A method for recommending intelligent follow-up services for mild cognitive impairment in the elderly in the community

Abstract

The present invention provides a method for recommending intelligent follow-up services for mild cognitive impairment in the elderly in a community. A prediction model based on feature evolution selection is proposed based on the collected elderly follow-up information. A recommendation model for elderly care services based on similarity calculation is set based on the prediction results of cognitive impairment; a questionnaire for community elderly cognition and chronic disease management is established by using a scale; the scale questionnaire is digitized; a web system is constructed to publish an electronic version of the questionnaire; scale query is supported, scale data is obtained, and the data is stored in a database; a prediction model based on feature evolution selection is used to predict the degree of Alzheimer's disease in the elderly; and a recommendation model for elderly care services based on similarity calculation is used to generate medical elderly care service recommendation results. Compared with the existing methods, the present invention can provide a more comprehensive MCI follow-up scale, more accurate prediction results, a more user-friendly follow-up system, and more personalized elderly service recommendations.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence technology applied to mild cognitive impairment prevention and control systems. Specifically, the present invention relates to the construction of an information service system for cognitive impairment prevention and control and the construction of a cognitive impairment prediction and service recommendation model based on an artificial intelligence algorithm, which is used to realize the networking of community surveys on mild cognitive impairment in the elderly, the electronicization of elderly cognitive impairment data, and the intelligent follow-up and service recommendation of elderly cognitive impairment. Background Art

[0002] As the problem of aging in my country gradually emerges, and the current community survey on the prevention and control of cognitive impairment in the elderly is basically a purely manual paper operation, which requires a lot of manpower and material resources, the electronic questionnaire survey is convenient for medical staff to operate and is also conducive to data entry and analysis. On the basis of the informatization of elderly follow-up data, the entered geriatric disease community survey data can be used in combination with the model adopted by the present invention to predict whether the elderly suffer from MCI and recommend corresponding online and offline care and nursing services.

[0003] For early screening of MCI, existing technologies often use machine learning models such as random forests to predict disease progression and trends based on features. However, existing artificial intelligence methods need to select features to achieve global optimality, and often use global search methods such as grid search. They cannot effectively screen the massive features in the field of cognitive impairment prevention and control, and need to design a more efficient feature search and selection method to improve algorithm training efficiency.

[0004] At present, there is no MCI service system with low information and intelligence levels, making it difficult to achieve personalized and intelligent cognitive impairment service recommendations. Collaborative filtering recommendation is currently the most widely used and successful recommendation system. The starting point of collaborative filtering is that users with similar interests may be interested in the same things. Collaborative filtering technology generates a user set that is most similar to the current user's behavioral interests by analyzing historical data, and uses the items they are most interested in as the recommendation results for the current user; different methods are used to calculate user preferences for items, and various calculation results are integrated to improve the reliability of the results. It is difficult for general-purpose service recommendation algorithms to combine information such as the physiological and sociological characteristics of the elderly, medical history, and peripheral blood biomarkers to make service recommendations. A recommendation system for elderly care, care services, and cognitive intervention that can combine disease screening information is needed to achieve accurate service recommendations and improve service docking efficiency and capabilities. Summary of the invention

[0005] In view of the above problems, the purpose of the present invention is to find an intelligent solution for community elderly MCI prevention and control follow-up and elderly care services. When conducting community elderly MCI surveys, it is as convenient as possible for the elderly and medical staff to operate. After obtaining data, it is convenient for patients to query, and for medical staff to analyze and recommend cognitive intervention services.

[0006] To achieve the above objectives, the present invention firstly designs a set of community elderly MCI follow-up information and elderly care service management schemes, then proposes a prediction model based on feature evolution selection based on the collected elderly follow-up information, and finally designs an elderly care service recommendation model based on similarity calculation based on the cognitive impairment prediction results.

