Chronic disease information management system based on behavior interaction model

Abstract

The invention belongs to the technical field of intelligent medical treatment, and discloses a chronic disease information management system based on a behavior interaction model. Comprising the following steps: acquiring background information, internal motivation and cognitive evaluation; based on the background information, a health education content library is constructed, and health education content is pushed in combination with cognitive evaluation; fusing the background information and the internal motivation to generate a multi-dimensional treatment scheme matrix; acquiring willingness information of a patient, performing quantitative evaluation on acceptance degrees of different treatment schemes, and screening out an optimal treatment scheme; acquiring and analyzing real-time interaction data, and dynamically formulating an emotion support strategy; patient health data are integrated, and the interaction effect is evaluated; according to the interaction effect, the health education content, the optimal treatment scheme and the emotion support strategy are intelligently optimized in sequence; behavior interaction is taken as the core, and accurate management of the whole life cycle of the chronic disease patient is realized, so that the health management effect of the patient is remarkably improved, and the complication risk is reduced.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical technology, and more specifically, to a chronic disease information management system based on a behavior interaction model. Background Art

[0002] With the acceleration of the aging population process and the transformation of lifestyle, chronic diseases have gradually become the main component of the global disease burden, bringing huge challenges to the medical and health system; among them, chronic diseases such as systemic lupus erythematosus and sarcopenia have become key and difficult problems in medical management due to their complexity, long-term nature, and multi-system involvement characteristics; the traditional chronic disease management model mainly relies on hospital outpatient follow-up, which has obvious limitations: the short contact time between doctors and patients, long follow-up intervals, poor patient compliance, the formality of health education, and the inability to achieve continuous monitoring and other problems are prominent; this fragmented and passive medical service model can no longer meet the long-term, continuous, and personalized health management needs of chronic disease patients.

[0003] In recent years, the rapid development of intelligent medical technology has provided new ideas for solving the above problems; the wide application of information technologies such as mobile Internet, big data, and cloud computing has made it possible for medical services to break through time and space limitations; for example, the patent with the publication number CN111933278A discloses a chronic disease management system based on artificial intelligence; including: inputting the basic information of patients, various examination results, and medical orders; evaluating the chronic disease situation of patients according to the basic information of patients and their corresponding various examination results; formulating a treatment plan for patients according to the evaluation results; formulating a reexamination plan for patients according to the evaluation results; realizing the management of reexamination situation data of patients, and when it is found that a patient who has not completed the reexamination plan within the preset time, starting a short message warning module for warning; this invention can achieve full monitoring and full recording of the chronic disease treatment process, while improving the efficiency of chronic disease management, providing guarantee for the on-demand treatment of chronic diseases.

[0004] However, although the above technology has realized chronic disease information management, it still stays at the level of information collection and passive monitoring, ignores the active participation and interactive feedback mechanism of patients in the whole process of health behavior change, lacks effective guidance for patients' behavior change, resulting in the difficulty of continuous patient health behavior, low treatment compliance, ultimately affecting the control effect of chronic diseases, increasing the risk of complications, and being difficult to fundamentally solve the core problem of long-term management of chronic diseases.

[0005] In view of this, the present invention proposes a chronic disease information management system based on a behavior interaction model to solve the above problems. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art and to achieve the above object, the present invention provides the following technical solutions: A chronic disease information management system based on a behavior interaction model, comprising: A data acquisition module for acquiring patient characteristic data, where the patient characteristic data includes background information, intrinsic motivation, and cognitive evaluation; An information push module for constructing a health education content library based on the background information, and combining with the cognitive evaluation to push health education content using a hierarchical push strategy; A professional support module for integrating the background information and intrinsic motivation, and using a multi-criteria decision analysis method to generate a multi-dimensional treatment plan matrix; A decision control module for acquiring patient willingness information, quantitatively evaluating the acceptance degree of different treatment plans in the multi-dimensional treatment plan matrix, and screening out the best treatment plan from the multi-dimensional treatment plan matrix according to the evaluation results; An emotional support module for acquiring real-time interaction data, analyzing the real-time interaction data using emotional analysis technology, dynamically formulating an emotional support strategy, and providing emotional support through a pre-established multi-level emotional support network; An interaction evaluation module for collecting patient health data and comprehensively evaluating the interaction effect using deep learning technology; A feedback optimization module for intelligently optimizing the health education content, the best treatment plan, and the emotional support strategy in sequence through a closed-loop feedback mechanism according to the interaction effect.

[0007] Further, the background information includes basic conditions, disease history, and living habits; the disease history includes past medical history and chronic disease status; the chronic disease status includes disease name, disease duration, and severity; The method for constructing the health education content library includes: Build a content resource library, which includes health education content corresponding to different chronic diseases; each piece of health education content in the content resource library corresponds to a set of content data, and the content data includes the name of the chronic disease, the cognitive difficulty, and the health field. According to the name of the chronic disease in the background information, select the corresponding health education content from all the health education content in the content resource library and mark it as appropriate content; determine whether health education content has been pushed to the patient; if so, obtain the historical content, and use the clustering algorithm to cluster the historical content and the appropriate content to obtain a content clusters, where a is an integer greater than 1; build a health education content library according to the appropriate content in the content cluster corresponding to the historical content; if not, obtain the historical patients, and use the clustering algorithm to cluster the historical patients and the current patient according to the patient characteristic data of the historical patients and the current patient to obtain b patient clusters, where b is an integer greater than 1; build a health education content library according to the appropriate content corresponding to the historical patients in the patient cluster corresponding to the current patient; where the historical content is the health education content pushed to the patient at a historical moment, the historical patient is the patient who received chronic disease management at a historical moment, and the current patient is the patient who is currently undergoing chronic disease management; The method for pushing health education content includes: Take the health education content with the same corresponding health field in the health education content library as a set of content; compare the cognitive difficulty corresponding to each piece of health education content in each set of content with the cognitive evaluation, and mark the health education content with the same cognitive difficulty and cognitive evaluation as the pushed content, and push the health education content to the current patient according to the pushed content.

[0008] Further, the method for generating the multi-dimensional treatment plan matrix includes: Obtain the treatment plan corresponding to the historical patients in the patient cluster corresponding to the current patient and mark it as the candidate treatment plan; mark all the data that is not a numerical value in the background information as text data, and use a pre-trained word embedding model to convert each piece of text data into the corresponding text vector; set different digital labels for different intrinsic motivations and mark them as motivation labels; replace all the text data in the background information with the corresponding text vectors, and use the replaced background information and the motivation labels as the analysis data; set different digital labels for different treatment plans and mark them as plan labels; mark the plan labels of the candidate treatment plans as candidate labels, and use each candidate label and the analysis data as a set of evaluation data, and the evaluation data corresponds to the candidate label one by one; input the evaluation data into the trained index evaluation model to predict the corresponding decision index; the decision index includes the treatment effect and the implementation ease, and the index evaluation model is a random forest model; A preset weight set, where the weight set includes the weight set corresponding to each data in the decision-making indicators; multiply each data in the decision-making indicators by the corresponding weight coefficient in the weight set and add them in sequence to obtain a decision value; preset a decision threshold, compare the decision value of each candidate treatment plan with the decision threshold respectively, mark the candidate treatment plans with a decision value greater than or equal to the decision threshold as alternative plans, and do not mark the candidate treatment plans with a decision value less than the decision threshold; generate a multi-dimensional treatment plan matrix according to all alternative plans.

