covers all aspects of nursing knowledge, such as chronic disease management, dietary nutrition, sports rehabilitation, and psychological care, but most of them exist in the form of natural language texts, lacking semantic structure, which is not conducive to direct computer application. To overcome this problem, we use manual or automatic methods to semantically annotate the knowledge base, extract the core concepts in the field of elderly care (such as hypertension, dizziness, low-salt diet, Tai Chi, etc.), and clarify the semantic relationships between concepts (such as "cause", "treatment", "relief", etc.). On this basis, with reference to ontology construction methodologies such as METHONTOLOGY and TOVE, knowledge engineers and domain experts work together to construct an ontology in the field of elderly care. First, determine the core concept level of the ontology, such as health status, symptoms and signs, treatment methods, nursing items, evaluation indicators, health care, etc. Then, the system defines the connotation and extension of each concept, clarifies the subordinate relationship between concepts (such as "disease-symptoms", "treatment-drugs", etc.), and describes the concept attributes (such as "drugs" have "usage and dosage" attributes). Next, the ontology axioms are defined to constrain the values ​​of concept attributes and enhance semantic integrity (such as "drug contraindications" specify which groups of people should use which drugs with caution). Finally, with the help of ontology editing tools such as Protégé, the ontology of the elderly care field is formally represented, and ontology evaluation methodologies such as OntoClean and OOPS! are used to evaluate the consistency, accuracy, redundancy, etc. of the ontology, forming a high-quality knowledge ontology library in the elderly care field.

[0041] S402, performing semantic extraction on the health status information to obtain a structured representation of the health status information, and mapping the structured representation to a concept node corresponding to the domain ontology to form a semantic representation of the health status information at the ontology concept level; Specifically, first, preprocess the collected health status information of the elderly, such as text denoising, segmenting sentences, etc., to improve the information quality. Then, use Chinese word segmentation tools such as jieba and THULAC to segment the health information text into meaningful lexical units. Next, use part-of-speech tagging tools such as LTP and StanfordPOS Tagger to determine the part-of-speech roles of each word. On this basis, with the help of syntactic analysis tools such as DD-Parser and Berkeley Parser, reveal the structured relationships such as modification and coordination between words. At the same time, use named entity recognition models in the medical field such as BERT-CRF and BiLSTM-CRF to extract key concepts such as diseases, symptoms, body parts, and examination indicators from the health information. For example, identify key entities such as "hypertension", "dizziness", and "headache" from "The patient's blood pressure is 140 / 95 mmHg, accompanied by dizziness and headache". Furthermore, use techniques such as semantic role labeling and dependency syntactic analysis to extract semantic relationships between concepts, such as triples in the form of <hypertension, accompanied by, dizziness> and <hypertension, accompanied by, headache>, so as to realize the conversion of health status information from "unstructured text" to "structured semantic representation". Finally, map and align the extracted structured representation with the ontology in the elderly care field. Use methods such as string matching, dictionary matching, and semantic matching to identify the semantic nodes corresponding to the key concepts in the health status representation in the ontology knowledge base, and construct semantic links between concepts. For example, map "hypertension" to the "disease" branch of the ontology, map "dizziness" and "headache" to the "symptom" branch, and establish a semantic association at the concept level of <disease, causes, symptom>. At the same time, based on the attribute definitions in the ontology, enrich the semantic information of the concept nodes, such as the blood pressure measurement value and disease grading of hypertension, and construct a semantic representation map of the health status information on the knowledge ontology.