[0007] The present invention provides a method for recommending intelligent follow-up services for mild cognitive impairment in the elderly in the community. A prediction model based on feature evolution selection is proposed based on the collected follow-up information of the elderly. A recommendation model for elderly care services based on similarity calculation is set based on the cognitive impairment prediction results. The collection of follow-up information of MCI in the elderly in the community and recommendation of elderly care services are realized. The implementation process includes the following steps:

[0008] Step 1.1, using the scale method, establish a community elderly cognition and chronic disease management questionnaire;

[0009] Step 1.2, digitize the questionnaire;

[0010] Step 1.3, build a web system and publish the electronic version of the questionnaire;

[0011] Step 1.4, supporting scale query, including obtaining scale data based on scale number, respondent name, respondent ID number, researcher name and / or follow-up time;

[0012] Step 1.5, storing the data obtained in step 1.4 in a database;

[0013] Step 1.6, based on the data obtained in step 1.4, the prediction model based on feature evolution selection is used to predict the severity of Alzheimer's disease in the elderly;

[0014] Step 1.7, based on the data obtained in step 1.4 and the prediction results in step 1.6, the elderly care service recommendation model based on similarity calculation is used to generate medical elderly care service recommendation results.

[0015] Moreover, the scale includes archival information, general data, physical examination, socioeconomic status, social network, handedness, personality, past medical history, family history, personal life and behavioral history, subjective cognitive decline questionnaire and some scale scores related to MCI determination, as well as peripheral blood biomarkers.

[0016] Moreover, when querying the scale, the scale data were obtained based on the scale number, respondent name, respondent ID number, researcher name and follow-up time.

[0017] Moreover, cognitive impairment prediction based on scale data is performed, including using a random forest algorithm based on feature evolution to process scale data for auxiliary prediction of MCI.

[0018] Moreover, service recommendations are made by combining collaborative filtering algorithms with popular item recommendations.

[0019] Moreover, after step 1.7, you can select the type of elderly care service and the service time and location information and place an order, and the backend will enter and save the order information.

[0020] Moreover, it is implemented on the web and mobile sides to adapt to a variety of application scenarios.

[0021] The advantages of the present invention are:

[0022] In terms of follow-up information of MCI in the community and intelligent elderly care services, the present invention combines a complete set of questionnaires to implement and predict MCI in the elderly and recommend medical elderly care services. It simplifies the steps of medical staff questionnaire surveys, makes it easier for patients to fill in and query data, and makes it easier for medical staff to organize and analyze data; it makes useful explorations into community elderly care issues and elderly care service recommendations.

[0023] In the prediction of cognitive impairment in the elderly, the present invention adopts a genetic algorithm to obtain the optimal parameters of the random forest MCI prediction model. After setting the random forest parameter range according to the MCI sample characteristics, the chromosome encoding rules are set according to the parameter range to generate a parameter group population. After the population evolves through generations, it finally converges to the optimal parameter solution. Compared with grid search, genetic algorithm is a random search and optimization method based on biological natural selection and genetic mechanisms, with fast random global feature search capabilities. The method proposed in the present invention has a high intelligent auxiliary value, which not only saves precious medical costs and reduces the workload of doctors, but also significantly improves the accuracy compared with the current MCI prediction method.

[0024] In terms of elderly care cognitive intervention recommendations, for different levels of MCI prediction results, the present invention adopts a collaborative filtering method to calculate the similarity between users in the same MCI result population, and recommends appropriate medical and elderly care services to the elderly in combination with the item popularity index. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A flowchart of community elderly health management and follow-up services according to an embodiment of the present invention;

[0026] Figure 2 This is a flow chart of the MCI prediction method based on feature evolution selection random forest according to an embodiment of the present invention.

