Citizen health condition prediction and dynamic adjustment algorithm based on deep learning

Abstract

The invention discloses a citizen health condition prediction and dynamic adjustment algorithm based on deep learning, and relates to the technical field of citizen health prediction.The citizen health condition prediction and dynamic adjustment algorithm comprises a user side, a medical terminal and a health prediction system.The citizen health condition prediction and dynamic adjustment algorithm comprises the steps that glycosylated hemoglobin detection values, continuous blood glucose monitoring data and carbohydrate intake data of citizens are processed; the method comprises the steps of obtaining a citizen health condition prediction result, searching a corresponding standard intervention scheme in a medical database according to the citizen health condition prediction result, performing analysis according to health improvement conditions of citizens after the citizens adopt the standard intervention scheme, and generating a personalized health intervention scheme in combination with current personal conditions of the citizens. The pertinence and effectiveness of the intervention scheme are improved, and the scheme can be adjusted in time by monitoring and analyzing the response condition of the citizens to the intervention scheme in real time, so that the health requirements of the citizens are better met, the health level and life quality of the citizens are improved, and the medical cost is reduced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of citizen health prediction, in particular to a citizen health condition prediction and dynamic adjustment algorithm based on deep learning. BACKGROUND

[0002] With the aggravation of population aging and the change of people's lifestyle, the incidence of chronic diseases such as diabetes and cardiovascular disease is rising, which poses a serious threat to the health of citizens. Effective health management is crucial for preventing and controlling these diseases and improving the overall health level of citizens.

[0003] At present, most health intervention programs are based on general medical guidelines or experience, and lack sufficient consideration of individual health conditions, living habits, genetic factors and other personal circumstances. For example, for citizens with diabetes, traditional intervention programs often only provide uniform dietary and exercise recommendations, which cannot be adjusted according to the patient's specific condition, blood glucose fluctuations, eating habits and exercise capacity, resulting in poor actual effect of the intervention program and difficulty in effectively improving the health status of citizens. Even unreasonable programs can cause other health problems.

[0004] Therefore, in view of the above, the present application proposes a citizen health condition prediction and dynamic adjustment algorithm based on deep learning to make up for and improve the shortcomings of the prior art. SUMMARY

[0005] To solve the above technical problems, the present application provides a citizen health condition prediction and dynamic adjustment algorithm based on deep learning to solve the corresponding technical problems proposed in the background art.

[0006] To achieve the above purpose, the technical solution adopted by the present application is as follows: a citizen health condition prediction and dynamic adjustment algorithm based on deep learning, comprising: a user terminal, a medical terminal and a health prediction system, the user terminal is used to establish a user account based on user basic information, and is associated with an intelligent wearable device to obtain user health status data and send it to the health prediction system;

[0007] The health prediction system comprises a data acquisition unit, a health prediction unit and a dynamic adjustment unit.

[0008] The data acquisition unit is used to collect the glycated hemoglobin detection value, continuous blood glucose monitoring data and carbohydrate intake quantification data of citizens, output as a comprehensive health data set, and preprocess the comprehensive health data set and send it to the health prediction unit.

[0009] The health prediction unit is configured to obtain a comprehensive health data set and perform fusion processing to obtain comprehensive health data set features, extract key features in the comprehensive health data set features by using a feature extraction algorithm, construct a prediction model based on a long short-term memory network, input the key features into the prediction model in a time sequence to output a citizen health condition prediction result, and send the citizen health condition prediction result to the dynamic adjustment unit.

[0010] The dynamic adjustment unit includes an intervention scheme generation module and a health behavior tracking module. The intervention scheme generation module is configured to obtain the citizen health condition prediction result, find a corresponding standard intervention scheme from a medical database, analyze health improvement of the citizen after using the standard intervention scheme, and generate a personalized health intervention scheme in combination with current personal conditions of the citizen.

[0011] The health behavior tracking module is configured to monitor health behavior data of the citizen after executing the personalized health intervention scheme by using an intelligent device, track health behavior change of the citizen, and adjust a focus and frequency of health behavior tracking in real time based on a current health risk of the citizen, and generate a health behavior tracking report periodically and send the health behavior tracking report to a user end.