[0042] S403, map the semantic representation of the health status information at the ontology concept level to the corresponding concept nodes of the domain ontology, obtain the ontology concepts, ontology instances, semantic relationships between concepts, and concept clustering of the health status information, and construct a multi-dimensional representation of the health status information in the ontology semantic space based on the ontology concepts, ontology instances, semantic relationships between concepts, and concept clustering; Specifically, first, based on the semantic mapping relationship between health status information and the domain ontology, the ontology concepts directly related to the health status are extracted from the ontology knowledge base. For example, from each health status instance mapped to the "disease" concept branch, specific disease types such as "hypertension", "diabetes", and "coronary heart disease" are extracted; from the instances mapped to the "symptom" concept branch, the main symptom manifestations such as "dizziness", "fatigue", and "palpitation" are extracted. These standardized ontology concepts eliminate the expression differences in the health information text and uniformly represent the health status at the semantic level. Secondly, the extracted ontology concepts are instantiated based on the ontology structure, and ontology instances are used to represent the specific health characteristics of a certain elderly person. For example, based on the ontology relationship of <disease, population with the disease, elderly people>, the ontology instance of "Zhang San has hypertension" is generated; based on the concept attribute definition of <symptom, duration, 3 years>, the instance knowledge of "Zhang San's dizziness symptom lasts for 3 years" is generated. By instantiation, the abstract concepts are associated with specific individuals, forming a fine-grained health characteristic representation for individuals. Thirdly, the ontology reasoning mechanism is used to mine the implicit semantic relationships between concepts and reveal the internal laws of the health status. For example, based on the prior knowledge such as <hypertension, complication, myocardial infarction> and <hypertension, risk factor, high-salt diet>, new knowledge such as "If Zhang San eats high-salt food for a long time, it is likely to cause myocardial infarction" is inferred to predict the deterioration trend of individual health. These implicit relationships help to reflect the action mechanism between health risk factors from multiple aspects. Finally, methods such as semantic similarity calculation are used to perform clustering analysis on the ontology concepts to discover the internal associations of different health characteristics. For example, the elderly groups with similar disease spectra and similar medication patterns are clustered into one category to form different health subgroups, which is convenient for targeted treatment and precise implementation of policies. Mapping the originally discrete health statuses to concept clusters can simplify the semantic representation dimension and highlight the common characteristics of different health subgroups. Combining the above information, a multi-dimensional health status representation including ontology concepts, ontology instances, concept relationships, and concept clusters can be constructed, showing the semantic panoramic view of elderly health from abstract to concrete and from appearance to essence.

[0043] S404, using the multi-dimensional representation as the retrieval condition to match the maintenance knowledge in the preset knowledge base, and obtaining multiple alternative maintenance plans.

[0044] Specifically, based on specific knowledge representation frameworks such as ontologies and knowledge graphs, explicit knowledge such as authoritative literature, expert experience, and diagnosis and treatment norms in the field of elderly care is formally and structurally represented and uniformly organized into a preset maintenance knowledge base. For example, elements such as drug ingredients, usage and dosage, and contraindications in drug instructions are extracted as ontology concepts, and the health preservation secrets of experts are manually sorted to form a knowledge graph of health preservation rules, forming a structured knowledge base with both breadth and depth, covering multiple dimensions of elderly care such as disease diagnosis, drug treatment, dietary nutrition, exercise rehabilitation, and psychological care. Secondly, using the multi-dimensional semantic representation of the health status as the retrieval condition, a semantic-based knowledge retrieval technology is adopted to discover nursing plans that highly match the health characteristics of the elderly from a vast amount of maintenance knowledge. For example, based on semantic similarity calculation, identify the treatment plan most relevant to the combination of the elderly's diseases and concurrent symptoms; based on association rule mining, discover the corresponding relationship between the elderly's symptoms, signs and specific traditional Chinese medicine prescriptions for preventing disease; based on ontology reasoning, according to the elderly's eating habits and exercise ability, match the appropriate dietary and exercise plans from the nutritional recipe library and physical exercise library. During the matching process, not only the semantic similarity between the characteristics of each dimension of the health status and the candidate plan should be considered, but also the confidence level, pros and cons weight of different maintenance knowledge should be weighed, and then the best candidates are selected from multiple alternative plans to generate several maintenance plans with the highest multi-dimensional matching degree. Since these plans can fully fit the health status and maintenance needs of the elderly, they are more likely to receive good intervention effects in subsequent nursing practices.

[0045] Based on the above embodiments, as an alternative embodiment, using the multi-dimensional representation as the retrieval condition to match the maintenance knowledge in the preset knowledge base, multiple alternative maintenance plans are obtained, including: Convert the multi-dimensional expression into a structured retrieval formula, and use the retrieval formula to match the maintenance knowledge in the preset knowledge base to obtain multiple alternative maintenance plans.