[0009] Further, the method for screening out the best treatment plan includes: Obtain the treatment characteristic data corresponding to each alternative plan, where the treatment characteristic data includes treatment goals and risk levels; use a pre-trained medical NLP model to perform entity recognition on each alternative plan in sequence to identify the treatment form corresponding to each alternative plan; regard the treatment form and treatment goals of each alternative plan as a set of plan characteristic data; calculate the cosine similarity between each data in each set of plan characteristic data and the corresponding data in the corresponding patient willingness information in sequence, and mark it as the acceptance probability; subtract each data in each set of risk levels from the corresponding data in the risk tolerance in sequence to obtain a data difference; the patient willingness information includes treatment goal preferences, treatment form preferences, and risk tolerance. Regard the acceptance probability and data difference corresponding to each set of alternative plans as a set of evaluation characteristic data; construct multiple fuzzy sets for each data in the evaluation characteristic data respectively; convert each set of evaluation characteristic data into the membership degree of the corresponding each fuzzy set through a fuzzification technique respectively; define fuzzy rules; match each fuzzified set of technical characteristic data with the fuzzy rules respectively, and perform fuzzy reasoning using a fuzzy reasoning method to obtain the fuzzy reasoning result corresponding to each alternative plan, where the fuzzy reasoning result is the membership degree of each acceptance level, and the acceptance levels include high acceptance level, medium acceptance level, low acceptance level, and unacceptable; set an acceptance interval, where the range of the acceptance interval is [0, c], and c is an integer greater than 1; divide the acceptance interval evenly into four grade intervals, and the grade intervals correspond one-to-one with the grades in the acceptance levels; add the maximum value and the corresponding minimum value of each grade interval and divide by 2 to obtain the interval mean of each grade interval; multiply each membership degree corresponding to each alternative plan by the corresponding interval mean to obtain an acceptance score; preset a ratio set, where the ratio set includes the ratio coefficients corresponding to each data in the patient willingness information; multiply each acceptance score of each alternative plan by the corresponding ratio coefficient in the ratio set and add them in sequence to obtain the acceptance total value of each alternative plan; add the membership degrees corresponding to each alternative plan in sequence to obtain the total membership degree; divide the acceptance total value of each alternative plan by the corresponding total membership degree respectively as the acceptance degree of each alternative plan, and regard the alternative plan with the highest acceptance degree as the best treatment plan.

[0010] Further, the method for dynamically formulating an emotional support strategy includes: Using a pre-trained emotion recognition model to analyze real-time interaction data to identify the patient's emotional response; setting a strategy set, where the strategy set includes a phone interaction set and an on-site interaction set; randomly selecting an interaction method from both the phone interaction set and the on-site interaction set to construct a group of interaction combinations, a total of d groups of interaction combinations are constructed, and the d groups of interaction combinations are all different. Set sequentially increasing numerical labels for the d groups of interaction combinations and mark them as set labels, and the range of the set labels is [1, d]; randomly select a set label as the initial iteration center and set the number of iterations to 0; Define the iteration process, where the iteration process is: generate m candidate solutions within the range of the set labels and calculate the emotional relief effect corresponding to each candidate solution, 1 < m < d, and the candidate solutions correspond one-to-one with the set labels; mark the candidate solution with the largest emotional relief effect as the temporarily optimal solution and move the iteration center to the temporarily optimal solution; Execute the iteration process, and each time the iteration process is executed, increment the number of iterations by one; preset an iteration threshold, and when the number of iterations is greater than or equal to the iteration threshold, stop executing the iteration process, mark the set label corresponding to the iteration center as the best label, and use the interaction combination corresponding to the best label as the emotional support strategy.

[0011] Further, the method for generating m candidate solutions includes: Preset a selection interval, use the iteration center as the center of the selection interval, and obtain m candidate solutions within the selection interval; The method for calculating the emotional relief effect corresponding to the candidate solution includes: Set different numerical labels for different emotion types and mark them as type labels; set different numerical labels for different emotion intensities and mark them as intensity labels; input the set label, the type label, and the intensity label corresponding to the emotional response into the trained emotion analysis model in sequence to predict the corresponding relief intensity; where the emotion analysis model is a random forest model, and the relief intensity is the emotional intensity of the patient after receiving the phone interaction and on-site interaction in the interaction combination; subtract the relief intensity from the emotional intensity to obtain the emotional relief effect.

[0012] Further, the method for providing emotional support includes: The multi-level emotional support network includes an automatic response layer, an artificial assistance layer, and an artificial intervention layer; a pre-trained word embedding model is used to convert each emotional type into a corresponding type vector; a clustering algorithm is used to cluster all type vectors to obtain three type clusters; the three type clusters respectively correspond to positive emotions, negative emotions, and neutral emotions; the emotional type of the patient is analyzed. If the emotional type belongs to positive emotions, the automatic response layer is used to provide emotional support. If the emotional type is neutral emotions, the artificial assistance layer is used for emotional support. If the emotional type is negative emotions, the artificial intervention layer is used for emotional support.

[0013] Further, the method for comprehensively evaluating the interaction effect includes: The patient's health data and historical health data are sequentially input into the trained effect evaluation model to predict the corresponding interaction effect; among them, the patient's health data are the physiological indicators after the patient receives chronic disease management, the historical health data are the physiological indicators before the patient receives chronic disease management, the effect evaluation model is a deep learning model, and the interaction effect is the improvement effect of chronic diseases after the patient receives chronic disease management; The training process of the effect evaluation model includes: n groups of different health data are collected in advance. Each group of health data includes the patient's health data and historical health data. A corresponding interaction effect is set for the n groups of health data. n is an integer greater than 1. The health data and the corresponding interaction effect are converted into a corresponding set of feature vectors; each set of feature vectors is used as the input of the effect evaluation model. The effect evaluation model takes a set of predicted interaction effects corresponding to each group of health data as the output, and takes the actual interaction effect corresponding to each group of health data as the prediction target. The actual interaction effect is the interaction effect preset corresponding to the health data; minimizing the sum of the prediction errors of all health data is used as the training target; the effect evaluation model is trained until the sum of the prediction errors reaches convergence and then the training stops.

[0014] Further, a preset effect threshold is set, and the interaction effect is compared with the effect threshold; if the interaction effect is greater than or equal to the effect threshold, the health education content, the best treatment plan, and the emotional support strategy are not intelligently optimized; if the interaction effect is less than the effect threshold, the health education content, the best treatment plan, and the emotional support strategy are sequentially and intelligently optimized; The method for intelligently optimizing the health education content includes: Subtract the interaction effect from the effect threshold to obtain the effect difference; set different numerical tags for different cognitive evaluations and mark them as cognitive tags; input the effect difference and cognitive tags into the trained cognitive optimization model to predict the corresponding optimization tags, and obtain the corresponding optimization evaluations according to the optimization tags. The optimization tag is the numerical tag corresponding to the optimization evaluation, and the optimization evaluation is the optimized cognitive evaluation; the cognitive optimization model is a deep neural network model, and the training process of the cognitive optimization model is the same as the training process of the effect evaluation model; in each set of content collections, compare the cognitive difficulty corresponding to each health education content with the optimization evaluation respectively, and mark the health education content with the same cognitive difficulty and optimization evaluation as the optimized content, and push the health education content to the current patient according to the optimized content.

