Sub-health state intelligent conversion strategy generation method and system based on intelligent wearable device

Abstract

The invention provides a sub-health state intelligent conversion strategy generation method and system based on intelligent wearable equipment. The method comprises the following steps: acquiring physiological data and non-privacy data of a user; dividing the obtained data into adjustable parameters and non-adjustable parameters, and performing discretization interval division on continuous value parameters; constructing a training set based on the discretized parameters and state labels, and training a data model to calculate an expected conversion rate of each adjustable parameter; taking the non-adjustable parameter as a root node, constructing a transformation tree according to the expected transformation rate and the selectable value of the adjustable parameter, and performing pruning optimization on the transformation tree; inputting user data and a conversion target, traversing the conversion tree to generate a compatible branch list, calculating the goodness of each branch and sorting to generate a goodness sequence; and the verification goodness sequence presets the first few conversion schemes, and the schemes conforming to the expectation serve as alternative strategies to be output. According to the method, the comprehensive balance and accurate evaluation of each conversion scheme are realized, and then the conversion alternative strategy with high application value is screened out.

Description

Technical Field

[0001] The present invention relates to the technical fields of artificial intelligence and health data monitoring. Specifically, it relates to a method and system for generating intelligent transformation strategies for sub-healthy states based on intelligent wearable devices. Background Art

[0002] People's attention to health has been increasing day by day. The sub-healthy state among the young population, as a physical state between health and disease, also widely exists. According to relevant statistical data, in a fast-paced life and work environment, such as long-term fatigue, poor sleep quality, and decreased physical resistance. From the sub-healthy state to the gradual development of various diseases, there is often a long process. However, traditional health management methods are difficult to provide accurate and personalized intervention measures to effectively prevent and improve the adverse changes from the healthy state to the emergence of disease symptoms.

[0003] Under the background of artificial intelligence technology, common wearable devices such as smart bracelets can real-time monitor a user's multiple physiological data, such as heart rate, blood pressure, walking steps, sleep duration, etc. With the continuous development of technology, the large amount of user data accumulated by them provides the possibility for analyzing, predicting, and improving the health state. But how to mine valuable cognitive information from these massive data and formulate scientific and feasible strategies for the transformation of poor health states has become an urgent problem to be solved in health management.

[0004] Patent application document CN118841120A discloses a health data transformation method and system for home health Internet of Things devices. The method includes: constructing symptom transformation rules; obtaining health-related data; querying whether there are symptom transformation rules; extracting the symptom transformation rules corresponding to the health device ID, and judging whether the data content meets the determination conditions; obtaining the latest health-related data. If the current health device ID meets the next device ID, and the acquisition time of the current health-related data meets the duration threshold, when the number of repetitions is greater than the repetition number flag in the symptom transformation rule, set the condition satisfaction flag and increment the condition satisfaction count; when the condition satisfaction count is greater than the condition satisfaction count threshold in the symptom transformation rule, encode the current symptom transformation rule, the health-related data of each health device, and the symptom name together and report them. However, this patent cannot completely solve the existing technical problems and also cannot meet the requirements of the present invention. Summary of the Invention

[0005] Aiming at the defects in the prior art, the purpose of the present invention is to provide a method and system for generating intelligent transformation strategies for sub-healthy states based on intelligent wearable devices.

[0006] According to the method for generating intelligent transformation strategies for sub-healthy states based on intelligent wearable devices provided by the present invention, it includes:

[0007] Step S1: Obtain the user's physiological data and non-privacy data, where the physiological data includes at least one of heart rate, sleep duration, number of exercise steps, sedentary time, blood oxygen saturation, and blood pressure value, and the non-privacy data includes at least one of age, gender, and chronic disease label;

[0008] Step S2: Divide the obtained data into adjustable parameters and non-adjustable parameters, and perform discretized interval division on continuous value parameters;

[0009] Step S3: Construct a training set based on the discretized parameters and status labels, and train a data model to calculate the expected conversion rate of each adjustable parameter;

[0010] Step S4: Use the non-adjustable parameters as the root nodes, construct a conversion tree according to the expected conversion rate and optional values of the adjustable parameters, and perform pruning optimization on the conversion tree;

[0011] Step S5: Input the user data and conversion target, traverse the conversion tree to generate a list of compatible branches, calculate the goodness of each branch and sort them to generate a goodness sequence;

[0012] Step S6: Verify the conversion schemes in the top few positions of the preset goodness sequence, and output the schemes that meet the expectations as alternative strategies.

[0013] Preferably, the specific steps for constructing the conversion tree in step S4 include:

[0014] Step S401: Construct a decision tree based on sample data, and extract all branches and clusters;

[0015] Step S402: Calculate the parameter adjustment coefficient according to the conversion difficulty score and the expected conversion rate, and generate a sequence of parameters to be adjusted;

[0016] Step S403: Select adjustable parameters in ascending order of the adjustment coefficient, divide the data set and generate child nodes;

[0017] Step S404: Traverse all adjustable parameters to generate a complete conversion tree, and remove redundant branches through pruning.