[0012] Preferably, the medical terminal is configured to obtain the citizen health condition prediction result and display the citizen health condition prediction result to a physician by using a visualization system, obtain early warning information by using the physician to assist in analyzing the citizen health condition prediction result, and send the early warning information to the user end.

[0013] Preferably, a specific process of obtaining the comprehensive health data set features is as follows.

[0014] S101, obtain a glycated hemoglobin detection value M, continuous blood glucose monitoring data P, and carbohydrate intake quantification data L of a citizen, and perform standardization processing on original data of the glycated hemoglobin detection value M, the continuous blood glucose monitoring data P, and the carbohydrate intake quantification data L to convert the original data into a standard normal distribution with a mean value of 0 and a standard deviation of 1. A calculation formula is as follows: wherein U represents the original data, μ represents a mean value of the original data, and σ represents a standard deviation of the original data.

[0015] S102, obtain the glycated hemoglobin detection value M, the continuous blood glucose monitoring data P, and the carbohydrate intake quantification data L after the standardization processing, perform fusion on the glycated hemoglobin detection value M, the continuous blood glucose monitoring data P, and the carbohydrate intake quantification data L by using a weighted summation method, and obtain comprehensive health data set features F. A formula is as follows: F = w M ×M + w P ×P + w L ×L, wherein w M , w P , and w L are weights of the glycated hemoglobin detection value M, the continuous blood glucose monitoring data P, and the carbohydrate intake quantification data L, respectively, and satisfy w M +wP +w L = 1.

[0016] As preferred, the specific process of extracting key features is as follows:

[0017] S201, obtain the comprehensive health data set features F, and calculate the Pearson correlation coefficient r between each feature and the health condition label, the calculation formula is:

[0018] Wherein, X i represents the feature value;

[0019] represents the feature mean value;

[0020] Y i represents the health condition label value;

[0021] represents the health condition label mean value;

[0022] n represents the sample size;

[0023] S202, according to the size of the calculated Pearson correlation coefficient r, extract the feature with |r|> theta as the key feature X' and output, wherein theta is a preset threshold.

[0024] As preferred, the specific process of outputting the citizen health condition prediction result is as follows:

[0025] S301, obtain the key features X', select the long short-term memory network model as the prediction model, divide the key features X' into training set and validation set, use the training set to train the prediction model, adjust the parameters of the prediction model through the back propagation algorithm, and use the validation set to evaluate the prediction model to prevent the prediction model from overfitting;

[0026] S302, arrange the key features X' according to the time sequence to form the input sequence X={x1, x2,..., x t};

[0027] Wherein, X t represents the key feature at time step t;

[0028] t represents the length of the time sequence;

[0029] S303, input the input sequence X={x1, x2,..., x t} into the prediction model, perform forward propagation calculation, output the hidden state sequence h t , and map the hidden state sequence h t through the full connection layer to obtain the citizen health condition prediction result The calculation formula is: Wherein, W represents the weight matrix of the full connection layer, and b represents the bias vector.

[0030] As preferred, the specific process of generating the personalized health intervention scheme is as follows:

[0031] S401, obtaining a citizen health status prediction result According to the comprehensive health data set, the health status of citizens is graded by using a multi-dimensional scoring algorithm, and the scoring function is set as Wherein, a g is the weight coefficient of the gth dimension, s g (D) is the scoring function of the gth dimension, D is the related data in the comprehensive health data set, h is the number of dimensions, and the health status of citizens is divided into healthy, sub-healthy, sick risk, etc. by calculating the scoring result;

[0032] S402, filtering out citizen data with the same health status grade as the current citizen from the medical database, analyzing the standard intervention scheme adopted by these citizens under the same health status grade and the health improvement situation, and evaluating the effect of the filtered standard intervention scheme, the calculation formula is:

[0033] S403, combined with the current personal situation of the citizen and the effect evaluation result of the standard intervention scheme, if the effect evaluation result of the citizens with the same health status grade as the standard intervention scheme is as expected, the standard intervention scheme is adopted, if not, the standard intervention scheme is adjusted to generate a personalized health intervention scheme that meets the current needs of the citizen, the current personal situation of the citizen includes sleep duration, eating habits, exercise habits and working environment.