[0015] Further, the method for intelligently optimizing the best treatment plan includes: Preset a contribution set, where the contribution set includes contribution coefficients corresponding to the acceptance degree and decision value; multiply the acceptance degree and decision value of each alternative plan by the corresponding contribution coefficients respectively, and add them up in sequence to obtain the plan excellence degree, and use the alternative plan with the highest plan excellence degree as the best treatment plan; The method for intelligently optimizing the emotional support strategy includes: Input the set label, the type label and intensity label corresponding to the emotional response into the trained threshold prediction model in sequence to predict the corresponding termination threshold; the threshold prediction model is a deep neural network model, and the training process of the threshold prediction model is the same as the training process of the effect evaluation model; in the iterative process, subtract the emotional relief effect corresponding to the previous iteration center from the emotional relief effect corresponding to the current iteration center to obtain the effect change amount; compare the effect change amount with the preset change threshold; if the effect change amount is greater than or equal to the change threshold, generate a change instruction; if the effect change amount is less than the change threshold, do not generate a change instruction; count the number of consecutive change instructions generated and mark it as the instruction quantity; when the instruction quantity is greater than or equal to the termination threshold, stop the iterative process; mark the set label corresponding to the iteration center as the best label, and use the interaction combination corresponding to the best label as the emotional support strategy.

[0016] The technical effects and advantages of the chronic disease information management system based on the behavior interaction model of the present invention: By comprehensively obtaining the patient's background information, internal motivation, and cognitive evaluation, constructing a health education content library and a multi-dimensional treatment plan matrix that meet the patient's personalized needs, and realizing precise guidance and support for the patient's behavior change process; adopting a method combining quantitative evaluation and fuzzy reasoning to comprehensively analyze the acceptance degree of different treatment plans, being able to screen out the best treatment plan that most meets the patient's preferences and needs, and improving treatment compliance; based on the sentiment analysis and adaptive optimization of real-time interaction data, dynamically formulating targeted emotional support strategies to effectively enhance the patient's mental health management; using deep learning technology to construct an effect evaluation model, being able to accurately evaluate the health improvement effect of the patient after receiving chronic disease management; through a closed-loop feedback mechanism, sequentially optimizing the health education content, treatment plan, and emotional support strategy to achieve continuous improvement of the system performance and enhance the overall effect of chronic disease management; with behavior interaction as the core, realizing precise management of the entire life cycle of chronic disease patients, thereby significantly improving the patient's health management effect, reducing the risk of complications, and providing strong support for the long-term control of chronic diseases. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 Schematic diagram of the chronic disease information management system based on the behavior interaction model according to Embodiment 1 of the present invention; Figure 2 Flowchart of the chronic disease information management system based on the behavior interaction model according to Embodiment 1 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0018] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0019] Embodiment 1

[0020] Please refer to Figure 1 and Figure 2 As shown, the chronic disease information management system based on the behavior interaction model in this embodiment includes a data acquisition module, an information push module, a professional support module, a decision control module, an emotional support module, an interaction evaluation module, and a feedback optimization module; each module is connected by wired and / or wireless means to realize data transmission between modules.

[0021] The data acquisition module is used to acquire patient characteristic data, and the patient characteristic data includes background information, internal motivation, and cognitive evaluation.

[0022] The background information includes basic conditions, disease history, and living habits; basic conditions such as the patient's age, gender, education level, occupation, marital status, etc.; disease history includes past medical history and chronic disease status; past medical history refers to the diseases the patient has had before, including the name of the disease, the time of onset, the severity, etc., and the chronic disease status refers to the chronic diseases the patient currently has, including the name of the chronic disease, the diagnosis time, the severity, etc.; living habits such as eating habits, exercise habits, smoking and drinking situations, work and rest routines, etc.; the background information is obtained through the hospital's electronic medical record system or a questionnaire filled out by the patient about personal situations, living habits, etc. Internal motivation refers to the internal driving force for patients to spontaneously carry out health management without external rewards or pressures, such as high, medium, low, etc. For example, Ms. Li, due to concerns about the exacerbation of systemic lupus erythematosus, hopes to reduce the disease burden through health management and actively participates in treatment, so Ms. Li's internal motivation is high; relatively speaking, Mr. Wang feels that the treatment has no obvious effect, lacks interest in health management, and is unwilling to change his living habits, so Mr. Wang's internal motivation is low; internal motivation is obtained through a motivation assessment scale filled out by the patient. Cognitive evaluation refers to the patient's cognitive level in different health fields (such as health status, disease treatment, health behaviors, etc.), such as high, medium, low, etc.; cognitive evaluation directly affects whether the patient believes that health behavior change is feasible and effective, thus determining whether to take action; cognitive evaluation includes the cognitive level of health status (that is, the patient's understanding of their own health status, such as the understanding of the severity and course of current chronic diseases), the cognitive level of disease treatment (that is, the patient's understanding of the treatment methods and medications used), the cognitive level of health behaviors (that is, the patient's understanding of the impact of health behavior changes on health, such as changes in exercise, diet, and sleep), etc. For example, Ms. Li understands the relationship between systemic lupus erythematosus and sarcopenia and knows that exercise can effectively improve sarcopenia, which motivates her to exercise according to the doctor's advice, so Ms. Li's cognitive evaluation is high; relatively speaking, Mr. Wang has an unclear understanding of sarcopenia, believes that exercise will exacerbate systemic lupus erythematosus, and hesitates to change his exercise behavior, so Mr. Wang's cognitive evaluation is low; cognitive evaluation is obtained through relevant questionnaires filled out by the patient or the learning effect during health education (such as the correct rate of answering questions on an online learning platform).

[0023] An information push module, used to build a health education content library based on the background information, and combined with the cognitive evaluation, adopt a hierarchical push strategy to push health education content.

[0024] The methods for building a health education content library include: Build a content resource library, which includes health education content corresponding to different chronic diseases. The health education content in the content resource library is obtained by technicians in this field through international guidelines (such as the American College of Rheumatology guidelines, the Clinical Practice Guidelines of the International Association of Gerontology and Geriatrics, etc.), evidence-based databases (such as UpToDate, PubMed, Cochrane Library, etc.), popular science materials (such as health manuals produced by hospitals, content on authoritative medical platforms, etc.). Each health education content in the content resource library corresponds to a set of content data, and the content data includes the name of the chronic disease, the cognitive difficulty, and the health field. The content data is analyzed and set by technicians in this field according to the health education content. The cognitive difficulty refers to the degree of difficulty for patients to understand and absorb the health education content, such as high, medium, low, etc. According to the name of the chronic disease in the background information, screen out the corresponding health education content from all the health education content in the content resource library and mark it as appropriate content. Judge whether health education content has been pushed to the patient. If so, obtain the historical content, and use clustering algorithms (such as Gaussian mixture model, DBSCAN, hierarchical clustering, etc.) to cluster the historical content and the appropriate content to obtain a content clusters, where a is an integer greater than 1. Construct a health education content library according to the appropriate content in the content cluster corresponding to the historical content. If not, obtain the historical patients, and use clustering algorithms to cluster the historical patients and the current patient according to the patient characteristic data of the historical patients and the current patient to obtain b patient clusters, where b is an integer greater than 1. Construct a health education content library according to the appropriate content corresponding to the historical patients in the patient cluster corresponding to the current patient. Among them, the historical content is the health education content pushed to the patient at a historical moment, the historical patient is the patient who has received chronic disease management at a historical moment, and the current patient is the patient who is currently undergoing chronic disease management. The historical content and patient characteristic data corresponding to the historical patients are all obtained through the patient information management in the hospital.