[0046] Specifically, first, perform syntactic parsing and syntactic analysis on the multi-dimensional semantic representation of the health status to identify different components such as concept words, instance words, and relationship words. Taking the disease concept as an example, extract disease concept words such as hypertension and diabetes; taking the symptom instance as an example, extract the symptom name "dizziness" and the duration "3 years" in "Patient Zhang San has had dizziness for 3 years". Secondly, based on the domain dictionary, semantic rules, etc., perform part-of-speech tagging and synonym normalization on the retrieval words of different types of components. For example, uniformly tag symptom words as "SYM", tag symptom durations as the numerical part-of-speech "NUM", and merge synonyms such as "dizziness" and "vertigo" into the canonical term "dizziness" in the canonical vocabulary. This standardization process eliminates the morphological variations of the retrieval words and unifies the semantic expressions, facilitating subsequent matching. Thirdly, extract retrieval words with standardized forms from different dimensions and construct a structured retrieval formula using Boolean logic, wildcards, semantic roles, etc. For example, organize the retrieval words in the disease dimension in the form of "DIS=hypertension AND DIS=coronary heart disease", and express the symptom characteristics in the form of "SYM=dizziness AND DUR>1 year". At the same time, use Boolean operators such as AND, OR, and NOT to connect different retrieval words to form a hierarchical structure of the retrieval formula with clear priorities. Finally, store the multi-dimensional retrieval formula in a standardized format to form a reusable structured query. When it is necessary to match the maintenance plan in the knowledge base, the system automatically loads the corresponding retrieval formula, extracts the keywords and logical structure therein, and drives the knowledge base to carry out accurate and efficient matching queries to quickly lock the most relevant candidate plans.

[0047] For example, for the elderly Mr. Wang who suffers from hypertension and diabetes and has symptoms such as dizziness and fatigue, the system constructs multiple retrieval expressions for diseases, symptoms, signs, lifestyle, etc. based on the multi-dimensional semantic representation of the health status. Taking the disease dimension as an example, disease concept words such as "hypertension" and "type 2 diabetes" are identified and normalized into formal expressions such as "DIS=HBP" and "DIS=T2DM", and organized into a disease retrieval expression of "DIS=HBP AND DIS=T2DM", indicating that the patient suffers from these two chronic diseases at the same time. Taking the symptom dimension as an example, symptom words such as "dizziness" and "fatigue" and the duration of "3 years" are extracted and standardized into a multi-condition retrieval expression of "SYM=dizziness AND SYM=fatigue AND DUR>2". The retrieval expressions of each dimension are connected by AND to form a structured total retrieval expression of "(DIS=HBP AND DIS=T2DM) AND (SYM=dizziness AND SYM=fatigue AND DUR>2) AND...", covering multiple conditions such as diseases, symptoms, and duration. The system parses this retrieval expression into a query statement and simultaneously performs matching in multiple sub-libraries such as the treatment plan, drug instructions, and health assessment in the maintenance knowledge base. For example, in the treatment plan library, the disease conditions of the retrieval expression match the indication conditions of the diabetes drug treatment plan, and the symptom characteristics match the syndrome characteristics of traditional Chinese medicine syndrome differentiation and treatment, and then alternative plans such as oral hypoglycemic drugs and blood-activating and stasis-removing decoctions are matched. Through the well-organized structured retrieval, the system quickly extracts the nursing measures that best meet the health needs of the elderly from the vast amount of information in the knowledge base, making the knowledge matching process no longer "searching for a needle in a haystack", but going straight to the point, accurately and efficiently.

[0048] On the basis of the above embodiments, as an optional embodiment, before inputting multiple alternative maintenance plans into a preset maintenance effect prediction model to obtain multiple maintenance effect prediction results, it further includes: S501, obtain expert knowledge and define a maintenance effect index system based on the expert knowledge; Specifically, senior experts in various fields such as medicine, nursing, psychology, and social work are invited to fully explore the valuable experience summarized and refined by experts in the practice of elderly care through brainstorming, in-depth interviews, Delphi method, etc., and sort out the key consideration factors for evaluating the maintenance effect. For example, clinical experts put forward medical indicators such as treatment effect and complication incidence rate, nursing experts focus on functional status indicators such as quality of life and self-care ability, psychological experts emphasize psychological indicators such as emotional state and cognitive function, and social work experts focus on social adaptation indicators such as social participation and interpersonal relationships. Secondly, multiple rounds of expert discussions are carried out. On the basis of widely adopting expert opinions and guided by professional consensus, methods such as brainstorming, KJ method, and analytic hierarchy process are used to explore the internal logic of each consideration factor, systematically sort out the hierarchical, parallel, progressive and other structural relationships among the indicators, and build an index system framework such as a pyramid-shaped or fishbone-shaped with clear goals and strict logic. On this basis, further refine the connotation and extension, measurement unit, and excellent standard of each indicator to form a specific measurement plan that is operable and assessable. For example, "treatment effect" is refined into sub-indicators such as drug efficiency, clinical cure rate, and disease control rate, and the calculation formula and percentage standard of each rate value are clarified. Finally, after the framework is determined, invite authority experts to demonstrate and review the scientificity, practicability, and completeness of the index setting. When necessary, conduct small-sample verification, and continuously improve and optimize according to the feedback opinions. Finally, a comprehensive index system covering multiple dimensions such as medical effect, quality of life, psychological state, and social integration is formed and reported to the relevant industry associations for filing to standardize the index definition and unify the evaluation caliber.