[0018] Preferably, the formula for calculating the goodness in step S5 is:

[0019] O = α1k1 + α2k2 + α3k3 + α4k4 + α5k5 + α6k6

[0020] where O represents the goodness, α1, α2, α3, α4, α5, and α6 represent weights, k1 represents the correlation degree of the number of walking steps, k2 represents the correlation degree of the daily exercise time, k3 represents the correlation degree of the sedentary time, k4 represents the correlation degree of the daily calorie consumption, k5 represents the correlation degree of the sleep duration, and k6 represents the expected conversion rate.

[0021] Preferably, the calculation formula for the correlation degree of the number of walking steps in step S5 is as follows:

[0022]

[0023] where k1(x) represents the correlation degree of the number of walking steps, ρ represents the calculation operation of the extension distance, D represents the calculation operation of the extension position value, x0 represents the optimal value of the daily walking steps, x represents the number of walking steps, X represents the range where the daily walking steps meet the requirements, and X0 represents several smaller ranges of the number of walking steps in X. represents the range that meets the requirements in X and in a specific situation. represents the selectable range of all the number of walking steps.

[0024] Preferably, the calculation formula for the expected conversion rate in step S5 is as follows:

[0025]

[0026] R(A i ,A j ) = P(C|A i ,X) - p′(C|A j ,X)

[0027] where, for parameter A, there are n optional values for parameter A, and these values are represented as A1, A2, …, A n ; R(A i ,A j ) represents the conversion rate of A i converted to A j ; E(A) is the expected conversion rate of parameter A; when two parameter values of A are converted into each other, an R value is obtained; in R(A i ,A j ), A i and A j are two different values of parameter A; only the positive value of R needs to be calculated, that is, R(A i ,A j ) > 0, ΣΣ represents the summation over all possible combinations of A i and A j , and R(A i ,A j ) > 0 means only calculating the conversion where R is positive, and m represents the total number of all positive R values;

[0028] For R(A i ,A j ), in the decision tree, for a specific input feature X, reaching the leaf node A with the A value of A i , the probability of being classified as the expected classification C is denoted as P(C|Ai , X); Change the value of leaf node A to A i to A j After that, for the same input feature X, the new probability of being classified into the desired classification C is denoted as p′(C|A j , X). The expected conversion rate is obtained from the average of the conversion rates. The average reduces a large number of data points to a single value, simplifying the data. The average provides a measure of the central position of a data set and provides a measure of the central tendency of the expected conversion rate. The conversion rate is based on a decision tree and is very intuitive. The conversion rate is the probability of a calculated value change, and this probability is based on the classification provided by the decision tree. From the perspective of the probability distribution, it is more in line with the distribution characteristics of the actual data. Due to the clear structure of the decision tree, it is easy to identify which decision paths have the greatest impact on the conversion rate. By optimizing these critical paths, the conversion rate can be more effectively improved. The decision tree model can dynamically adjust the classification rules according to new data, which means that as the user behavior characteristics change, the model can be continuously optimized, and then a more suitable conversion rate can be obtained.

[0029] According to the sub - health state intelligent conversion strategy generation system based on intelligent wearable devices provided by the present invention, it includes:

[0030] Module M1: Obtain the physiological data and non - privacy data of the user. The physiological data includes at least one of heart rate, sleep duration, number of exercise steps, sedentary time, blood oxygen saturation, blood pressure value, etc. The non - privacy data includes at least one of age, gender, chronic disease label, etc.;

[0031] Module M2: Divide the obtained data into adjustable parameters and non - adjustable parameters, and perform discrete interval division on continuous value parameters;

[0032] Module M3: Construct a training set based on the discretized parameters and status labels, and train a data model to calculate the expected conversion rate of each adjustable parameter;

[0033] Module M4: Use the non - adjustable parameter as the root node, construct a conversion tree according to the expected conversion rate and optional values of the adjustable parameters, and perform pruning optimization on the conversion tree;

[0034] Module M5: Input user data and conversion targets, traverse the conversion tree to generate a compatible branch list, calculate the goodness of each branch and sort them to generate a goodness sequence;

[0035] Module M6: Verify the conversion schemes of the first few preset positions in the goodness sequence, and output the schemes that meet the expectations as alternative strategies.

[0036] Preferably, the specific steps of constructing the conversion tree in the module M4 include:

[0037] Module M401: Construct a decision tree based on sample data, extract all branches and clusters;

[0038] Module M402: Calculate the parameter adjustment coefficient according to the conversion difficulty score and the expected conversion rate, and generate a sequence of parameters to be adjusted;

[0039] Module M403: Select adjustable parameters in ascending order of the adjustment coefficient, divide the data set and generate child nodes;

[0040] Module M404: Traverse all adjustable parameters to generate a complete conversion tree, and remove redundant branches by pruning.