[0025] The method for pushing health education content includes: Take the health education content with the same corresponding health field in the health education content library as a set of content collections, and the content collections correspond one-to-one with the health fields. Compare the cognitive difficulty corresponding to each health education content in each set of content collections with the cognitive evaluation respectively, mark the health education content with the same cognitive difficulty and cognitive evaluation as the pushed content, and push the health education content to the current patient according to the pushed content. Exemplarily, for patients with systemic lupus erythematosus, push health education content such as sun protection, drug management, immune regulation, regular monitoring, fatigue management, etc. For patients with sarcopenia, push health education content such as protein supplementation, progressive resistance training, functional activities, nutritional balance, etc.

[0026] A professional support module is used to integrate background information and intrinsic motivation, and adopts a multi-criteria decision analysis method to generate a multi-dimensional treatment plan matrix.

[0027] The method for generating a multi-dimensional treatment plan matrix includes: Obtain the treatment plans corresponding to historical patients in the patient cluster corresponding to the current patient, and mark them as candidate treatment plans. The candidate treatment plans are obtained through the patient information management system in the hospital; mark all non-numerical data in the background information as text data, and use a pre-trained word embedding model (such as the FastText model, ELMo model, GloVe model) to convert each text data into a corresponding text vector. The text data includes, for example, gender, education level, name of chronic disease, etc.; set different digital labels for different intrinsic motivations and mark them as motivation labels; replace all text data in the background information with the corresponding text vectors, and use the replaced background information and motivation labels as analysis data; set different digital labels for different treatment plans and mark them as plan labels; mark the plan labels of the candidate treatment plans as candidate labels, and use each candidate label and the analysis data as a group of evaluation data, and the evaluation data corresponds to the candidate label one by one; input the evaluation data into the trained index evaluation model to predict the corresponding decision-making indicators; the decision-making indicators include treatment effect and implementation ease; Among them, the treatment effect is the expected effect of the treatment plan on improving the chronic disease status of the patient, including clinical effects such as symptom relief, improvement of chronic disease indicators, and prevention of complications. The higher the treatment effect, the more effectively the treatment plan can control or improve the chronic disease status of the patient medically; the implementation ease is the ease of implementing the treatment plan in the patient's daily life, which involves the objective convenience of the patient implementing the treatment plan, including factors such as the frequency of medication, operation complexity, economic burden, and time cost; the higher the implementation ease, the easier it is to implement the treatment plan in the patient's daily life, reducing implementation barriers, such as a simplified medication plan, reduced side effects, reduced economic burden, and reduced frequency of hospital visits; the index evaluation model is a random forest model, and the random forest model is an existing technology, and the specific training process will not be elaborated here; A preset weight set, and the weight set includes the weight set corresponding to each data in the decision-making indicators; multiply each data in the decision-making indicators by the corresponding weight coefficient in the weight set and add them in turn to obtain a decision value; preset a decision threshold, compare the decision value of each candidate treatment plan with the decision threshold respectively, mark the candidate treatment plans with a decision value greater than or equal to the decision threshold as alternative plans, and do not mark the candidate treatment plans with a decision value less than the decision threshold; generate a multi-dimensional treatment plan matrix according to all alternative plans; the weight set and the decision threshold are both preset by those skilled in the art according to the actual situation.

[0028] A decision control module is used to obtain patient preference information, quantitatively evaluate the acceptance degree of different treatment plans in a multi-dimensional treatment plan matrix, and screen out the best treatment plan from the multi-dimensional treatment plan matrix according to the evaluation results.

[0029] The method for screening out the best treatment plan includes: Patient preference information includes treatment goal preference, treatment form preference, and risk tolerance; among them, treatment goal preference is the ultimate effect pursued by the patient for the treatment plan or the subjective focus of the treatment result, reflecting individual differences in the patient's treatment expectations, such as prolonging life, relieving symptoms, reducing side effects, etc.; treatment form preference is the patient's preference for the treatment execution method, such as treatment route (such as oral medication or injection), treatment environment (such as home treatment or inpatient treatment), etc.; risk tolerance is the degree to which the patient can accept the risks that may occur during the treatment process (such as side effects, complications, etc.), including severity and probability; patient preference information is obtained through questionnaires filled out by patients, doctor-patient interviews, and other means. Obtain the treatment characteristic data corresponding to each alternative plan. The treatment characteristic data includes treatment goals and risk levels, and is obtained through the electronic medical record system in the hospital; among them, the treatment goal is the expected effect or key point of action that the treatment plan can achieve medically, corresponding to the treatment goal preference in the patient preference information; the risk level is the severity and probability of the risks that the treatment plan may trigger, corresponding to the risk tolerance in the patient preference information; use a pre-trained medical NLP model (such as Med-BERT model, ClinicalBERT model, etc.) to perform entity recognition on each alternative plan in turn to identify the treatment form corresponding to each alternative plan, corresponding to the treatment form preference in the patient preference information; regard the treatment form and treatment goal of each alternative plan as a set of plan characteristic data, and the plan characteristic data corresponds to the alternative plan one by one; calculate the cosine similarity between each data in each set of plan characteristic data and the corresponding data in the corresponding patient preference information in turn, and mark it as the acceptance probability; the calculation method of the cosine similarity is a prior art, and the specific calculation process will not be elaborated here; subtract each data in each set of risk levels from the corresponding data in the risk tolerance in turn to obtain the data difference. Take the acceptance probability corresponding to each group of alternative solutions and the data difference as a set of evaluation characteristic data, with the evaluation characteristic data corresponding one-to-one to the alternative solutions; construct multiple fuzzy sets for each data in the evaluation characteristic data respectively; for example: the fuzzy sets corresponding to the acceptance probability are high acceptance probability, medium acceptance probability, low acceptance probability, etc., and the fuzzy sets corresponding to the data difference are positive difference (i.e., the data difference is greater than 0), zero difference (i.e., the data difference is equal to 0), and negative difference (i.e., the data difference is less than 0); convert each group of evaluation characteristic data into the membership degree of the corresponding fuzzy set respectively through the fuzzification technique; the fuzzification technique is the process of converting accurate numerical values into the membership degree corresponding to the fuzzy set, and the fuzzification technique is, for example, the triangular membership function, the trapezoidal membership function, etc.; for example, if the numerical value of the acceptance probability is relatively high, then it is inferred that the membership degree of high acceptance probability is 0.9, the membership degree of medium acceptance probability is 0.3, and the membership degree of low acceptance probability is 0; Define fuzzy rules, which are defined according to expert knowledge or relevant literature; for example, if in a set of evaluation characteristic data, more than half of the acceptance probabilities are high acceptance probabilities and more than half of the data differences are negative differences, then it is inferred that the membership degree of the acceptance degree of the corresponding alternative solution belonging to the high acceptance degree is high; if in a set of evaluation characteristic data, more than half of the acceptance probabilities are low acceptance probabilities and more than half of the data differences are positive differences, then it is inferred that the membership degree of the acceptance degree of the corresponding alternative solution belonging to the unacceptable is high; match each group of fuzzified technical characteristic data with the fuzzy rules respectively, and use the fuzzy inference method (such as Mamdani fuzzy inference model, Sugeno fuzzy inference model, etc.) to perform fuzzy inference to obtain the fuzzy inference result corresponding to each alternative solution, and the fuzzy inference result is the membership degree of each acceptance degree level, and the acceptance degree levels include high acceptance degree, medium acceptance degree, low acceptance degree, and unacceptable; the fuzzy inference result is, for example, the membership degree of high acceptance degree is 0.6, the membership degree of medium acceptance degree is 0.8, the membership degree of low acceptance degree is 0.1, and the membership degree of unacceptable is 0; Set an acceptance interval, where the range of the acceptance interval is [0, c], c is an integer greater than 1, and in this embodiment, c is preferably 100; evenly divide the acceptance interval into four grade intervals, and the grade intervals correspond one by one to the grades in the acceptance degree grades; add the maximum value of each grade interval to the corresponding minimum value and then divide by 2 to obtain the interval mean of each grade interval; multiply each membership degree corresponding to each alternative by the corresponding interval mean to obtain the acceptance score; preset a proportion set, and the proportion set includes the proportion coefficients corresponding to each data in the patient's will information, and the proportion set is preset by those skilled in the art according to the actual situation of the patient; multiply each acceptance score of each alternative by the corresponding proportion coefficient in the proportion set and add them in turn to obtain the acceptance total value of each alternative; add up each membership degree corresponding to each alternative to obtain the total membership degree; divide the acceptance total value of each alternative by the corresponding total membership degree as the acceptance degree of each alternative, and take the alternative with the highest acceptance degree as the best treatment plan.