[0049] S502, obtain the data related to the maintenance effect, and analyze the influencing factors of the maintenance effect on the data related to the maintenance effect to obtain the influencing characteristic data; Specifically, collect a wide range of historical maintenance cases from elderly care institutions and communities, covering structured and unstructured data in links such as admission assessment, maintenance services, and discharge follow-up. Through preprocessing such as data cleaning and association fusion, a high-quality maintenance big dataset is formed. Secondly, starting from the index system, select data tables and fields closely related to the maintenance effect, such as admission assessment forms, nursing record sheets, satisfaction questionnaires, follow-up forms, etc. Based on methods such as statistical analysis and data mining, characterize the correlation between the maintenance effect and multi-dimensional elements such as service items, object characteristics, service frequency, and participation. For example, use chi-square test to analyze the association strength between nursing items and satisfaction, and use factor analysis to explore the combined effects of characteristics such as object age, disease type, and self-care ability on the maintenance effect. On this basis, further introduce modeling methods such as multiple regression and structural equation to quantitatively explore the magnitude and direction of the effects of key influencing factors on the maintenance effect, and characterize the relative importance of elements based on model coefficients, weights, etc., and preliminarily screen out the main effect factors with the greatest contribution to the maintenance effect. Thirdly, use combined learning algorithms such as decision trees and random forests to automatically discover the interaction patterns most relevant to the maintenance effect from a large number of feature combinations, and combine feature engineering methods to construct comprehensive features such as portraits of maintenance objects and portraits of maintenance measures to characterize the interaction effects of key influencing factors, and further refine the high-order combined features that have a significant impact on the maintenance effect. Finally, evaluate the stability, interpretability, and acquisition difficulty of each feature, weigh the information gain and usage cost, and select a set of concise, robust, and usable influencing feature data to support the training and optimization of the subsequent effect prediction model.

[0050] For example, for the life care services in the field of community-based home care for the elderly, the system first collects multi-source heterogeneous data such as the home assessment form, service record sheet, and satisfaction questionnaire of the elderly in the community, cleans, integrates, and correlates them to form a multi-dimensional data mart covering elements such as demographic characteristics, health status, service type, service frequency, and service evaluation. Taking satisfaction as an example, cross-tabulation analysis is used to examine the differences in the impact of different types such as meal assistance services, cleaning services, and accompaniment services on the satisfaction distribution, and it is found that the satisfaction with meal assistance services is the highest. Furthermore, the dose-response relationship between service frequency and satisfaction is explored, and it is found that when the meal assistance frequency is more than 3 times a week, the satisfaction is significantly improved. In addition, it is also found that the elderly over 80 years old with moderate limitations in their daily living abilities generally have a high satisfaction with meal assistance services. Based on this, it is initially determined that service type, service frequency, age, and self-care ability are the main effect factors affecting satisfaction. On the basis of the main effects, further using association rule mining algorithms such as Apriori and FP-Growth, it is found that the elderly over 80 years old with moderate limitations in their self-care ability have the highest satisfaction when receiving meal assistance services more than 3 times a week, and thus the high-order combined feature of "elderly + moderately limited + high-frequency meal assistance" is refined. After multiple rounds of iterative analysis and expert demonstration, finally, 4 main effect features including service type, service frequency, elderly age, and self-care ability and 1 interaction effect feature are determined as the key influencing factors for predicting the satisfaction of home care for the elderly, and are used to guide the design and development of the satisfaction prediction model. At the same time, data such as service type and service frequency at the maintenance site are included in the scope of rapid collection to form a real-time data monitoring and analysis closed-loop throughout the service process, providing a basis for dynamically adjusting the service plan and continuously optimizing the service experience.