[0027] Figure 3 This is a flowchart of collaborative filtering service recommendation based on MCI prediction results and user similarity in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0029] The intelligent method for recommending follow-up services for elderly people with MCI in the community provided by the embodiment of the present invention mainly includes 1) design and development of follow-up information and service management system, 2) development of MCI prediction random forest model based on feature evolution, and 3) development of service recommendation model combining collaborative filtering and project popularity. That is, a prediction model based on feature evolution selection is proposed based on the collected follow-up information for the elderly, and a recommendation model for elderly care services based on similarity calculation is set based on the prediction results of cognitive impairment, so as to realize the collection of follow-up information for elderly people with MCI in the community and recommendation of elderly care services.

[0030] The main contents include:

[0031] 1) Define a MCI screening system for the elderly in the community, and effectively judge the cognitive function and possible pathological changes of the respondents through multiple information such as the respondents' personal physical condition and lifestyle, family illness, neuropsychological assessment scale assessment, and peripheral blood new marker detection; 2) Make the paper survey scale electronic, and propose the use of random forest method based on feature evolution selection to automatically diagnose MCI while facilitating data management, and investigate and analyze the MCI symptoms of the elderly in the community; 3) Provide a convenient and easy-to-operate MCI follow-up interactive method, which respondents and researchers can operate directly through this system. Compared with most existing methods, it is more convenient for respondents to fill in and query information; 4) Recommend cognitive intervention services based on the results of MCI diagnosis, group the population according to the MCI results, calculate the similarity between users in the same group, and recommend services based on the similarity ranking. Recommending among the same group of people improves the accuracy of the recommendation system; in addition, when calculating the similarity between users, in order to eliminate the long-tail effect of the recommendation system and avoid the more popular items being more easily recommended and the less popular items being less likely to be recommended, the concept of popularity, that is, the popularity of the item, is introduced to reduce the impact of the popularity of the item on the user similarity, so as to achieve a better recommendation effect. In general, compared with the existing methods and systems, the present invention can provide a more comprehensive MCI follow-up scale, more accurate prediction results, a more user-friendly follow-up system, and more personalized elderly service recommendations.

[0032] The specific implementation of each part of the embodiment is described below.

[0033] 1. Aiming at the needs of community elderly health management and follow-up information services, the embodiment of the present invention designs a community MCI intelligent follow-up and service recommendation method.

[0034] The proposed implementation process is as follows: 1) define a complete set of community survey questionnaires on geriatric diseases; 2) digitize the community survey questionnaires on geriatric diseases; 3) analyze and manage the questionnaire data; 4) export the questionnaire data to an xls file; 5) predict the degree of Alzheimer's disease in the elderly; 6) automatically recommend medical and elderly care services based on the prediction results.

[0035] The main implementation in the embodiment includes the following steps: Figure 1 As shown:

[0036] Step 1.1: Use the scale method to establish a complete set of community elderly cognition and chronic disease management questionnaires;

[0037] Preferably, the scale includes archival information, general information, physical examination, socioeconomic status, social network, handedness, personality, past history, family history, personal life and behavioral history, subjective cognitive decline questionnaire and some scale scores related to MCI determination, as well as peripheral blood biomarkers, each of which contains its own detailed content information.

[0038] In the embodiment, this step mainly designs the following scale:

[0039] 1) Use evidence-based medicine analysis and Delphi expert consultation method to study the evaluation index system of elderly cognitive impairment. According to the five steps of evidence-based medicine methodology, guided by the health needs of community elderly people and the prevention and control goals of cognitive impairment, for traditional elderly cohort field surveys, community elderly physical examinations, medical diagnosis and treatment, and Internet of Things automatic monitoring and other multi-dimensional data collection forms and application scenarios, evidence-based evaluation of relevant literature and evidence, to build a comprehensive evaluation index system for cognitive impairment. Using the Delphi expert consultation method, from the perspectives of data source, data quality and data scale, establish data acceptance rules to form a community elderly cognitive impairment evaluation index system and digital list;