[0030] An emotional support module, which is used to obtain real-time interaction data, analyze the real-time interaction data by using emotional analysis technology, dynamically formulate emotional support strategies, and provide emotional support through a pre-established multi-level emotional support network.

[0031] The real-time interaction data is the text-based interaction content generated instantaneously when the patient communicates with the chronic disease management system during the process of receiving chronic disease management; for example, information such as "Will taking this medicine have side effects?" and "What should I do if my blood sugar doesn't drop?" entered by the patient; the real-time interaction data is obtained through the doctor-patient communication platform in the hospital.

[0032] The method for dynamically formulating emotional support strategies includes: Use a pre-trained emotion recognition model to analyze the real-time interaction data and identify the patient's emotional response; the emotional response refers to the patient's emotional experience in the face of the disease and the treatment process, which directly affects the patient's treatment compliance and the effect of behavior change; the emotional response includes the emotion type and the emotion intensity, the emotion type such as anxiety, worry, depression, anger, positive, etc., and the emotion intensity such as mild, moderate, severe, etc.; the emotion recognition model is obtained by fine-tuning based on a pre-trained language model (such as ClinicalBERT model, BioBERT model, RoBERTa model, etc.) for application in specific emotion analysis tasks; the emotion recognition model is an existing technology, and the specific training process will not be elaborated here. A strategy set is set, which includes a telephone interaction set and an on-site interaction set; telephone interaction refers to communication conducted over the phone, which is highly convenient and flexible. The telephone interaction set includes emotional counseling (i.e., medical staff provide emotional counseling to patients with emotional fluctuations over the phone to help them relieve negative emotions), health education (i.e., medical staff provide personalized health education content to patients), psychological counseling (i.e., medical staff provide psychological counseling to patients with more serious emotional problems over the phone to reduce the psychological pressure of patients), etc.; on-site interaction refers to face-to-face communication between patients and medical staff, which has a high emotional support effect. The on-site interaction set includes face-to-face consultation (i.e., medical staff communicate with patients face-to-face and answer patients' questions), support group meetings (i.e., medical staff organize patients to participate in patient group meetings to share treatment experiences with other patients, encourage and comfort each other), social support activities (i.e., medical staff organize leisure activities to enhance patients' social support and help them release emotional pressure, such as group sports, meditation courses, etc.); the strategy set is obtained by technical personnel in this field by referring to clinical practice, standard guidelines, user surveys, etc.; Randomly select an interaction method from the telephone interaction set and the on-site interaction set to construct a set of interaction combinations. A total of d groups of interaction combinations are constructed. The d groups of interaction combinations are all different. Set increasing numerical labels for the d groups of interaction combinations and mark them as set labels. The range of the set label is [1, d]. Randomly select a set label as the initial iteration center and set the number of iterations to 0. Define the iterative process: generate m candidate solutions within the range of the set label, and calculate the emotional relief effect corresponding to each candidate solution, 1<m<d, and the candidate solutions correspond to the set labels one by one; mark the candidate solution with the largest emotional relief effect as the temporary optimal solution, and move the iteration center to the temporary optimal solution; An iterative process is executed, and each time an iterative process is executed, the number of iterations is increased by one; an iteration threshold is preset, and the iteration threshold is pre-set by technical personnel in this field according to actual conditions; when the number of iterations is greater than or equal to the iteration threshold, the iterative process is stopped, the set label corresponding to the iteration center is marked as the best label, and the interaction combination corresponding to the best label is used as the emotional support strategy.

[0033] Methods for generating m candidate solutions include: A preset selection interval is set in advance by a person skilled in the art according to actual conditions; the iteration center is taken as the center of the selection interval, and m candidate solutions in the selection interval are obtained; exemplarily, the selection interval is [-2,2], and the set label corresponding to the iteration center is 5, so 5 is taken as the center of the selection interval, that is, the set labels included in the selection interval are 3, 4, 5, 6, and 7.

[0034] The method for calculating the emotional relief effect corresponding to the candidate solution includes: Set different numerical tags for different emotional types and mark them as type tags; set different numerical tags for different emotional intensities and mark them as intensity tags; input the set tag, the type tag, and the intensity tag corresponding to the emotional response into the trained sentiment analysis model in sequence to predict the corresponding relief intensity; wherein, the sentiment analysis model is a random forest model, and the relief intensity is the emotional intensity of the patient after receiving the phone interaction and on-site interaction in the interaction combination; subtract the relief intensity from the emotional intensity to obtain the emotional relief effect.

[0035] The method for providing emotional support includes: The multi-level emotional support network includes an automatic response layer, an artificial assistance layer, and an artificial intervention layer; use a pre-trained word embedding model to convert each emotional type into a corresponding type vector; use a clustering algorithm to cluster all type vectors to obtain three type clusters; the three type clusters correspond to positive emotions (such as positive, confident, etc.), negative emotions (such as anxiety, worry, etc.), and neutral emotions (such as calm, confused, etc.) respectively; analyze the emotional type of the patient, if the emotional type belongs to positive emotion, use the automatic response layer to provide emotional support, if the emotional type is neutral emotion, use the artificial assistance layer to provide emotional support, if the emotional type is negative emotion, use the artificial intervention layer to provide emotional support; Among them, the automatic response layer uses robots (such as chatbots, virtual health assistants, etc.) to automatically reply to the patient's real-time interaction data by using methods such as preset dialogue templates and natural language generation (NLG) technology; the purpose is to consolidate and continue the patient's positive emotional state, so no complex manual intervention is required; the artificial assistance layer automatically replies to the patient's real-time interaction data by combining the initial response of the robot and manual review; that is, medical staff judge whether modification is needed and then send it to the patient after reviewing the robot's reply; at this time, the patient does not show strong emotions, but may still hide needs or potential risks and requires appropriate attention; the artificial intervention layer is that medical staff reply to the patient's real-time interaction data in real time; at this time, the patient has obvious emotional distress or psychological pressure and requires timely and in-depth manual intervention.