[0051] S503. Build an initial maintenance effect prediction model based on the impact feature data, and train the initial maintenance effect preset model based on the maintenance effect-related data. When the prediction accuracy reaches the preset prediction accuracy threshold, obtain the preset maintenance effect prediction model.

[0052] Specifically, based on the obtained impact feature data, targeted initial prediction models are constructed for different prediction targets (such as satisfaction, complication incidence rate) and maintenance scenarios. According to the feature type and data structure, a variety of basic learners such as linear regression, logistic regression, decision tree, and support vector machine are flexibly adopted, and combined with strategies such as ensemble learning and transfer learning to build a personalized prediction model framework. During model training, with the impact features as the input and the effect indicators as the output, supervised learning is carried out using the data related to the maintenance effect. Through optimization strategies such as error backpropagation, the model parameters gradually converge, and the fitting ability between the feature combination and the maintenance effect is continuously improved. For example, for the prediction of maintenance satisfaction, the non-linear relationship between features such as service type, service frequency, and self-care ability and the level of satisfaction can be characterized by a logistic regression model with L1 regularization. When tuning the model, methods such as cross-validation are used to evaluate the prediction accuracy of the model, hyperparameters are optimized by techniques such as grid search, and strategies such as data augmentation and sample weighting are introduced to improve the generalization ability and robustness of the model until the prediction accuracy stably reaches the preset threshold (such as 90%), that is, a mature prediction model that can be used in actual business scenarios is formed. During application deployment, the inference performance is optimized through hierarchical design, parallel computing, etc., and based on model interpretation and visualization techniques, the prediction results are transformed into intuitive and credible decision-making references to assist professionals in scientifically formulating and dynamically optimizing the maintenance plan.

[0053] For example, for the satisfaction prediction in the field of community-based home care for the elderly, the system first constructs a satisfaction prediction model based on logistic regression, using the four main effect features of service type, service frequency, elderly age, and self-care ability, and one interaction effect feature that have been refined. The features and satisfaction are mapped into a vector matrix, and the sigmod function is used to fit the non-linear relationship between the feature combination and satisfaction. Parameter estimation is carried out based on the maximum likelihood method to form an initial prediction model that converts the feature input into a satisfaction probability output. During model training, historical home care data is used for supervised learning, and the parameters are iteratively optimized through the mini-batch gradient descent algorithm. L1 regularization is used to prevent overfitting. At the same time, a cross-validation mechanism is introduced, and the sample data is divided into a training set and a validation set. Hyperparameters such as the regularization coefficient and learning rate are optimized through techniques such as grid search to continuously improve the prediction accuracy of the model. In addition, for the problem of sample imbalance, oversampling algorithms such as SMOTE are used for data augmentation, and the minority class samples are weighted to further improve the robustness of the model. Finally, when the prediction accuracy of the model on the validation set stably reaches over 95%, the evolution from the initial model to the mature model is completed. In practical applications, community workers enter information such as the elderly's age and self-care ability during home visits, and real-time service records such as meal assistance and cleaning are collected. The system automatically conducts satisfaction prediction and generates early warning prompts to assist managers in continuously improving service quality. At the same time, by visually presenting the weight coefficients of each feature and the satisfaction sensitivity, service personnel can easily understand the correlation logic between the influencing factors and satisfaction, and optimize the service targeted.

[0054] S106. Evaluate the prediction results of the maintenance effect to obtain the evaluation scores of each alternative maintenance plan, and use the alternative maintenance plan with the highest evaluation score as the target maintenance plan.

[0055] Specifically, first invite experts in the fields of medicine, nursing, psychology, management, etc. to subjectively score the expected effects of the candidate plans from the perspectives of the improvement amplitude of the maintenance effect, implementation feasibility, cost economy, and elderly comfort, etc., to obtain the authoritative scores of each plan on each index. Then, the analytic hierarchy process is used to determine the weight coefficients of each evaluation index, and the weighted average method is used to calculate the comprehensive evaluation scores of each plan. On this basis, the scoring results are manually reviewed and fine-tuned in combination with the actual situation of the elderly. For example, if the elderly have limited financial conditions, the weight of the economic index can be appropriately increased; if the elderly have poor compliance, the feasibility score can be appropriately emphasized. Finally, the alternative maintenance plan with the highest evaluation score is determined as the optimal target maintenance plan to guide subsequent maintenance practices. At the same time, the system stores the decision-making process as a case in the case base for optimizing the decision-making algorithm and evaluation rules.