[0040] 2) Use evidence-based medicine analysis and hierarchical analysis method to study the comprehensive health management quality evaluation indicators and health management grading standards for cognitive impairment. On the one hand, use the elderly memory cohort to quickly apply evidence-based evidence. On the other hand, sort out the indicator evaluation standards for health management, complication management, physical therapy rehabilitation, life assistance, emergency warning, education and training, humanistic care and life care of the elderly cognitive impairment in domestic and foreign literature. Use hierarchical analysis method to refine and integrate grading and classification schemes with high evidence level and feasibility, and form a multi-module cognitive impairment risk quantitative grading standard, intervention service list and comprehensive health management quality standard;

[0041] 3) Using the Delphi expert consultation method and normalization method, we studied the data governance standards of the cognitive impairment prevention and control system and built a multimodal large database and data management system standard. We integrated heterogeneous data from different sources to establish a cognitive impairment prevention and control database; then we parsed the text data, structured the data, systematically clarified the semantics of the fields, and standardized the data structure; finally, we normalized the non-standard terms to form cleaned data and compiled a standard data set for cognitive impairment risk assessment.

[0042] 4) Archive main information scale, by entering the main information of the interviewee and the researcher to create the interviewee's personal scale, which is convenient for the subsequent input and storage of all scale contents and also facilitates the scale query operation. The main field table in the database is shown in the following table:

[0043] Table 1: Archive information

[0044]

[0045] 5) The main personal information of the interviewer is conducive to the subsequent scale classification and research analysis. The main field table in the database is shown in the following table:

[0046] Table 1 General information

[0047] Field Name Field Type Field Description a1 varchar gender a2 int age a3 int Education level (years) a4 date Date of Birth

[0048] 6) B Physical Examination Scale, records the respondent's personal physical health and saves it in the database. The main fields in the database are shown in the following table:

[0049] Table 2 Physical examination

[0050]

[0051]

[0052] 7) C Socioeconomic Status Scale, which records the economic status of the respondents. In subsequent studies, it can be used to determine the impact of economic status on the elderly in the community. The main fields in the database are shown in the following table:

[0053] Table 3 Socioeconomic status

[0054]

[0055] 8) D Social Network Scale, which records the relationship between the respondents and others, is helpful for studying the mental state of the elderly and whether there is a phenomenon that people become more lonely as they age. The main fields in the database are shown in the following table:

[0056] Table 4 Social networks

[0057] Field Name Field Type Field Description d1 varchar Who have you lived with in the past year? d2 varchar Do you have children? (Number of children) d3 varchar Do you have any living brothers and sisters? d4 varchar How many close friends do you have who you can turn to for support and help? d5 varchar Your relationship with your neighbors d6 varchar The number of social events you attend

[0058] 9) E Handedness Scale and F Personality Scale record the respondents’ dominant hand and their main personality traits, which can predict the differences in their cognitive functions. The main fields in the database are shown in the following table:

[0059] Table 5 Handedness + Personality

[0060] Field Name Field Type Field Description e varchar Handedness (right-handed, left-handed, ambidextrous) f varchar Personality traits (introversion or extroversion, social adaptability)

[0061] 10) G Past medical history scale, which records the medical history of the respondent, can clearly understand the respondent's previous and current health status, and facilitate subsequent analysis of whether there are any residual symptoms that affect their personal situation. The main field tables in the database are shown in the following table:

[0062] Table 6 Past medical history

[0063]

[0064] 11) H Family History Scale, which records the medical history of the respondent's family members, can help understand whether the mild cognitive impairment symptoms of the elderly in the community are related to family inheritance. The main fields in the database are shown in the following table:

[0065] Table 7 Family history

[0066] Field Name Field Type Field Description h_1_1 varchar Has the father ever been ill? h_1_2 varchar If the father is ill, select the type of illness h_2_1 varchar Has the mother ever been ill? h_2_2 varchar If the mother is ill, select the type of illness h_3_1 varchar Have your siblings ever been ill? h_3_2 varchar If siblings have illness, select the type of illness