[0034] As preferred, the specific process of generating the warning information is as follows:

[0035] S501, obtaining real-time monitoring data and historical health data of citizens through intelligent wearable devices, combining with the citizen health status prediction result, and analyzing by using a risk assessment model to evaluate various health risks faced by citizens, and the risk assessment index is set as Wherein, k j is the weight of the jth risk factor, r j is the risk value of the jth risk factor, and m is the number of risk factors;

[0036] S502, according to the risk assessment result, when the risk value E exceeds the preset threshold, the warning information is generated, the content of the warning information includes the current health problems of the citizen, the serious consequences caused, the emergency measures recommended to be taken and the key points needing to be paid attention to in the future, and the warning information is sent to the family end.

[0037] As preferred, the specific process of generating the health behavior tracking report is as follows:

[0038] S601, obtaining health behavior data of the citizen after executing the personalized health intervention scheme through the intelligent device, the intelligent device including a smart wearable device, a smart tableware and a smart medicine box, the health behavior data including diet intake, exercise condition, medicine taking record and sleep quality;

[0039] S602, dynamically adjusting the focus and frequency of the health behavior tracking according to the health risk currently faced by the citizen, comparing the actual health behavior of the citizen with the requirement in the personalized health intervention scheme, analyzing the change of the citizen in each health behavior, and calculating the compliance rate of the citizen in each health behavior index, the calculation formula being:

[0040] S603, regularly generating the health behavior tracking report and sending it to the family end, the health behavior tracking report including the execution condition of the citizen in diet, exercise, medicine taking and sleep, the influence of the health behavior change on the health condition, the execution condition feedback of the personalized health behavior improvement suggestion and the subsequent focus and suggestion.

[0041] Compared with the prior art, the beneficial effects of the present application are: by processing the glycated hemoglobin detection value, continuous blood glucose monitoring data and carbohydrate intake quantification data of the citizen, obtaining the health condition prediction result of the citizen, searching for the corresponding standard intervention scheme in the medical database according to the health condition prediction result of the citizen, and analyzing the health improvement condition of the citizen after adopting the standard intervention scheme, generating the personalized health intervention scheme combined with the current personal condition of the citizen, not only improving the pertinence and effectiveness of the intervention scheme, but also through real-time monitoring and analysis of the response condition of the citizen to the intervention scheme, the scheme can be adjusted in time, so as to better meet the health needs of the citizen, improve the health level and life quality of the citizen, and reduce the medical cost. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 It is a whole structure schematic view of a preferred embodiment of the present application. DETAILED DESCRIPTION

[0043] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application.

[0044] Embodiments of the present application:

[0045] Referring to Figure 1 Fig. 1 shows a deep learning-based citizen health status prediction and dynamic adjustment algorithm, which comprises a user terminal, a medical terminal and a health prediction system. The user terminal is used to establish a user account based on user basic information and associate with an intelligent wearable device to obtain user health status data and send the data to the health prediction system.

[0046] The health prediction system comprises a data acquisition unit, a health prediction unit and a dynamic adjustment unit.

[0047] The data acquisition unit is used to acquire the glycated hemoglobin detection value, continuous blood glucose monitoring data and carbohydrate intake quantification data of citizens, output the data as a comprehensive health data set, and pre-process the comprehensive health data set and send it to the health prediction unit.

[0048] The health prediction unit is used to obtain the comprehensive health data set and perform fusion processing to obtain comprehensive health data set features. A feature extraction algorithm is used to extract key features from the comprehensive health data set features. A prediction model is constructed based on a long short-term memory network. The key features are input into the prediction model in time sequence to output a citizen health status prediction result and send it to the dynamic adjustment unit.

[0049] The dynamic adjustment unit comprises an intervention scheme generation module and a health behavior tracking module. The intervention scheme generation module is used to obtain the citizen health status prediction result, find the corresponding standard intervention scheme from a medical database, analyze the health improvement of the citizen after adopting the standard intervention scheme, and generate a personalized health intervention scheme combined with the current personal situation of the citizen. The medical database is used to store patient (citizen) medical information, including but not limited to patient basic information, medical records, diagnosis results, treatment schemes, drug prescriptions, laboratory results, genetic information, lifestyle data and environmental factor data.