[0041] Preferably, the formula for the goodness in Module M5 is:

[0042] O = α1k1 + α2k2 + α3k3 + α4k4 + α5k5 + α6k6

[0043] Where O represents the goodness, α1, α2, α3, α4, α5 and α6 represent weights, k1 represents the correlation degree of the number of walking steps, k2 represents the correlation degree of the daily exercise time, k3 represents the correlation degree of the sedentary time, k4 represents the correlation degree of the daily calorie consumption, k5 represents the correlation degree of the sleep duration, and k6 represents the expected conversion rate.

[0044] Preferably, the formula for the correlation degree of the number of walking steps in Module M5 is as follows:

[0045]

[0046] Where k1(x) represents the correlation degree of the number of walking steps, ρ represents the extension distance calculation operation, D represents the extension position value calculation operation, x0 represents the optimal value of the daily walking steps, x represents the number of walking steps, X represents the range of the daily walking steps that meets the requirements, and X0 represents several smaller ranges of the number of walking steps in X. represents the range that meets the requirements in X and in a specific case, represents the selectable range of all walking steps.

[0047] Preferably, the formula for the expected conversion rate in Module M5 is as follows:

[0048]

[0049] R(A i ,A j ) = P(C|A i ,X) - p′(C|A j ,X)

[0050] Where, for parameter A, parameter A has n optional values, and these values are represented as A1, A2, …, A n; R(A i ,A j ) means A i Convert to A j E(A) is the expected conversion rate of parameter A; when two parameter values ​​of A are converted to each other, an R value is obtained; R(A i ,A j ), A i and A j are two different values ​​of parameter A; only the positive value R needs to be calculated, that is, R(A i ,A j )>0, ΣΣ means for all possible A i and A j The combination of R(A i ,A j )>0 means that only the transformations with positive R are counted, and m represents the total number of all positive R values;

[0051] For R(A i ,A j ), in the decision tree, for a specific input feature X, when reaching the leaf node A, the value of A is A i , the probability of being classified as the expected classification C, denoted as P(C|A i ,X); Change the value of leaf node A i A j After that, for the same input feature X, the new probability of being classified as the expected category C is recorded as p′(C|A j ,X). Compared with the prior art, the present invention has the following beneficial effects:

[0052] (1) The present invention uses the rich data collected by the smart bracelet to build a transformation tree system for the sub-health state of the human body, so as to deeply analyze the key position and improvement priority of human parameters such as heart rate, sleep duration, exercise time, sedentary time and sleep quality in the sub-health transformation process. By comprehensively traversing the transformation tree, various values ​​of the above parameters are systematically replaced and combined, thereby continuously generating new and diverse transformation solutions;

[0053] (2) The merit ranking mechanism introduced in the present invention cleverly integrates the cost factors such as manpower and time required for the implementation of the scheme and the expected effect of achieving substantial improvement in sub-health status, thereby achieving a comprehensive balance and accurate evaluation of various transformation schemes, and then screening out transformation alternative strategies with great application value;

[0054] (3) The solution generated by the present invention has strong adaptability. Whether it is faced with the regular time series data generated by smart bracelets or unstructured information such as users' sleep feelings, it can efficiently carry out data mining work, providing strong technical support and strategic guidance for the intelligent transformation of the sub-healthy state of the human body. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] By reading the following detailed description of the non-restrictive embodiments with reference to the accompanying drawings, other features, objects, and advantages of the present invention will become more apparent:

[0056] Figure 1 Flow chart of the method for transforming the sub-healthy state of the human body for a smart bracelet;

[0057] Figure 2 Flow chart for constructing a transformation tree;

[0058] Figure 3 Flow chart for traversing the transformation tree. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0059] The present invention will be described in detail below with reference to specific embodiments. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the present invention in any form. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several changes and improvements can still be made. These all belong to the protection scope of the present invention.

[0060] Embodiment

[0061] The object of the present invention is to overcome the deficiency of the existing technology in quantitative analysis for the prevention and transformation of the health state, and to provide a method for intelligent transformation of the health state based on data from intelligent wearable devices such as smart bracelets. By using the extension set principle and extension intelligent methods of extension theory, effectively utilize the data generated by intelligent wearable devices such as smart bracelets to generate prediction and transformation strategies for future health states. By processing and deeply mining the key health-related information in the multi-source heterogeneous data of intelligent devices, construct a personalized health state transformation model, and obtain feasible measures for the transformation of the sub-healthy state to the healthy state through intelligent algorithms, providing scientific and personalized health state improvement solutions for users, and promoting the development and application of health intelligent management technology.