[0056] For example, when recommending a personalized care plan for the elderly Mr. Wang who has lost his ability to take care of himself, after knowledge matching and effect prediction, the system gave three candidate plans: Plan A is to hire professional caregivers to provide home care every day, with excellent care effects but high costs; Plan B is for family members to take turns taking care of him and conduct regular medical consultations and follow-ups, with good care effects and moderate costs; Plan C is to use intelligent wearable devices for monitoring and cooperate with online doctor guidance, with acceptable effects and the lowest costs. After comprehensively scoring the candidate plans, it was found that although Plan A had the best health improvement effect, considering the financial situation of the elderly's family, the comprehensive score was not the highest. Plan B achieved a better balance between care effects and affordability, with the highest final score, and was selected as the target plan and implemented. Another example is when formulating a rehabilitation and nursing plan for the elderly Mrs. Li with cognitive impairment. Although Plan A had the highest intensity of rehabilitation training and the greatest potential for improvement, the elderly had poor compliance; Plan B adopted cognitive stimulation games that combined education with entertainment and traditional Chinese medicine intervention, with a high participation rate of the elderly; Plan C mainly relied on drug treatment, which was worry-free and labor-saving, but the long-term effects were difficult to guarantee. After model calculation, the comprehensive score of Plan B exceeded that of Plan A and was more suitable for Mrs. Li's actual situation, becoming the final recommended target care plan.

[0057] Please refer to Figure 2 , Figure 2 which is an architecture diagram of an AI-based intelligent elderly care service system provided by an embodiment of the present application. The AI-based intelligent elderly care service system may include: A data acquisition module 1, configured to collect the vital sign data, behavior data, and environmental data of the elderly in real time through wearable devices worn on the elderly and smart home sensors arranged at the place where the elderly live; A first data processing module 2, configured to obtain the past medical record data of the elderly and perform standardization processing on the past medical record data to obtain standard case data; A second data processing module 3, configured to manage and integrate the standard case data, vital sign data, behavior data, and environmental data to obtain multi-dimensional health data; A health status assessment module 4, configured to input the multi-dimensional health data into a preset health status assessment model to obtain health status information; A care effect prediction module 5, configured to match the health status information with the care knowledge in a preset knowledge base to obtain multiple alternative care plans, and input the multiple alternative care plans into a preset care effect prediction model to obtain multiple care effect prediction results; A care plan determination module 6, configured to evaluate the care effect prediction results to obtain the evaluation scores of each alternative care plan, and use the alternative care plan with the highest evaluation score as the target care plan.

[0058] It should be noted that: when the system provided in the above embodiments realizes its functions, only the division of the above function modules is used for illustration. In actual applications, the above functions can be allocated to different function modules according to needs, that is, the internal structure of the device is divided into different function modules to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept. For the specific implementation process, please refer to the method embodiments and will not be elaborated here.

[0059] Please refer to Figure 3 This application also discloses an electronic device. Figure 3 FIG. is a schematic structural diagram of an electronic device disclosed in an embodiment of this application. The electronic device 300 may include: at least one processor 301, at least one network interface 304, a user interface 303, a memory 305, and at least one communication bus 302 or end-to-end wireless communication.

[0060] Among them, the communication bus 302 is used to realize the connection and communication between these components.

[0061] Among them, the user interface 303 may include a display screen (Display) and a camera (Camera). Optionally, the user interface 303 may further include a standard wired interface and a wireless interface.

[0062] Among them, the network interface 304 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface).

[0063] Among them, the processor 301 may include one or more processing cores. The processor 301 connects various parts within the entire server through various interfaces and lines, and executes various functions of the server and processes data by running or executing instructions, programs, code sets, or instruction sets stored in the memory 305, and by calling the data stored in the memory 305. Optionally, the processor 301 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 301 may integrate a combination of one or more of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user interface, application programs, etc.; the GPU is responsible for rendering and drawing the content to be displayed on the display screen; the modem is used to process wireless communication. It can be understood that the above-mentioned modem may not be integrated into the processor 301 and may be implemented separately by a single chip.