[0036] It should be understood that the dynamically formulated emotional support strategy is a plan for future interaction behaviors and belongs to a delayed behavior, that is, the chronic disease management system will not respond immediately to ensure appropriate support in future interactions; while providing emotional support through the pre-established multi-level emotional support network is a timely feedback on the current interaction and belongs to a real-time behavior, that is, immediately respond and reply to the patient in real time when the patient communicates with the chronic disease management system.

[0037] An interactive evaluation module, which is used to collect patients' health data and comprehensively evaluate the interactive effect by using deep learning technology.

[0038] The methods for comprehensively evaluating the interactive effect include: Input the patients' health data and historical health data into the trained effect evaluation model in sequence to predict the corresponding interactive effect; among them, the patients' health data are the physiological indicators after the patients receive chronic disease management, such as blood pressure, blood sugar, weight, etc.; the historical health data are the physiological indicators before the patients receive chronic disease management; the effect evaluation model is a deep learning model; the interactive effect is the improvement effect of chronic diseases after the patients receive chronic disease management.

[0039] The training process of the effect evaluation model includes: Pre-collect n groups of different health data. Each group of health data includes patients' health data and historical health data, and set corresponding interactive effects for the n groups of health data. n is an integer greater than 1. Convert the health data and the corresponding interactive effects into a corresponding set of feature vectors; the interactive effects corresponding to the health data are collected by those skilled in the art during the process of historical chronic disease management. n groups of different health data are collected, and each group of health data is analyzed in sequence in combination with the actual situation to evaluate the interactive effect corresponding to each group of health data, and corresponding interactive effects are set for the n groups of different health data in sequence; Take each set of feature vectors as the input of the effect evaluation model. The effect evaluation model takes a set of predicted interactive effects corresponding to each group of health data as the output, and takes the actual interactive effect corresponding to each group of health data as the prediction target. The actual interactive effect is the pre-set interactive effect corresponding to the health data; take minimizing the sum of the prediction errors of all health data as the training target; among them, the calculation formula of the prediction error is η w =(θ w -ε w ) 2 , where η w is the prediction error, w is the group number of the feature vector corresponding to the health data, θ w is the predicted interactive effect corresponding to the w-th group of health data, and ε w is the actual interactive effect corresponding to the w-th group of health data; train the effect evaluation model until the sum of the prediction errors reaches convergence and then stop training.

[0040] A feedback optimization module, which is used to intelligently optimize the health education content, the best treatment plan and the emotional support strategy in sequence through a closed-loop feedback mechanism according to the interactive effect.

[0041] A preset effect threshold, which is pre-set by those skilled in the art according to the actual situation; compare the interaction effect with the effect threshold; if the interaction effect is greater than or equal to the effect threshold, do not perform intelligent optimization on the health education content, the best treatment plan, and the emotional support strategy; if the interaction effect is less than the effect threshold, perform intelligent optimization on the health education content, the best treatment plan, and the emotional support strategy in sequence.

[0042] The method for intelligent optimization of health education content includes: Subtract the interaction effect from the effect threshold to obtain an effect difference; set different digital tags for different cognitive evaluations and mark them as cognitive tags; input the effect difference and the cognitive tags into the trained cognitive optimization model to predict the corresponding optimization tags, and obtain the corresponding optimization evaluations according to the optimization tags. The optimization tags are the digital tags corresponding to the optimization evaluations, and the optimization evaluations are the optimized cognitive evaluations; the cognitive optimization model is a deep neural network model, and the training process of the cognitive optimization model is the same as that of the effect evaluation model; in each set of content collections, compare the cognitive difficulty corresponding to each health education content with the optimization evaluation respectively, mark the health education content with the same cognitive difficulty and optimization evaluation as the optimized content, and push the health education content to the current patient according to the optimized content.

[0043] The method for intelligent optimization of the best treatment plan includes: Preset a contribution set, which includes contribution coefficients corresponding to the acceptance degree and the decision value; the contribution set is pre-set by those skilled in the art according to the actual situation; multiply the acceptance degree and the decision value of each alternative plan by the corresponding contribution coefficients respectively, and add them up in sequence to obtain the plan excellence degree, and take the alternative plan with the highest plan excellence degree as the best treatment plan.

[0044] The method for intelligent optimization of the emotional support strategy includes: The set label, the type label corresponding to the emotional response, and the intensity label are sequentially input into the trained threshold prediction model to predict the corresponding termination threshold. The threshold prediction model is a deep neural network model, and the training process of the threshold prediction model is the same as that of the effect evaluation model. During the iteration process, subtract the emotional mitigation effect corresponding to the previous iteration center from the emotional mitigation effect corresponding to the current iteration center to obtain the effect change amount. Compare the effect change amount with a preset change threshold, and the change threshold is preset by those skilled in the art according to the actual situation. If the effect change amount is greater than or equal to the change threshold, a change instruction is generated. If the effect change amount is less than the change threshold, no change instruction is generated. Count the number of consecutive change instructions generated and mark it as the instruction quantity. When the instruction quantity is greater than or equal to the termination threshold, the iteration process is stopped. Mark the set label corresponding to the iteration center as the best label, and use the interaction combination corresponding to the best label as the emotional support strategy.

[0045] In this embodiment, by comprehensively obtaining the patient's background information, internal motivation, and cognitive evaluation, a health education content library and a multi-dimensional treatment plan matrix that meet the patient's personalized needs are constructed to achieve precise guidance and support for the patient's behavior change process; a method combining quantitative evaluation and fuzzy reasoning is adopted to comprehensively analyze the acceptance degree of different treatment plans, and the best treatment plan that most meets the patient's preferences and needs can be screened out to improve treatment compliance; based on the emotional analysis and adaptive optimization of real-time interaction data, targeted emotional support strategies are dynamically formulated to effectively enhance the patient's mental health management; a deep learning technology is used to construct an effect evaluation model, which can accurately evaluate the health improvement effect of the patient after receiving chronic disease management; through a closed-loop feedback mechanism, the health education content, treatment plan, and emotional support strategy are optimized in sequence to continuously improve the system performance and enhance the overall effect of chronic disease management; with behavior interaction as the core, precise management of the entire life cycle of chronic disease patients is achieved, thereby significantly improving the patient's health management effect, reducing the risk of complications, and providing strong support for the long-term control of chronic diseases.

[0046] Embodiment 2

[0047] The present application also provides an electronic device. The electronic device may include one or more processors and one or more memories. Among them, computer-readable code is stored in the memory, and when the computer-readable code is run by one or more processors, it can execute the chronic disease information management system based on the behavior interaction model as described above.

[0048] The method or system according to the embodiments of the present application can also be implemented by means of the architecture of the electronic device shown in the present application. The electronic device may include a bus, one or more CPUs, a ROM, a RAM, a communication port connected to a network, an input / output, a hard disk, etc. The storage device in the electronic device, such as the ROM or the hard disk, can store the chronic disease information management system based on the behavior interaction model provided by the present application. Further, the electronic device may also include a user interface. Of course, the architecture shown in the present application is only exemplary. When implementing different devices, one or more components shown in the electronic device of the present application can be omitted according to actual needs.