[0067] 12) I Personal Life and Behavior History Scale, which records the respondent's own behavior and lifestyle, from which we can analyze whether there are chronic non-communicable diseases, which is helpful for subsequent MCI research. The main fields in the database are shown in the following table:

[0068] Table 8 Personal life and behavior history

[0069]

[0070]

[0071] 13) Subjective cognitive decline questionnaire scale. Subjective cognitive decline is a high-risk factor for Alzheimer's disease. This part of the scale records the information of the respondents. Through analysis and comparison, the memory status of the elderly population in the community can be well obtained. The main fields in the database are shown in the following table:

[0072] Table 9 Questionnaire on Subjective Cognitive Decline

[0073]

[0074]

[0075] If j2_x (x is 1 / 2 / 3 / 4 / 5) is selected as yes, the following fields will be displayed:

[0076] Field Name Field Type Field Description j2_x_1 varchar Are you worried j2_x_2 varchar Do you feel that you perform worse than your peers in this area? j2_x_3 varchar When do you think things got worse? j2_x_4 varchar Have you been to the doctor? j2_x_5 varchar When did you first talk to your doctor about these issues?

[0077] 14) H The scores of each scale, record the scores of the respondents on some medical MCI scales, which is convenient for subsequent prediction and judgment of medical MCI, and is also helpful to know the prediction results of the respondents' mild cognitive impairment. The main fields in the database are shown in the following table:

[0078] Table 10 Scores of each scale

[0079]

[0080] 15) Added proteomics detection, serum inflammatory factors, plasma Aβ1-42, Aβ1-40, Aβ1-42 / Aβ1-40, T-tau, P-tau181, neurofilament protein light chain (NfL)

[0081] 16) Add genetic testing and ApoE allele detection indicators

[0082] Step 1.2: Electronicize the questionnaire in 1.1; this step mainly implements the front-end design of the questionnaire based on the Vue framework, uses a paged and step-by-step input method, and adopts a design method suitable for operation on the Android and desktop terminals; Step 1.3: Build a web system and publish the electronic version of the questionnaire in 1.2 in the system; the system implementation of this step is based on the SSM framework. The Web system consists of three parts: the background service program, the desktop front-end system, and the Android front-end system. The background service program implements background functions such as data management, chronic disease analysis, and service docking. The desktop front-end mainly provides data entry, query, import and export, and service ordering functions. The Android front-end mainly provides data entry and service ordering functions;

[0083] Step 1.4: Support scale query, and obtain scale data according to scale number, respondent name, respondent ID number, researcher name and / or follow-up time; in specific implementation, community survey subjects use Android or desktop front-end interface to complete the questionnaire under the guidance of medical staff;

[0084] Step 1.5: The system stores the data obtained in 1.4 in the MySql database;

[0085] Step 1.6: Based on the data obtained in 1.4, use the model in content 2 below to predict the severity of Alzheimer's disease in the elderly; the data obtained can be used as a reference for MCI determination, and the use of the random forest algorithm based on feature evolution to process these high-dimensional data can efficiently and accurately assist in MCI prediction.

[0086] Step 1.7: Based on the data obtained in 1.4 and the prediction results in 1.6, use the model in content 3 to recommend medical and elderly care services to the elderly; the recommendation method can use a combination of collaborative filtering algorithm and popular item recommendation, which is conducive to improving the accuracy of the recommendation system;

[0087] Step 1.8: Support the elderly’s family members to select the type of elderly care services, service time and location, and other information on the MCI prevention and control service interface and place an order. The backend will enter and save the order information.

[0088] During the specific implementation, it is recommended that the system be divided into web and mobile terminals, which can adapt to a variety of application scenarios, making it more convenient to conduct intelligent follow-up of community cognitive impairment prevention and control, and providing convenience for respondents and researchers.