[0062] As Figure 1 , the method for transforming the sub-healthy state of the human body proposed by the smart bracelet based on the collected data is as follows:

[0063] S1: Obtain non-privacy data such as the user's age, gender, permanent residence, heart rate, sleep duration, number of steps during exercise time, sedentary time, bedtime, wake-up time, deep sleep and light sleep duration, ratio, meal time, etc. At the same time, label or obtain their certain chronic disease conditions, such as whether they have hypertension? Whether they have diabetes, etc. as status labels;

[0064] S2: Divide these parameters into adjustable parameters and non-adjustable parameters. For example, age, gender, etc. can be used as non-adjustable parameters, while wake-up time, breakfast time, exercise time, etc. are set as adjustable parameters, and set the degree of difficulty of conversion for the adjustable (convertible) parameters. After the division, judge the discreteness of the parameter values. If they are continuous values (such as heart rate data), then perform discrete interval division of the parameter values.

[0065] S3: According to the discrete values and status labels of each parameter, divide them into a training set, a test set and a validation set, construct and train a data model, and obtain a conversion strategy and calculate the expected conversion rate through calculation. For example: for men aged 35 - 40, with a bedtime after 23:30 and an average daily number of exercise steps less than 3000, the probability of getting hypertension after the age of 55 is 76.2%. If the number of exercise steps is adjusted to 5000 - 6000 steps, then the expected conversion rate of getting hypertension in the future is 69.7%.

[0066] S4: Set non-adjustable parameters (such as age, gender) as the root nodes, and construct a conversion tree according to the expected conversion rates and optional values of each adjustable parameter (such as sleep duration, exercise time). Each leaf branch of the conversion tree corresponds to a plan, and prune and optimize the conversion tree;

[0067] S5: Input or select the target to be converted, such as reducing the probability of getting diabetes. Then traverse the leaf branch nodes on the generated conversion tree to determine several conversion plans, and calculate the goodness of each conversion plan; sort all conversion plans according to the goodness of the conversion plans to obtain a goodness sequence;

[0068] S6: Test the results of the conversion plans in the front several positions of the goodness sequence. If they do not meet the expectations of improving the sub-healthy state, then feedback the new test results, re-supplement the samples and execute S1, and output the plans that meet the expectations as alternative conversion strategies.

[0069] In a specific embodiment, apply this method to a health analysis database, with the target being whether the object is healthy, and extract gender, age, work and rest time, blood pressure, heart rate parameters, bedtime, average daily number of exercise steps in the database as several parameters in step S1.

[0070] The formula for the information gain ratio is as follows:

[0071]

[0072] Further, in step S1, after dividing a number of parameters into adjustable parameters and setting the difficulty level of conversion, it further includes performing a discreteness judgment on the adjustable parameters. If the adjustable parameter is a continuous value, a parameter discretization operation is performed.

[0073] Further, as Figure 2 , in step S4, the construction of the conversion tree includes:

[0074] S401: Construct a decision tree based on the sample data, obtain all branches of the decision tree, and for each branch, obtain the leaf nodes and all other nodes on the branch to form a branch clustering.

[0075] S402: Score the difficulty level of conversion of all parameters (the easier the conversion, the higher the score), calculate the expected conversion rate of all parameters, multiply the conversion difficulty level score by the expected conversion rate to obtain a parameter adjustment coefficient, and sort the parameters in ascending order of the parameter adjustment coefficient to obtain a sequence of parameters to be adjusted. The root node obtained through sorting is the parameter that cannot be converted or is the most difficult to convert or has the lowest expected conversion value;

[0076] S403: Sequentially select a parameter with a low parameter adjustment coefficient from the sequence of parameters to be adjusted as the first adjustable parameter;

[0077] S404: According to the optional values of the first adjustable parameter, divide the data set into several subsets, and sequentially derive corresponding child nodes from the nodes of each root node group;

[0078] S405: Replace the first adjustable parameter, and execute step S404 in each of the subsets until each adjustable parameter in the sequence of adjustable parameters is traversed or there is only one state in the subset or the features of all samples in the subset are the same, and finally generate a tree;

[0079] S406: Traverse all branches of the tree generated in step S405 and compare with the branch clustering described in step S401;

[0080] S407: Judge whether the branch clustering is compatible. If it is compatible, keep it; otherwise, prune it. The judgment criterion is that the parameters and optional values of the leaf nodes are both the same, and the parameters of the other nodes are the same;

[0081] S408: Prune and optimize the conversion tree.

[0082] S409: Traverse all branches of the conversion tree, and calculate the expected conversion rate of the leaf nodes according to the parameters and their optional values of the branch nodes.