[0064] Among them, the memory 305 may include random access memory (RAM) and may also include read-only memory. Optionally, the memory 305 includes a non-transitory computer-readable medium. The memory 305 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 305 may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing the operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-mentioned method embodiments, etc.; the data storage area may store the data involved in the above-mentioned method embodiments. Optionally, the memory 305 may also be at least one storage system located far from the aforementioned processor 301. Refer to Figure 3 , in the memory 305 as a computer storage medium, there may be included an operating system, a network communication module, a user interface module, and an application program of an AI-based intelligent elderly care service method.

[0065] In Figure 3In the electronic device 300 shown, the user interface 303 is mainly used to provide an interface for the user to input and obtain the data input by the user; while the processor 301 can be used to call the application program stored in the memory 305 and based on the AI-based intelligent elderly care service method. When executed by one or more processors 301, the electronic device 300 is caused to execute the method of one or more of the above-described embodiments. It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that this application is not limited by the described action sequence, because according to this application, certain steps can be in other sequences or performed simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily essential to this application. In the above embodiments, the descriptions of the respective embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0066] In several implementation manners provided by this application, it should be understood that the disclosed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of modules is only a logical function division. In actual implementation, there can be other division methods. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the couplings or direct couplings or communication connections shown or discussed with each other can be through some service interfaces. The indirect couplings or communication connections of the system or modules can be in an electrical or other form.

[0067] The modules described as separate components may or may not be physically separated. The components shown as modules may or may not be physical modules, that is, they can be located in one place or distributed to multiple network modules. Some or all of the modules can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0068] This application embodiment also provides a computer storage medium. The computer storage medium can store multiple instructions, and the instructions are suitable for being loaded and executed by a processor to perform the AI-based intelligent elderly care service method as shown in the above Figure 1 shown embodiment. The specific execution process can refer to the specific description of the Figure 1 shown embodiment and will not be elaborated here.

[0069] In addition, in each embodiment of this application, the various functional modules can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module. The above integrated modules can be implemented in the form of hardware or in the form of software functional modules.

[0070] When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can be stored in a computer-readable memory. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a memory and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present application. The aforementioned memory includes: various media such as USB flash drives, mobile hard disks, magnetic disks, or optical discs that can store program codes.

[0071] The above are only exemplary embodiments of the present disclosure, and the scope of the present disclosure cannot be limited thereby. That is, any equivalent changes and modifications made in accordance with the teachings of the present disclosure still fall within the scope covered by the present disclosure. After considering the specification and the practice of the present disclosure, those skilled in the art will easily think of other implementation manners of the present disclosure.

[0072] The present application aims to cover any variations, uses, or adaptive changes of the present disclosure that follow the general principles of the present disclosure and include well-known common knowledge or conventional technical means in the technical field not recorded in the present disclosure. The specification and the embodiments are only regarded as exemplary, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. An AI-based smart elderly care service method, characterized in that: The method comprises: Through wearable devices worn by the elderly and smart home sensors placed in their residences, the elderly’s vital signs, behavior and environmental data are collected in real time; Obtaining the elderly's past medical record data, and standardizing the past case data to obtain standard case data; Managing and integrating the standard case data, the vital sign data, the behavior data, and the environmental data to obtain multi-dimensional health data; Inputting the multidimensional health data into a preset health status assessment model to obtain health status information; Matching the health status information with maintenance knowledge in a preset knowledge base to obtain multiple alternative maintenance plans, and inputting the multiple alternative maintenance plans into a preset maintenance effect prediction model to obtain multiple maintenance effect prediction results; The maintenance effect prediction result is evaluated to obtain an evaluation score for each of the alternative maintenance plans, and the alternative maintenance plan with the highest evaluation score is used as the target maintenance plan.

2. The method according to claim 1, characterized in that Before inputting the multi-dimensional health data into a preset health status assessment model to obtain health status information, the method further includes: Acquire historical sample data, and classify the historical sample data into health levels based on expert knowledge to obtain the health level of each of the historical sample data, and construct a training data set based on the historical sample data and the health level of each of the historical sample data; Extracting features from the historical sample data to obtain sample feature data, and using a feature screening algorithm to screen the sample feature data to obtain a feature sub-dataset; A preliminary health status assessment model is constructed based on the characteristic sub-data, and the preliminary health status assessment model is trained based on the training data set. When the assessment accuracy reaches a preset accuracy threshold, the preset health status assessment model is obtained.