[0089] 2. The present invention designs an early risk screening model for elderly MCI based on feature evolution selection random forest, and the realization of its main functions includes the following steps, such as Figure 2 As shown:

[0090] Step 2.1: Set the random forest parameter range according to the MCI feature. Since the number of decision trees in the random forest generally tends to be stable after 100, the number of decision trees is set to [1,121]. The number of MCI features in this system is 22. To prevent the decision tree from overfitting, the maximum depth of the decision tree should be slightly smaller than the number of MCI features, so it is set to [1,16]. The minimum number of leaf features is set to [1,8].

[0091] Step 2.2: Set the chromosome encoding rule, and combine a set of parameters of the random forest MCI prediction model as a chromosome. A chromosome is also called an individual. A chromosome contains multiple genes. A gene is represented by a binary number. A parameter is represented by one or more genes. The number of genes is determined by the range of the parameter and the correspondence between the binary encoding and the parameter's real value. The more genes, the greater the accuracy. According to the parameter value range obtained in step 2.1, the number of genes is set to 14, and the number of decision trees, the maximum depth of the decision tree, the maximum number of features, and the minimum number of leaf features can correspond to 4, 4, 3, and 3 genes, respectively;

[0092] Step 2.3: Initialize the population with the parameter group, and use the method in step 2.2 to generate a series of chromosomes to represent the parameter group of the random forest MCI prediction model, forming a population, and then proceeding with the selection, crossover and mutation process. The initialization population is usually generated randomly;

[0093] Step 2.4: Calculate the fitness of the chromosome through the MCI prediction accuracy. The fitness of the chromosome corresponds to the quality of the current individual and is generally determined by the objective function. The individual fitness of this method is calculated based on the accuracy of the parameter group corresponding to the individual in the random forest MCI prediction model;

[0094] Step 2.5: Select individuals from the population. According to the natural selection law of survival of the fittest, select individuals according to the fitness of chromosomes, that is, screen the parameter combination of the random forest MCI prediction model. Individuals with high fitness are more likely to be retained, but it does not mean that individuals with low fitness will definitely disappear. Some individuals may have excellent genes, which serve as the basis for subsequent genetic development, although their fitness is not high.

[0095] Step 2.6: Chromosome gene crossover: According to the genetic laws of nature, individuals will cross gene fragments in pairs during the genetic process. The process of exchanging gene fragments will produce new possible solution combinations, that is, it is possible to produce new random forest MCI prediction model parameter combinations;

[0096] Step 2.7: Gene mutation: In the process of heredity, a small probability of gene mutation may occur in the population, which increases the diversity of the population and creates raw materials for heredity. This process may also produce a more favorable parameter group for the random forest MCI prediction model.

[0097] Step 2.8: Determine the termination condition, loop the process from step 2.3 to step 2.6, and terminate the loop when the number of loops reaches the predetermined number;

[0098] Step 2.9: Use the optimal parameters obtained by the feature selection method to train the random forest to predict MCI, and you can get the random forest automatic prediction MCI model.

[0099] 3. The present invention designs a collaborative filtering service recommendation method based on MCI prediction results and user similarity, including the following sub-steps: Figure 3 As shown:

[0100] Step 3.1: Classify users into different groups according to the MCI results predicted in step 1.6, so as to search for similar users in a smaller range, improve the efficiency of the recommendation system, and make the recommendation results more accurate;

[0101] Step 3.2: Calculate the similarity between users in the same group based on the cognitive intervention service selected by the user; each piece of data contains three attributes: user_id (user number), item_id (item number), and mci_result (MCI result). First, divide the data into different groups according to the different mci_result values, calculate the similarity between users, and obtain the top 5 similar users of the target user, that is, the top 5 users who are most similar to the items selected by the target user. Since in the traditional cos similarity calculation method, the number of times an item is selected is approximately exponentially related to the size of the similarity, the influencing factor of the number of times an item is selected is amplified, resulting in the more popular items being more likely to be recommended, and the less popular items being less likely to be recommended. In order to eliminate this drawback, define the popularity of each item: ppl i =ln(1+|N i |), N i Represents the number of times each item is selected, so the user similarity calculation formula is:

[0102]

[0103] U1 and U2 represent the sets of items selected by user 1 and user 2, which reduces the impact of the number of times an item is selected on the recommendation results, making the recommendation results more reasonable.