[0050] The health behavior tracking module is used to monitor the health behavior data of the citizen after executing the personalized health intervention scheme through the intelligent device, track the health behavior change of the citizen, and adjust the focus and frequency of health behavior tracking in real time based on the current health risk faced by the citizen, and generate a health behavior tracking report periodically and send it to the user terminal.

[0051] The medical terminal is used to obtain the citizen health status prediction result and display it to the physician through a visualization system. The physician analyzes the citizen health status prediction result to obtain early warning information and sends it to the user terminal.

[0052] The specific process of obtaining the comprehensive health data set features is as follows:

[0053] S101, obtain the glycated hemoglobin detection value M, continuous blood glucose monitoring data P and carbohydrate intake quantification data L of citizens, and standardize the original data to convert the original data into a standard normal distribution with a mean of 0 and a standard deviation of 1, and the calculation formula is: Wherein, U represents the original data, μ represents the mean of the original data, and σ represents the standard deviation of the original data;

[0054] S102, obtain the glycated hemoglobin detection value M, continuous blood glucose monitoring data P and carbohydrate intake quantification data L after standardization, and fuse them by using the weighted summation method to obtain the comprehensive health data set feature F, and the formula is: F = w M ×M+w P ×P+w L ×L, wherein w M , w P and w L are the weights of the glycated hemoglobin detection value M, continuous blood glucose monitoring data P and carbohydrate intake quantification data L, and satisfy w M +w P +w L =1.

[0055] The specific process of extracting key features is as follows:

[0056] S201, obtain the comprehensive health data set feature F, and calculate the Pearson correlation coefficient r between each feature and the health status label, and the calculation formula is:

[0057] Wherein, X i represents the feature value;

[0058] represents the feature average value;

[0059] Y i represents the health status label value;

[0060] represents the health status label average value;

[0061] n represents the sample size;

[0062] S202, according to the size of the calculated Pearson correlation coefficient r, extract the feature with |r|>θ as the key feature X' and output, wherein θ is a preset threshold.

[0063] The specific process of outputting the citizen health status prediction result is as follows:

[0064] S301, acquire the key feature X', select a long short-term memory network model as a prediction model, divide the key feature X' into a training set and a validation set, train the prediction model using the training set, adjust the parameters of the prediction model through a back propagation algorithm, and evaluate the prediction model using the validation set to prevent overfitting of the prediction model;

[0065] S302, arrange the key feature X' in a time sequence to form an input sequence X={x1, x2,..., x t};

[0066] Wherein, X t represents the key feature at time step t;

[0067] t represents the length of the time sequence;

[0068] S303, input the input sequence X={x1, x2,..., x t} into the prediction model, perform forward propagation calculation, output a hidden state sequence h t , and map the hidden state sequence h t through a fully connected layer to obtain a citizen health status prediction result The calculation formula is: Wherein, W represents the weight matrix of the fully connected layer, and b represents the bias vector.

[0069] The specific process of generating a personalized health intervention plan is as follows:

[0070] S401, acquire the citizen health status prediction result According to the comprehensive health data set, a multi-dimensional scoring algorithm is used to grade the citizen health status, and the scoring function is set as Wherein, a g is the weight coefficient of the gth dimension, s g (D) is the scoring function of the gth dimension, D is the related data in the comprehensive health data set, and h is the number of dimensions. By calculating the scoring result, the citizen health status is divided into healthy, sub-healthy, and high-risk levels;

[0071] S402, filter out citizen data with the same citizen health status level from the medical database, analyze the standard intervention plan adopted by these citizens under the same citizen health status level and their health improvement, and evaluate the effect of the filtered standard intervention plan, the calculation formula is:

[0072] S403, combine the current personal situation of the citizen and the effect evaluation result of the standard intervention scheme, if the effect evaluation result of the citizen with the same health status level using the standard intervention scheme meets the expectation, the standard intervention scheme is adopted, if it does not meet the expectation, the standard intervention scheme is personalized adjusted to generate a personalized health intervention scheme meeting the current needs of the citizen, the current personal situation of the citizen includes sleep duration, eating habit, exercise habit and working environment;