[0083] Further, as Figure 3 , in step S5, the traversal includes:

[0084] Format user data (e.g., male aged 35 - 40, bedtime after 23:30, average daily step count less than 3000 steps) into a target feature combination of parameters and their optional values.

[0085] S502: Traverse all branches of the conversion tree and generate a feature combination of parameters and their optional values for each branch;

[0086] S503: Compare the compatibility between the target feature combination and the current branch feature combination. If they are compatible, save the current branch to form a compatible branch list, and go to step S504. Otherwise, traverse the next branch. The compatibility judgment criterion is that the number of nodes is the same and the parameters and optional values are all consistent;

[0087] S504: Compare the expected conversion rate of the leaf node of the current branch with the conversion target. If it is greater than or equal to the conversion target, form a conversion plan. Otherwise, traverse the next branch;

[0088] S505: Save the conversion plan to a list to form a single-parameter conversion plan list;

[0089] S506: Traverse the compatible branch list formed in step S503;

[0090] S507: Traverse each compatible branch in reverse order;

[0091] S508: When traversing in reverse order, obtain the combination of parameters to be tested formed by the current node and all nodes before it, and remove the combinations with the number of parameters to be tested less than 2. For example, the combination of parameters to be tested is bedtime between 22:00 - 23:00 and average daily step count between 3000 steps - 5000 steps.

[0092] S509: Calculate the expected conversion rate of the combination of parameters to be tested. If the expected conversion rate is greater than or equal to the conversion target, form a conversion plan. Otherwise, traverse the next node and go to step S508;

[0093] S510: Save the conversion plan to a list to form a multi-parameter conversion plan list.

[0094] In a specific embodiment, the method is applied to a sub-healthy analysis database, with the goal of determining whether an object is in a healthy state. It is assumed that the expected conversion rates of sleep schedule, eating habits, and exercise frequency increase in sequence, and a conversion tree is constructed accordingly. Assume the user's data is as follows: aged 35 - 40, male, falling asleep after 23:30, with an average daily step count of less than 3000 steps, sedentary time exceeding 8 hours, waking up at 7:00, and meal times being 8:00 for breakfast, 13:00 for lunch, and 20:00 for dinner. The goal is to reduce the future risk of hypertension by adjusting adjustable parameters. First, format the user data into a target feature combination (S501), and traverse all branches of the conversion tree to generate the feature combination of each branch (S502). By comparing the compatibility between the target feature combination and the current branch feature combination (S503), filter out the compatible branches, for example: falling asleep between 22:00 - 23:00, step count of 3000 - 5000 steps, and sedentary time less than 6 hours. Then, calculate the expected conversion rate of each branch to generate single-parameter and multi-parameter conversion plans (S504 - S510). The single-parameter plan includes adjusting the falling asleep time to 22:00 - 23:00 (expected conversion rate of 70%) or increasing the step count to 3000 - 5000 steps (expected conversion rate of 65%). The multi-parameter plan includes simultaneously adjusting the falling asleep time to 22:00 - 23:00, increasing the step count to 3000 - 5000 steps (expected conversion rate of 75%), or further reducing the sedentary time to 6 - 8 hours (expected conversion rate of 80%). These plans significantly reduce the future risk of hypertension by optimizing the user's sleep schedule and exercise habits.

[0095] Further, before the step S5, it further includes:

[0096] Determine whether the number of conversion plans reaches a preset number; if so, directly execute step S5; otherwise, increase the parameters and / or the optional values of the adjustable parameters, and repeat steps S1 - S4.

[0097] Further, conduct a goodness evaluation on the generated plans. Since there are many indicators for judging whether a person is healthy, it is necessary to calculate the correlation degree of each measurement indicator in sequence. The formula for goodness is as follows:

[0098] O = α1k1 + α2k2 + α3k3 + α4k4 + α5k5 + α6k6

[0099] Where O represents goodness, α1, α2, α3, α4, α5, and α6 represent weights, k1 represents the correlation degree of the daily step count, k2 represents the correlation degree of the daily exercise time, k3 represents the correlation degree of the sedentary time, k4 represents the correlation degree of the daily calorie consumption, k5 represents the correlation degree of the sleep duration, and k6 represents the expected conversion rate.

[0100] Furthermore, the calculation formula for the expected conversion rate is as follows:

[0101]

[0102] R(A i ,A j ) = P(C|A i ,X) - p′(C|A j ,X)

[0103] Among them, for the parameter A, there are n optional values for the parameter A, and these values are represented as A1, A2, …, A n ; R(A i ,A j ) represents the conversion rate of A i converted to A j , and E(A) is the expected conversion rate of the parameter A; when two parameter values of A are converted with each other, an R value is obtained; in R(A i ,A j ), A i and A j are two different values of the parameter A; only the positive R needs to be calculated, that is, R(A i ,A j ) > 0, ΣΣ represents the summation over all possible combinations of A i and A j , and R(A i ,A j ) > 0 means only calculating the conversions where R is positive, and m represents the total number of all positive R values;

[0104] For R(A i ,A j ), in the decision tree, for a specific input feature X, reaching the leaf node A with the A value of A i , the probability of being classified as the expected classification C is denoted as P(C|A i ,X); after changing the value A i of the leaf node A to A j , for the same input feature X, the new probability of being classified as the expected classification C is denoted as p′(C|A j ,X). Furthermore, in step S1, the parameters include conventional physical parameters, conventional information parameters, special abstract parameters, and method codes.