[0104] Step 3.3: Take the top five users with the highest similarity to the target user, and take the five items that these users have selected most times and that the target user has not selected as the recommendation results;

[0105] Step 3.4: If the target user is a new user, the five most popular items in the user group that the target user has not selected are used as recommendation results.

[0106] In specific implementation, the method proposed in the technical solution of the present invention can be implemented by technical personnel in this field using computer software technology to realize automatic operation process. System devices for implementing the method, such as computer-readable storage media storing the corresponding computer program of the technical solution of the present invention and computer equipment running the corresponding computer program, should also be within the scope of protection of the present invention.

[0107] For example, a community MCI intelligent follow-up and service recommendation system can be implemented accordingly, and the user use process includes the following steps:

[0108] Step 1: Log in;

[0109] Enter the correct account and password to enter the system;

[0110] Step 2: Add the scale;

[0111] Respondents and researchers use the system to create a personal scale for respondents and fill in the main content information of the scale according to the respondents' own situation. After filling in, submit it and the data will be saved in the background database;

[0112] Step 3: Scale maintenance;

[0113] Respondents and researchers can view and delete existing scales on the main interface. Under the viewing operation, the scale content can be modified and saved;

[0114] Step 4: Query the scale;

[0115] Respondents and researchers input the information they want to query on the front end, and the back-end database returns the corresponding scale data based on the information obtained;

[0116] Step 5: Scale export;

[0117] Through the export operation, the researchers summarized and exported all the scale information stored in the database in the form of an xls file. The researchers can analyze the obtained data and conduct more effective research on the prediction of medical MCI symptoms.

[0118] Step 6: Prediction of cognitive impairment and service recommendation;

[0119] After entering the questionnaire data, the system will give the predicted results of the elderly's MCI and recommend appropriate cognitive impairment intervention services to the elderly based on the results.

[0120] Step 7: Place service order;

[0121] The elderly and their families select the type of cognitive impairment intervention, service time, service location and other information before placing an order, and the service provider will provide services based on the order content.

[0122] The specific embodiments described herein are merely examples of the spirit of the present invention. Those skilled in the art may make various modifications or additions to the specific embodiments described or replace them in similar ways, but they will not deviate from the spirit of the present invention or exceed the scope defined by the appended claims.