[0049] Embodiment III

[0050] One embodiment of the present application discloses a computer-readable storage medium. Computer-readable instructions are stored on the computer-readable storage medium. When the computer-readable instructions are run by a processor, the chronic disease information management system based on the behavior interaction model according to the embodiments of the present application described with reference to the above drawings can be executed. The storage medium includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory, etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0051] In addition, according to the embodiments of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium storing machine-readable instructions that can be run by a processor to execute instructions corresponding to the method steps provided by the present application, such as: the chronic disease information management system based on the behavior interaction model. When the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.

[0052] The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

[0053] It should be noted that in this text, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of additional identical elements in the process, method, article or device comprising said element.

[0054] In the description of the present invention, it should be understood that terms such as "first", "second", etc. are only used for distinguishing descriptions and cannot be construed as indicating or implying relative importance.

[0055] In the description of the present invention, unless otherwise specified, the meaning of "a plurality" is two or more.

[0056] In the description of the present invention, the meaning of "several" is one or more, and the meaning of "a large number" is two or more.

[0057] In the description of this specification, the description referring to terms such as "one embodiment", "some embodiments", "examples", "specific examples", or "some examples", etc. means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0058] For the formulas in this specification, the dimensional quantities are removed and only the numerical values are calculated. The formula is obtained by collecting a large amount of data and performing software simulation to get a formula that is closest to the actual situation. The preset parameters and threshold values in the formula are set by those skilled in the art according to the actual situation.

[0059] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the claims and their equivalents.

Claims

1. A chronic disease information management system based on a behavioral interaction model, characterized by: include: A data acquisition module is used to obtain patient characteristic data, including background information, intrinsic motivation, and cognitive evaluation; The information push module is used to build a health education content library based on background information, and to push health education content using a layered push strategy in combination with cognitive evaluation; A professional support module is used to integrate background information and intrinsic motivation, using a multi-criteria decision analysis approach to generate a multi-dimensional treatment plan matrix; The decision control module is used to obtain patient willingness information, quantitatively evaluate the acceptance of different treatment plans in the multi-dimensional treatment plan matrix, and select the best treatment plan from the multi-dimensional treatment plan matrix based on the evaluation results; The emotional support module is used to obtain real-time interaction data, analyze the real-time interaction data using emotional analysis technology, dynamically formulate emotional support strategies, and provide emotional support through a pre-established multi-level emotional support network; The interactive evaluation module is used to collect patient health data and use deep learning technology to comprehensively evaluate the interactive effect; The feedback optimization module is used to intelligently optimize the health education content, optimal treatment plan and emotional support strategy in sequence according to the interaction effect through a closed-loop feedback mechanism.

2. The chronic disease information management system based on the behavioral interaction model according to claim 1 is characterized in that: The background information includes basic information, medical history and living habits; the medical history includes previous medical history and chronic disease conditions; The chronic disease condition includes the name of the disease, duration of the disease and severity; The method for constructing a health education content library comprises: Construct a content resource library, which includes health education content corresponding to different chronic diseases; each health education content in the content resource library corresponds to a set of content data, which includes the name of the chronic disease, cognitive difficulty and health field. According to the name of the chronic disease in the background information, the corresponding health education content is screened out from all the health education content in the content resource library and marked as appropriate content; determine whether the health education content has been pushed to the patient; if so, obtain historical content, use a clustering algorithm to cluster the historical content and the appropriate content, and obtain a content clusters, where a is an integer greater than 1; construct a health education content library based on the appropriate content in the content cluster corresponding to the historical content; if not, obtain historical patients, and use a clustering algorithm to cluster the historical patients and the current patients based on the patient characteristic data of the historical patients and the current patients, and obtain b patient clusters, where b is an integer greater than 1; construct a health education content library based on the appropriate content corresponding to the historical patients in the patient cluster corresponding to the current patient; wherein the historical content is the health education content pushed to the patient at a historical moment, the historical patient is the patient who has received chronic disease management at a historical moment, and the current patient is the patient who is currently undergoing chronic disease management; Methods for delivering health education content include: The health education contents in the health education content library that correspond to the same health field are taken as a content collection; the cognitive difficulty corresponding to each health education content in each content collection is compared with the cognitive evaluation respectively, and the health education contents with the same cognitive difficulty and cognitive evaluation are marked as push content, and the health education content is pushed to the current patient according to the push content.

3. The chronic disease information management system based on the behavioral interaction model according to claim 2 is characterized in that: The method for generating a multi-dimensional treatment plan matrix comprises: Obtain the treatment plans corresponding to the historical patients in the patient cluster corresponding to the current patient, and mark them as candidate treatment plans; mark all non-numerical data in the background information as text data, and use the pre-trained word embedding model to convert each text data into a corresponding text vector; set different digital labels for different intrinsic motivations and mark them as motivation labels; replace the text data in the background information with the corresponding text vectors, and use the replaced background information and motivation labels as analysis data; set different digital labels for different treatment plans and mark them as plan labels; mark the plan labels of the candidate treatment plans as candidate labels, and use each candidate label and the analysis data as a set of evaluation data, and the evaluation data and the candidate labels correspond one to one; input the evaluation data into the trained indicator evaluation model to predict the corresponding decision indicators; the decision indicators include treatment effect and ease of implementation, and the indicator evaluation model is a random forest model; A weight set is preset, and the weight set includes a weight set corresponding to each data in the decision indicator; each data in the decision indicator is multiplied by the corresponding weight coefficient in the weight set, and the results are added in sequence to obtain a decision value; a decision threshold is preset, and the decision value of each candidate treatment plan is compared with the decision threshold, and candidate treatment plans with decision values ​​greater than or equal to the decision threshold are marked as alternative plans, and candidate treatment plans with decision values ​​less than the decision threshold are not marked; a multi-dimensional treatment plan matrix is ​​generated based on all alternative plans.

4. The chronic disease information management system based on the behavioral interaction model according to claim 3 is characterized in that: Methods for selecting the best treatment options include: Obtain treatment characteristic data corresponding to each alternative plan, the treatment characteristic data including treatment goals and risk levels; use a pre-trained medical NLP model to perform entity recognition on each alternative plan in turn, and identify the treatment form corresponding to each alternative plan; use the treatment form and treatment goal of each alternative plan as a set of plan characteristic data; calculate the cosine similarity between each data in each set of plan characteristic data and the corresponding data in the corresponding patient willingness information in turn, and mark them as acceptance probability; subtract the corresponding data in the risk tolerance from each data in each set of risk levels in turn to obtain the data difference; the patient willingness information includes treatment goal preference, treatment form preference and risk tolerance; The acceptance probability and data difference corresponding to each set of alternative plans are taken as a set of evaluation characteristic data; multiple fuzzy sets are constructed for each data in the evaluation characteristic data; each set of evaluation characteristic data is converted into the membership of each corresponding fuzzy set through fuzzification technology; fuzzy rules are defined; each set of fuzzified technical feature data is matched with the fuzzy rules respectively, and fuzzy reasoning method is used for fuzzy reasoning to obtain the fuzzy reasoning result corresponding to each alternative plan, and the fuzzy reasoning result is the membership of each acceptance level, and the acceptance levels include high acceptance, medium acceptance, low acceptance and unacceptable; the acceptance interval is set, and the range of the acceptance interval is [0, c], where c is an integer greater than 1; the acceptance interval is evenly divided into four level intervals, and the level interval One-to-one correspondence with the levels in the acceptance level; add the maximum value of each level interval to the corresponding minimum value and then divide it by 2 to obtain the interval mean of each level interval; multiply each membership degree corresponding to each alternative plan by the corresponding interval mean to obtain the acceptance score; preset a proportion set, the proportion set includes the proportion coefficient corresponding to each data in the patient's willingness information; multiply each acceptance score of each alternative plan by the corresponding proportion coefficient in the proportion set, and add them up in sequence to obtain the total acceptance value of each alternative plan; add each membership degree corresponding to each alternative plan in sequence to obtain the total membership degree; divide the total acceptance value of each alternative plan by the corresponding total membership degree as the acceptance degree of each alternative plan, and take the alternative plan with the highest acceptance degree as the best treatment plan.