[0105] For example, first obtain various data of the user, such as sleep time, sleep quality, heart rate, etc., so as to judge that their daily routine is a sub-healthy one. Now, it is necessary to use the constructed conversion tree algorithm to propose several healthy routine plans. For example, Plan 1 is to change the number of walking steps from 500 steps per day to 5000 steps per day, and the sedentary time from 9 hours per day to 7 hours per day. Plan 2 is to change the sleep time from 500 steps per day to 8000 steps per day, the sedentary time from 9 hours per day to 6 hours per day, and the daily exercise time is 3 hours. Sort the two plans by calculating their goodness, and select the optimal plan.

[0106] Calculate the goodness of each plan, and propose an optimal plan to the user. For example, it is recommended that the user walk 8000 steps per day, exercise 2 km per day, increase the sleep duration to 8 hours, reduce the sedentary time to 6 hours, increase the daily calorie consumption to 300 kcal, and exercise 2 hours per day, so that the user can be transformed from a sub-healthy state to a healthy state.

[0107] The present invention also provides a conversion smart bracelet for intelligently generating a method for transforming the sub-healthy state of the human body, which is used to implement the intelligent method for intelligently generating a method for transforming the sub-healthy state of the human body, including:

[0108] Parameter division module: Obtain several parameters in the sample, the optional values of each parameter, and the corresponding number of reaching the target and the number of not reaching the target, divide the several parameters into adjustable parameters and non-adjustable parameters, and set the conversion difficulty level;

[0109] Expected conversion rate calculation module: Calculate the expected conversion rate of each adjustable parameter according to the number of reaching the target and the number of not reaching the target corresponding to the optional values of each adjustable parameter;

[0110] Conversion tree construction module: Set the non-adjustable parameters as the root nodes, and construct a conversion tree according to the expected conversion rates and optional values of each adjustable parameter. Each leaf branch of the conversion tree corresponds to a plan;

[0111] Traverse conversion tree module: Input a plan, traverse its corresponding leaf branch, determine several conversion plans, and calculate the goodness of each conversion plan;

[0112] Goodness ranking module: Rank all conversion plans according to the goodness of the conversion plans to obtain a goodness sequence;

[0113] Optimal selection module: Use the conversion plans in the front several positions of the goodness sequence as conversion alternative strategies.

[0114] Here is an example to illustrate the intelligent conversion algorithm. For example, a user wears a smart bracelet, which collects various life data of the user, such as gender, age, daily walking steps, exercise distance, exercise duration, calorie consumption, heart rate, blood oxygen saturation, sleep duration, sleep quality, body temperature, hourly stress value, sedentary time, etc. When establishing the conversion tree, it has been judged through historical data that the user's physical condition is already in a sub-healthy state, specifically, the time of falling asleep is often 23:45, the average daily number of exercise steps is only 2,000 steps, the sedentary time is as long as 9 hours, the time of getting up is 6:30, and the three meals are eaten at 7:40, 12:40, and 19:50 respectively. At the same time, it is marked that the user is in sub-health and has a risk of hypertension. Then the parameters were divided and discretized, with age and gender classified as non-adjustable parameters, and bedtime and average daily exercise steps as adjustable parameters. They were scored according to the difficulty of adjustment. The average daily exercise steps were scored 8 points for being easy to adjust, and the moderate bedtime was scored 6 points. Continuous value parameters such as the average daily exercise steps and bedtime were also divided into intervals. After that, the database data was divided into training sets, test sets, and validation sets according to the discrete values ​​of the parameters and the health status labels to build a training model. It was found that in the current state, the possibility of his sub-health in the next year was 80%, and if the exercise steps were adjusted to 5000-6000 steps or the bedtime was adjusted to 22:00-23:00, the expected conversion rate could be reduced to 70% and 72% respectively. When constructing the conversion tree, a decision tree was constructed based on the sample data to obtain branch clustering, and the adjustment coefficients of each parameter were calculated and sorted. The data set was divided and nodes were derived with the wake-up time as the first adjustable parameter. After the parameters were continuously changed to generate the tree, pruning optimization was performed, and the expected conversion rate of each branch leaf node was calculated. When determining the conversion plan, the user data is formatted into a target feature combination, the tree branches are traversed to find compatible branches, and the conversion rates are compared to form single-parameter and multi-parameter conversion plans. The single-parameter plan has the possibility of sub-health reduced to 72% by adjusting the bedtime to 22:00-23:00, and the possibility of sub-health reduced to 70% by adjusting the average daily exercise steps to 5000-6000 steps; the multi-parameter plan is to adjust the bedtime, exercise steps and reduce the sedentary time to 6 hours at the same time, which can reduce the possibility to 60%. Finally, the top-ranked plan of the superiority sequence is tested. If it meets the expectation of reducing the possibility of sub-health, it will be output as an alternative strategy. If it does not meet the requirements, the results will be fed back and the samples will be collected for analysis again.