Claims

1. A method for recommending intelligent follow-up services for mild cognitive impairment in the elderly in the community, characterized by: According to the collected elderly follow-up information, a prediction model based on feature evolution selection is proposed, and a pension service recommendation model based on similarity calculation is set based on the cognitive impairment prediction results to realize the collection of community elderly MCI follow-up information and pension service recommendation. The implementation process includes the following steps: Step 1.1, using the scale method, establish a community elderly cognition and chronic disease management questionnaire; Step 1.2, digitize the questionnaire; Step 1.3, build a web system and publish the electronic version of the questionnaire; Step 1.4, supporting scale query, including obtaining scale data based on scale number, respondent name, respondent ID number, researcher name and / or follow-up time; Step 1.5, storing the data obtained in step 1.4 in a database; Step 1.6, based on the data obtained in step 1.4, the prediction model based on feature evolution selection is used to predict the severity of Alzheimer's disease in the elderly, and the predicted MCI result is obtained; Step 1.7, according to the data obtained in step 1.4 and the MCI result predicted in step 1.6, the elderly care service recommendation model based on similarity calculation is used to generate medical elderly care service recommendation results; The prediction model based on feature evolution selection adopts an early risk screening model for elderly MCI based on feature evolution selection random forest, and the model execution includes the following steps: Step 2.1, set the random forest parameter range according to the MCI features. The maximum depth of the decision tree should be slightly smaller than the number of MCI features. Step 2.2, set the chromosome encoding rule and combine a set of parameters of the random forest MCI prediction model as a chromosome; Step 2.3, initialize the population with the parameter group, generate a series of chromosomes representing the parameter group of the random forest MCI prediction model, form a population, and carry out the following selection, crossover and mutation process; Step 2.4, calculate the fitness of the chromosome through the MCI prediction accuracy. The fitness of the chromosome corresponds to the quality of the current individual. The fitness is calculated based on the accuracy of the parameter group corresponding to the individual in the random forest MCI prediction model. Step 2.5, select individuals from the population, select individuals according to the fitness of the chromosome, and screen the parameter combination of the random forest MCI prediction model; Step 2.6, chromosome gene crossover, is used to generate new possible solution combinations, i.e., new random forest MCI prediction model parameter combinations; Step 2.7, gene variation, used to generate a more favorable parameter set for the random forest MCI prediction model; Step 2.8, determine the termination condition, loop the process from step 2.3 to step 2.6, and terminate the loop when the number of loops reaches the predetermined number; Step 2.9, using the optimal parameters obtained by the feature selection method to train the random forest to predict MCI, and obtain the random forest automatic prediction MCI model; The generation of medical and elderly care service recommendation results adopts a collaborative filtering service recommendation method based on MCI prediction results and user similarity, and the implementation process includes the following sub-steps: Step 3.1, classify users into different groups according to the MCI results predicted in step 1.6; Step 3.2, based on the cognitive intervention service selected by the user, calculate the similarity between users in the same group; each data contains three attributes: user ID user_id, item ID item_id, and MCI result mci_result. First, divide the data into different groups according to the different mci_result values, calculate the similarity between users, and obtain several similar users before the target user. Define the popularity of each item: ppl i =ln(1+|N i |), N i Represents the number of times each item is selected. The formula for calculating user similarity is: U1 and U2 represent the sets of items selected by user 1 and user 2; Step 3.3: Take the top several users with the highest similarity to the target user and select the items that are most frequently selected by these users and have not been selected by the target user as the recommendation results; Step 3.4: If the target user is a new user, the most popular items in the user group that the target user has not selected are selected as recommendation results.

2. The method for recommending intelligent follow-up services for mild cognitive impairment in the elderly in the community according to claim 1, characterized in that: The scale includes archival information, general data, physical examination, socioeconomic status, social network, handedness, personality, past medical history, family history, personal life and behavioral history, subjective cognitive decline questionnaire and some scale scores related to MCI determination, as well as peripheral blood biomarkers.

3. The method for recommending intelligent follow-up services for mild cognitive impairment in the elderly in the community according to claim 1, characterized in that: When querying the scale, the scale data were obtained based on the scale number, respondent name, respondent ID number, researcher name and follow-up time.

4. The method for recommending intelligent follow-up services for mild cognitive impairment in the elderly in the community according to claim 1, characterized in that: Predict cognitive impairment based on scale data, including using a random forest algorithm based on feature evolution to process scale data for auxiliary prediction of MCI.

5. The method for recommending intelligent follow-up services for mild cognitive impairment in the elderly in the community according to claim 1, characterized in that: Service recommendations are made by combining collaborative filtering algorithms with popular item recommendations.

6. The method for recommending intelligent follow-up services for mild cognitive impairment in the elderly in the community according to claim 1, characterized in that: After step 1.7, you can select the type of elderly care service and the service time and location information and place an order. The backend will enter and save the order information.

7. The method for recommending intelligent follow-up services for mild cognitive impairment in the elderly in the community according to claim 1, 2, 3, 4, 5 or 6, characterized in that: It is divided into web and mobile implementations to adapt to various application scenarios.

Images