[0073] The sleep duration of the citizen is recorded by the intelligent wearable device, and a sleep duration standard value a is set, if the sleep duration of the citizen is less than a, the exercise time and diet suggestion in the standard intervention scheme are preferentially adjusted to avoid high-intensity exercise at night and increase food rich in tryptophan to promote sleep;

[0074] The daily meal structure of the citizen is recorded by the intelligent device, and high-risk eating habits (high salt) are marked, a daily average salt intake standard value b is set, if the daily average salt intake of the citizen is greater than b, the proportion of low-sodium salt replacement in the standard intervention scheme is increased, and the daily salt intake monitoring target is set;

[0075] The citizen's sedentary situation is obtained by the intelligent wearable device, if the citizen is in a sedentary working environment, five minutes of micro-movement (such as stretching, standing on tiptoe) per hour is added in the standard intervention scheme, and office table lifting equipment is recommended.

[0076] The specific process of generating the warning information is as follows:

[0077] S501, the real-time monitoring data and the historical health data of the citizen are obtained by the intelligent wearable device, combined with the health status prediction result of the citizen, the risk assessment model is used for analysis, the various health risks faced by the citizen are evaluated, and the risk assessment index is Wherein, k j is the weight of the jth risk factor, r j is the risk value of the jth risk factor, and m is the number of risk factors;

[0078] S502, according to the risk assessment result, when the risk value E exceeds the preset threshold, the warning information is generated, the content of the warning information includes the current health problems of the citizen, the serious consequences caused, the emergency measures suggested to be taken and the key points needing to be paid attention to in the future, and the warning information is sent to the family end.

[0079] The specific process of generating the health behavior tracking report is as follows:

[0080] S601, the health behavior data of the citizen after executing the personalized health intervention scheme is obtained by the intelligent device, the intelligent device includes intelligent wearable device, intelligent tableware and intelligent medicine box, the health behavior data includes diet intake, exercise, medicine taking record and sleep quality;

[0081] S602, according to the current health risk faced by the citizen, dynamically adjusting the focus and frequency of health behavior tracking, comparing the actual health behavior of the citizen with the requirements in the individualized health intervention plan, analyzing the change of the citizen in each health behavior, and calculating the compliance rate of the citizen in each health behavior index, the calculation formula is:

[0082] S603, generating a health behavior tracking report periodically and sending it to the family end, the health behavior tracking report includes the execution of the citizen in diet, exercise, medication and sleep, the influence of health behavior change on health status, the execution feedback of individualized health behavior improvement suggestion and the subsequent focus and suggestion.

[0083] By processing the glycosylated hemoglobin detection value, continuous blood glucose monitoring data and carbohydrate intake quantitative data of the citizen, the health status prediction result of the citizen is obtained, according to the health status prediction result of the citizen, the corresponding standard intervention plan is found in the medical database, and the health improvement situation of the citizen after adopting the standard intervention plan is analyzed, and the individualized health intervention plan is generated according to the current personal situation of the citizen, which not only improves the pertinence and effectiveness of the intervention plan, but also through real-time monitoring and analysis of the response of the citizen to the intervention plan, the scheme can be adjusted in time, so as to better meet the health needs of the citizen, improve the health level and life quality of the citizen, and reduce the medical cost.

[0084] The size of the interval and the threshold is set for easy comparison, and the size of the threshold depends on the number of sample data and the base number set by the person skilled in the art for each group of sample data; as long as it does not affect the proportional relationship between the parameters and the quantized values.