[0115] Table 1 Some sample data on human health

[0116]

[0117]

[0118] Those skilled in the art know that, in addition to implementing the systems, devices, and their respective modules provided by the present invention in the form of pure computer-readable program code, the method steps can be logically programmed to enable the systems, devices, and their respective modules provided by the present invention to be implemented in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. to achieve the same program. Therefore, the systems, devices, and their respective modules provided by the present invention can be considered as a kind of hardware components, and the modules included therein for implementing various programs can also be regarded as the structures within the hardware components; the modules for implementing various functions can also be regarded as either software programs for implementing methods or structures within hardware components.

[0119] The specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the above specific embodiments, and those skilled in the art can make various changes or modifications within the scope of the claims, which does not affect the essence of the present invention. Without conflict, the embodiments of the present application and the features in the embodiments can be combined with each other arbitrarily.

Claims

1. A strategy generation method for intelligent transformation of sub-health status based on smart wearable devices, characterized in that: include: Step S1: Acquire the user's physiological data and non-privacy data, wherein the physiological data includes at least one of heart rate, sleep duration, exercise steps, sedentary time, blood oxygen saturation, and blood pressure value, and the non-privacy data includes at least one of age, gender, and chronic disease label; Step S2: Divide the acquired data into adjustable parameters and non-adjustable parameters, and divide the continuous value parameters into discrete intervals; Step S3: construct a training set based on the discretized parameters and state labels, and train the data model to calculate the expected conversion rate of each adjustable parameter; Step S4: Taking the non-adjustable parameter as the root node, constructing a transformation tree according to the expected transformation rate and optional value of the adjustable parameter, and pruning and optimizing the transformation tree; Step S5: input user data and conversion target, traverse the conversion tree to generate a compatible branch list, calculate the goodness of each branch and sort it, and generate a goodness sequence; Step S6: Verify the transformation solutions of the first few presets in the goodness sequence, and output the solutions that meet the expectations as alternative strategies.

2. The strategy generation method for intelligent transformation of sub-health status based on smart wearable devices according to claim 1 is characterized in that: The specific steps of constructing the transformation tree in step S4 include: Step S401: construct a decision tree based on sample data and extract all branches and clusters; Step S402: Calculate the parameter adjustment coefficient according to the conversion difficulty score and the expected conversion rate to generate a sequence of parameters to be adjusted; Step S403: Selecting adjustable parameters in ascending order according to the adjustment coefficients, dividing the data set and generating child nodes; Step S404: traverse all adjustable parameters to generate a complete transformation tree, and remove redundant branches by pruning.

3. The strategy generation method for intelligent transformation of sub-health status based on smart wearable devices according to claim 1 is characterized in that: The calculation formula of the goodness in step S5 is: O=α1k1+α2k2+α3k3+α4k4+α5k5+α6k6 Among them, O represents goodness, α1, α2, α3, α4, α5 and α6 represent weights, k1 represents the correlation of walking steps, k2 represents the correlation of daily exercise time, k3 represents the correlation of sedentary time, k4 represents the correlation of daily calorie consumption, k5 represents the correlation of sleep duration, and k6 represents the expected conversion rate.

4. The strategy generation method for intelligent transformation of sub-health status based on smart wearable devices according to claim 3 is characterized in that: The calculation formula of the correlation degree of the walking steps in step S5 is as follows: Wherein, k1(x) represents the correlation of walking steps, ρ represents the extension distance calculation operation, D represents the extension position value calculation operation, x0 represents the optimal value of daily walking steps, x represents the walking steps, X represents the daily walking steps that meet the required range, X0 represents several ranges of smaller walking steps in X, represents X and the range of requirements that are met in a particular case, Indicates the selectable range of all walking steps.