5. The chronic disease information management system based on the behavioral interaction model according to claim 4 is characterized in that: The method of dynamically developing emotional support strategies includes: A pre-trained emotion recognition model is used to analyze real-time interaction data and identify the patient's emotional response; a strategy set is set, which includes a telephone interaction set and an on-site interaction set; an interaction method is randomly selected from the telephone interaction set and the on-site interaction set to construct a set of interaction combinations, and a total of d groups of interaction combinations are constructed. The d groups of interaction combinations are all different, and the d groups of interaction combinations are set with increasing numerical labels in sequence and marked as set labels. The range of the set labels is [1, d]; a set label is randomly selected as the initial iteration center, and the number of iterations is set to 0; Define the iterative process: generate m candidate solutions within the range of the set label, and calculate the emotional relief effect corresponding to each candidate solution, 1<m<d, and the candidate solutions correspond to the set labels one by one; mark the candidate solution with the largest emotional relief effect as the temporary optimal solution, and move the iteration center to the temporary optimal solution; The iterative process is executed, and each time the iterative process is executed, the number of iterations is increased by one; an iteration threshold is preset, and when the number of iterations is greater than or equal to the iteration threshold, the iterative process is stopped, the set label corresponding to the iteration center is marked as the best label, and the interaction combination corresponding to the best label is used as the emotional support strategy.

6. The chronic disease information management system based on the behavioral interaction model according to claim 5 is characterized in that: Methods for generating m candidate solutions include: Preset the selection interval, take the iteration center as the center of the selection interval, and obtain m candidate solutions in the selection interval; The method for calculating the emotional relief effect corresponding to the candidate solution includes: Different numerical labels are set for different emotion types and marked as type labels; different numerical labels are set for different emotion intensities and marked as intensity labels; the set labels and the type labels and intensity labels corresponding to the emotional responses are input into the trained sentiment analysis model in sequence to predict the corresponding relief intensity; the sentiment analysis model is a random forest model, and the relief intensity is the emotional intensity of the patient after receiving the telephone interaction and on-site interaction in the interaction combination; the relief intensity is subtracted from the emotional intensity to obtain the emotional relief effect.

7. The chronic disease information management system based on the behavioral interaction model according to claim 6 is characterized in that: Ways to provide emotional support include: The multi-level emotional support network includes an automatic response layer, an artificial assistance layer and a artificial intervention layer; a pre-trained word embedding model is used to convert each emotion type into a corresponding type vector; a clustering algorithm is used to cluster all type vectors to obtain three type clusters; the three type clusters correspond to positive emotions, negative emotions and neutral emotions, respectively; the patient's emotion type is analyzed, if the emotion type is positive, the automatic response layer is used to provide emotional support, if the emotion type is neutral, the artificial assistance layer is used for emotional support, if the emotion type is negative, the artificial intervention layer is used for emotional support.

8. The chronic disease information management system based on the behavioral interaction model according to claim 7 is characterized in that: Methods for comprehensively evaluating the effectiveness of interactions include: The patient's health data and historical health data are sequentially input into the trained effect evaluation model to predict the corresponding interactive effect; wherein the patient's health data are the physiological indicators of the patient after receiving chronic disease management, the historical health data are the physiological indicators of the patient before receiving chronic disease management, the effect evaluation model is a deep learning model, and the interactive effect is the improvement effect of the chronic disease after the patient receives chronic disease management; The training process of the effect evaluation model includes: Collect n groups of different health data in advance, each group of health data includes patient health data and historical health data, set corresponding interaction effects for the n groups of health data, n is an integer greater than 1, and convert the health data and the corresponding interaction effects into a corresponding set of feature vectors; use each set of feature vectors as the input of the effect evaluation model, the effect evaluation model uses a set of predicted interaction effects corresponding to each group of health data as the output, and uses the actual interaction effect corresponding to each group of health data as the prediction target, and the actual interaction effect is the pre-set interaction effect corresponding to the health data; minimize the sum of the prediction errors of all health data as the training target; train the effect evaluation model until the sum of the prediction errors converges and stops training.

9. The chronic disease information management system based on the behavioral interaction model according to claim 8, characterized in that: Preset the effect threshold and compare the interaction effect with the effect threshold; If the interaction effect is greater than or equal to the effect threshold, the health education content, the best treatment plan and the emotional support strategy will not be intelligently optimized; if the interaction effect is less than the effect threshold, the health education content, the best treatment plan and the emotional support strategy will be intelligently optimized in turn; Methods for intelligent optimization of health education content include: Subtract the interaction effect from the effect threshold to obtain the effect difference; set different numerical labels for different cognitive evaluations and mark them as cognitive labels; The effect difference and cognitive label are input into the trained cognitive optimization model, the corresponding optimization label is predicted, and the corresponding optimization evaluation is obtained according to the optimization label. The optimization label is the digital label corresponding to the optimization evaluation, and the optimization evaluation is the cognitive evaluation after optimization. The cognitive optimization model is a deep neural network model, and the training process of the cognitive optimization model is consistent with the training process of the effect evaluation model. The cognitive difficulty corresponding to each health education content in each content set is compared with the optimization evaluation, and the health education content with the same cognitive difficulty and optimization evaluation is marked as optimized content, and the health education content is pushed to the current patient according to the optimized content.

10. The chronic disease information management system based on the behavioral interaction model according to claim 9, characterized in that: Methods for intelligent optimization of the best treatment options include: A contribution set is preset, and the contribution set includes contribution coefficients corresponding to the acceptance degree and the decision value; the acceptance degree and the decision value of each alternative plan are multiplied by the corresponding contribution coefficient, and the results are added in sequence to obtain the plan excellence, and the alternative plan with the highest plan excellence is taken as the best treatment plan; Ways to intelligently optimize emotional support strategies include: The set label and the type label and intensity label corresponding to the emotional response are input into the trained threshold prediction model in sequence to predict the corresponding termination threshold; the threshold prediction model is a deep neural network model, and the training process of the threshold prediction model is consistent with the training process of the effect evaluation model; in the iterative process, the emotional relief effect corresponding to the center of this iteration is subtracted from the emotional relief effect corresponding to the center of the previous iteration to obtain the effect change; the effect change is compared with the preset change threshold; if the effect change is greater than or equal to the change threshold, a change instruction is generated; if the effect change is less than the change threshold, no change instruction is generated; the number of times the change instruction is generated continuously is counted and marked as the number of instructions; when the number of instructions is greater than or equal to the termination threshold, the iteration process is stopped; the set label corresponding to the iteration center is marked as the best label, and the interaction combination corresponding to the best label is used as the emotional support strategy.

Images