[0085] The above formulas are dimensionless values, the formulas are obtained by software simulation of a large amount of data to obtain the most real situation, and the preset parameters in the formula are set by the person skilled in the art according to the actual situation;

[0086] In the two embodiments provided in the present application, it should be understood that the disclosed device and system can be implemented in other ways; for example, the device embodiments described above are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, a plurality of modules or components can be combined or integrated into another system, or some features can be ignored or not executed; another point, the coupling or direct coupling or communication connection between the displayed or discussed each other can be through some interface, indirect coupling or communication connection between the devices or modules, which can be electrical, mechanical or other forms;

[0087] The above merely describes preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art, according to the technical solution and inventive concept of the present application, makes equivalent replacement or change within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A deep learning-based citizen health condition prediction and dynamic adjustment algorithm, characterized in that, The application relates to a user terminal, a medical terminal and a health prediction system, wherein the user terminal is used for establishing a user account based on user basic information and acquiring user health state data from an intelligent wearable device and sending the data to the health prediction system; the health prediction system comprises a data acquisition unit, a health prediction unit and a dynamic adjustment unit; the data acquisition unit is used for acquiring a glycated hemoglobin detection value of citizens, continuous blood glucose monitoring data and a quantitative data of carbohydrate intake, outputting the data as a comprehensive health data set, and preprocessing the comprehensive health data set and sending the data to the health prediction unit; the health prediction unit is used for acquiring the comprehensive health data set, performing fusion processing to obtain comprehensive health data set features, extracting key features in the comprehensive health data set features by using a feature extraction algorithm, constructing a prediction model based on a long short-term memory network, inputting the key features into the prediction model in a time sequence, and outputting a citizen health condition prediction result and sending the result to the dynamic adjustment unit; the dynamic adjustment unit comprises an intervention scheme generation module and a health behavior tracking module; the intervention scheme generation module is used for acquiring the citizen health condition prediction result, searching for a corresponding standard intervention scheme from a medical database, analyzing health improvement conditions of the citizen after the citizen adopts the standard intervention scheme, and generating an individualized health intervention scheme in combination with current personal conditions of the citizen; the health behavior tracking module is used for monitoring health behavior data of the citizen after the citizen executes the individualized health intervention scheme through an intelligent device, tracking health behavior change conditions of the citizen, and adjusting a focus and a frequency of health behavior tracking in real time based on current health risks of the citizen, and generating a health behavior tracking report and sending the report to the user terminal at regular time intervals; the medical terminal is used for acquiring the citizen health condition prediction result and displaying the result to a doctor through a visual system, assisting the doctor in analyzing the citizen health condition prediction result to obtain early warning information, and sending the early warning information to the user terminal. The specific process of obtaining the comprehensive health data set features is as follows: The specific process of extracting the key features is as follows: n represents a sample number; S202, according to the size of the calculated Pearson correlation coefficient r, a feature of r|>theta is extracted as a key feature X' and outputted, wherein theta is a preset threshold. The specific process of outputting the citizen health condition prediction result is as follows: S301, the key features X' are acquired, a long short-term memory network model is selected as a prediction model, the key features X' are divided into a training set and a verification set, the training set is used for training the prediction model, parameters of the prediction model are adjusted through a back propagation algorithm, the verification set is used for evaluating the prediction model, and overfitting of the prediction model is prevented; 2. The algorithm for predicting and dynamically adjusting the health status of citizens based on deep learning according to claim 1, characterized in that, t represents a length of a time sequence; 3. The algorithm for predicting and dynamically adjusting the health status of citizens based on deep learning according to claim 2, characterized in that, The specific process of generating the individualized health intervention scheme is as follows: S101, obtain the glycated hemoglobin detection value M of the citizen, the continuous blood glucose monitoring data P and the carbohydrate intake quantification data L, and standardize the original data thereof, convert the original data into a standard normal distribution with a mean of 0 and a standard deviation of 1, and the calculation formula is: wherein U represents the original data, μ represents the mean of the original data, and σ represents the standard deviation of the original data; S102, obtain the standardized glycosylated hemoglobin detection value M, continuous blood glucose monitoring data P and carbohydrate intake quantification data L, adopt a weighted summation method to fuse them, and obtain a comprehensive health data set feature F, and the formula is: F=w M ×M+w P ×P+w L ×L, wherein w M , w P and w L are weights of the glycosylated hemoglobin detection value M, the continuous blood glucose monitoring data P and the carbohydrate intake quantification data L respectively, and satisfy w M +w P +w L =1.