5. The strategy generation method for intelligent transformation of sub-health status based on smart wearable devices according to claim 3 is characterized in that: The calculation formula of the expected conversion rate in step S5 is as follows: R(A i ,A j )=P(C|A i ,X)-p′(C|A j ,X) Among them, for parameter A, parameter A has n optional values, which are represented as A1, A2, ..., A n ; R(A i ,A j ) means A i Convert to A j E(A) is the expected conversion rate of parameter A; when two parameter values ​​of A are converted to each other, an R value is obtained; R(A i ,A j ), A i and A j are two different values ​​of parameter A; only the positive value R needs to be calculated, that is, R(A i ,A j )>0, ΣΣ means for all possible A i and A j The combination of R(A i ,A j )>0 means that only the transformations with positive R are counted, and m represents the total number of all positive R values; For R(A i ,A j ), in the decision tree, for a specific input feature X, when reaching the leaf node A, the value of A is A i , the probability of being classified as the expected classification C, denoted as P(C|A i ,X); Change the value of leaf node A i A j After that, for the same input feature X, the new probability of being classified as the expected category C is recorded as p′(C|A j ,X).

6. A strategy generation system for intelligent transformation of sub-health status based on smart wearable devices, characterized in that: include: Module M1: Acquire the user's physiological data and non-privacy data, wherein the physiological data includes at least one of heart rate, sleep duration, exercise steps, sedentary time, blood oxygen saturation, and blood pressure value, and the non-privacy data includes at least one of age, gender, and chronic disease label; Module M2: divides the acquired data into adjustable parameters and non-adjustable parameters, and divides the continuous value parameters into discrete intervals; Module M3: Construct a training set based on the discretized parameters and state labels, and train the data model to calculate the expected conversion rate of each adjustable parameter; Module M4: Taking the non-adjustable parameters as the root nodes, constructing a transformation tree according to the expected transformation rates and optional values ​​of the adjustable parameters, and pruning the transformation tree; Module M5: Input user data and conversion target, traverse the conversion tree to generate a list of compatible branches, calculate the goodness of each branch and sort them, and generate a goodness sequence; Module M6: Verify the transformation solutions of the first few presets in the goodness sequence, and output the solutions that meet the expectations as alternative strategies.

7. The strategy generation system for intelligent transformation of sub-health status based on smart wearable devices according to claim 6 is characterized in that: The specific steps of constructing the transformation tree in the module M4 include: Module M401: Build a decision tree based on sample data and extract all branches and clusters; Module M402: Calculate the parameter adjustment coefficient according to the conversion difficulty score and the expected conversion rate, and generate a sequence of parameters to be adjusted; Module M403: Select adjustable parameters in ascending order of adjustment coefficients, divide the data set and generate child nodes; Module M404: traverses all adjustable parameters to generate a complete transformation tree and removes redundant branches by pruning.

8. The strategy generation system for intelligent transformation of sub-health status based on smart wearable devices according to claim 6 is characterized in that: The calculation formula of the goodness in the module M5 is: O=α1k1+α2k2+α3k3+α4k4+α5k5+α6k6 Among them, O represents goodness, α1, α2, α3, α4, α5 and α6 represent weights, k1 represents the correlation of walking steps, k2 represents the correlation of daily exercise time, k3 represents the correlation of sedentary time, k4 represents the correlation of daily calorie consumption, k5 represents the correlation of sleep duration, and k6 represents the expected conversion rate.

9. The strategy generation system for intelligent transformation of sub-health status based on smart wearable devices according to claim 8 is characterized in that: The calculation formula of the correlation degree of the walking steps in the module M5 is as follows: Wherein, k1(x) represents the correlation of walking steps, ρ represents the extension distance calculation operation, D represents the extension position value calculation operation, x0 represents the optimal value of daily walking steps, x represents the walking steps, X represents the daily walking steps that meet the required range, X0 represents several ranges of smaller walking steps in X, represents X and the range of requirements that are met in a particular case, Indicates the selectable range of all walking steps.

10. The strategy generation system for intelligent transformation of sub-health status based on smart wearable devices according to claim 8, characterized in that: The calculation formula of the expected conversion rate in the module M5 is as follows: R(A i ,A j )=P(C|A i ,X)-p′(C|A j ,X) Among them, for parameter A, parameter A has n optional values, which are represented as A1, A2, ..., A n ; R(A i ,A j ) means A i Convert to A j E(A) is the expected conversion rate of parameter A; when two parameter values ​​of A are converted to each other, an R value is obtained; R(A i ,A j ), A i and A j are two different values ​​of parameter A; only the positive value R needs to be calculated, that is, R(A i ,A j )>0, ΣΣ means for all possible A i and A j The combination of R(A i ,A j )>0 means that only the transformations with positive R are counted, and m represents the total number of all positive R values; For R(A i ,A j ), in the decision tree, for a specific input feature X, when reaching the leaf node A, the value of A is A i , the probability of being classified as the expected classification C, denoted as P(C|A i ,X); Change the value of leaf node A i A j After that, for the same input feature X, the new probability of being classified as the expected category C is recorded as p′(C|A j ,X).

Images