3. The method according to claim 1, characterized in that The health status information is matched with maintenance knowledge in a preset knowledge base to obtain multiple alternative maintenance plans, including: Acquire maintenance knowledge and construct a domain ontology based on the maintenance knowledge; Performing semantic extraction on the health status information to obtain a structured representation of the health status information, and mapping the structured representation to a concept node corresponding to the domain ontology to form a semantic representation of the health status information at the ontology concept level; Mapping the semantic representation of the health status information at the ontology concept level to the concept node corresponding to the domain ontology, obtaining the ontology concept, ontology instance, semantic relationship between concepts and concept clustering of the health status information, and constructing a multi-dimensional representation of the health status information in the ontology semantic space based on the ontology concept, the ontology instance, the semantic relationship between concepts and the concept clustering; The multi-dimensional representation is used as a search condition to match the maintenance knowledge in the preset knowledge base to obtain a plurality of alternative maintenance plans.

4. The method according to claim 3, characterized in that The multi-dimensional representation is used as a search condition to match the maintenance knowledge in the preset knowledge base to obtain a plurality of alternative maintenance plans, including: The multi-dimensional expression is converted into a structured search formula, and the search formula is used to match the maintenance knowledge in the preset knowledge base to obtain a plurality of alternative maintenance plans.

5. The method according to claim 1, characterized in that Before inputting the plurality of alternative maintenance schemes into a preset maintenance effect prediction model to obtain a plurality of maintenance effect prediction results, the method further includes: Acquire expert knowledge, and define a maintenance effect indicator system based on the expert knowledge; Acquire maintenance effect related data, and analyze the influencing factors of the maintenance effect on the maintenance effect related data to obtain influencing characteristic data; An initial maintenance effect prediction model is constructed based on the influencing feature data, and the initial maintenance effect preset model is trained based on the maintenance effect related data. When the prediction accuracy reaches a preset prediction accuracy threshold, the preset maintenance effect prediction model is obtained.

6. The method according to claim 1, characterized in that The standardization of the previous case data to obtain standard case data includes: Cleaning the past case data to obtain cleaned past case data; The cleaned previous case data is structured to obtain the standard case data.

7. The method according to claim 1, characterized in that The standard case data, the vital sign data, the behavior data and the environmental data are managed and integrated to obtain multi-dimensional health data, including: Performing quality assessment on the standard case data, the vital sign data, the behavior data, and the environmental data to obtain a quality assessment score for each data; Constructing a health data semantic mapping knowledge base based on medical concepts, and fusing the quality assessment scores of each data with the health data semantic mapping knowledge base to obtain a multidimensional data fusion model; The standard case data, the vital sign data, the behavior data and the environmental data are input into a multidimensional data fusion model to obtain the multidimensional health data.

8. An AI-based smart elderly care service system, characterized in that: The system comprises: The data acquisition module is used to collect the elderly's vital signs data, behavior data and environmental data in real time through wearable devices worn by the elderly and smart home sensors placed in the elderly's residence; The first data processing module is used to obtain the old patient's medical record data and perform standardization on the old case data to obtain standard case data; A second data processing module is used to manage and integrate the standard case data, the vital sign data, the behavior data and the environmental data to obtain multi-dimensional health data; A health status assessment module, used to input the multi-dimensional health data into a preset health status assessment model to obtain health status information; A maintenance effect prediction module, used for matching the health status information with maintenance knowledge in a preset knowledge base to obtain a plurality of alternative maintenance plans, and inputting the plurality of alternative maintenance plans into a preset maintenance effect prediction model to obtain a plurality of maintenance effect prediction results; The maintenance plan determination module is used to evaluate the maintenance effect prediction result to obtain an evaluation score for each of the alternative maintenance plans, and take the alternative maintenance plan with the highest evaluation score as the target maintenance plan.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor and executing the method according to any one of claims 1 to 7.

10. An electronic device, characterized in that: It includes a processor, a memory and a transceiver, the memory is used to store instructions, the transceiver is used to communicate with other devices, and the processor is used to execute the instructions stored in the memory so that the electronic device executes the method according to any one of claims 1 to 7.

Images