4. The algorithm for predicting and dynamically adjusting the health status of citizens based on deep learning according to claim 3, characterized in that, ​ S201、Obtain the comprehensive health data set feature F, calculate the Pearson correlation coefficient r between each feature and the health condition label, and the calculation formula is: wherein X i represents an eigenvalue; represents the average value of the characteristic; Y i represents a health condition label value; represents the average value of the health condition label; ​ ​ 5. The algorithm for predicting and dynamically adjusting the health status of citizens based on deep learning according to claim 4, characterized in that, ​ ​ S302, arrange the key features X' in time sequence to form an input sequence X = {x1, x2,..., xT} of the key features X'. t} wherein X t represents a key feature at time step t; ​ S303, Given the input sequence X = {x1, x2, ..., x...} t The input is fed into the prediction model, where forward propagation is performed, and the hidden state sequence h is output. t and the hidden state sequence h t As input, the data is mapped through a fully connected layer to obtain the predicted results of citizens' health status. The calculation formula is as follows: Where W represents the weight matrix of the fully connected layer, and b represents the bias vector.

6. The algorithm for predicting and dynamically adjusting the health status of citizens based on deep learning according to claim 5, characterized in that, ​ S401、Obtaining a citizen health condition prediction result According to the comprehensive health data set, the health condition of citizens is graded by using a multi-dimensional scoring algorithm, and a scoring function is set as Wherein, a g is the weight coefficient of the gth dimension, s g (D) is the scoring function of the gth dimension, D is the related data in the comprehensive health data set, h is the number of dimensions, and the health condition of citizens is divided into healthy, sub-healthy, sick risk and other levels by calculating the scoring result. S402, screening citizen data with the same citizen health condition level from the medical database, analyzing the standard intervention scheme adopted by these citizens under the same citizen health condition level and their health improvement, and evaluating the effect of the screened standard intervention scheme, the calculation formula of which is: S403, combine the current personal situation of the citizen and the effect evaluation result of the standard intervention scheme, if the effect evaluation result of the same citizen health status level using the standard intervention scheme meets the expectation, the standard intervention scheme is adopted, if it does not meet the expectation, the standard intervention scheme is personalized adjusted to generate a personalized health intervention scheme meeting the current citizen demand, the current personal situation of the citizen includes sleep duration, eating habit, exercise habit and working environment.

7. The algorithm for predicting and dynamically adjusting the health status of citizens based on deep learning according to claim 6, characterized in that, The specific process of generating the early warning information is as follows: S501, acquire real-time monitoring data and historical health data of the citizen through the smart wearable device, combine the citizen health condition prediction result, analyze by using a risk assessment model, assess various health risks faced by the citizen, and set the risk assessment index as wherein, k j is the weight of the jth risk factor, r j is the risk value of the jth risk factor, and m is the number of risk factors; S502, according to the risk evaluation result, when the risk value E exceeds the preset threshold, generate the early warning information, the early warning information content includes the current health problem of the citizen, the serious consequences caused, the emergency measures suggested to be taken and the key to be concerned subsequently, and send the early warning information to the family end.

8. The algorithm for predicting and dynamically adjusting the health status of citizens based on deep learning according to claim 7, characterized in that, The specific process of generating the health behavior tracking report is as follows: S601, obtain the health behavior data of the citizen after executing the personalized health intervention scheme through the intelligent device, the intelligent device includes the intelligent wearable device, the intelligent tableware and the intelligent medicine box, the health behavior data includes the diet intake, the exercise condition, the medicine taking record and the sleep quality; S602, according to the current health risk faced by the citizen, dynamically adjust the focus and frequency of health behavior tracking, compare the actual health behavior of the citizen with the requirements in the individualized health intervention plan, analyze the change of the citizen in each health behavior, and calculate the compliance rate of the citizen in each health behavior index, and the calculation formula is: S603, generate the health behavior tracking report regularly and send it to the family end, the health behavior tracking report includes the execution condition of the citizen in the diet, the exercise, the medicine taking and the sleep, the influence of the health behavior change on the health condition, the execution condition feedback of the personalized health behavior improvement suggestion, and the subsequent key to be concerned and the suggestion.

Images