Personalized exercise scheme recommendation method and device based on intelligent algorithm

Abstract

The embodiment of the invention provides a personalized exercise scheme recommendation method and device based on an intelligent algorithm. The method comprises the steps of obtaining health data of a target user and an exercise type set corresponding to the health data; acquiring the exercise demand of the target user, wherein the exercise demand comprises at least one of exercise type preference, exercise frequency per week and exercise duration per time; using a multi-target XGBoost-Pareto model to optimize and adjust the health data based on the exercise demand and the exercise type set to obtain a target exercise video scheme meeting a target constraint condition, the target constraint condition being related to at least one of health income, exercise risk and user exercise type preference; the target exercise video scheme is recommended to the target user, at least personalized exercise schemes can be given for different users, exercise preferences of the users are met, fitness benefits are improved, exercise injuries are reduced, the exercise effect is comprehensively improved, and the satisfaction degree of the users to the exercise recommendation scheme is improved.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of information processing, and particularly to a personalized exercise plan recommendation method and device based on intelligent algorithms. Background Art

[0002] With the rapid development of technology and the increasing life pressure, more and more people relieve the pressure in life by swiping various videos on mobile phones. People gradually neglect the importance of exercise, resulting in an increasing number of people who are slightly overweight or overweight. When a person's weight exceeds their standard weight and they lack exercise, it will cause more diseases in themselves, endangering human health. Therefore, everyone should attach importance to physical exercise to enhance physical fitness and manage weight. Based on this, China has proposed a national health action of "Weight Management Year", aiming to improve the national weight management awareness, popularize healthy lifestyles and improve abnormal weight conditions over a three-year period. However, in the prior art, most of the physical exercises carried out during weight management are randomly following the teaching videos of online fitness bloggers, so that the content of the exercises may not meet the actual physical needs of users, that is, it cannot meet the actual fitness benefits required. At the same time, without arranging various types of exercise teaching videos in an orderly manner, it will also greatly increase the exercise risk during the user's exercise. It can be seen that the prior art lacks personalized physical exercise methods for individuals, making people lack personal pertinence during physical exercise and resulting in poor exercise effects. Summary of the Invention

[0003] The purpose of the present invention is to provide at least a personalized exercise plan recommendation method and device based on intelligent algorithms, which can at least solve the technical problem of the lack of personalized physical exercise methods for individuals in the prior art, and can at least achieve giving personalized exercise plans for different users, while meeting the user's exercise preferences, improving fitness benefits and reducing exercise injuries, so as to comprehensively improve the exercise effect and enhance the user's satisfaction with the exercise recommendation plan.

[0004] To solve the above technical problems, at least one embodiment of the present application provides a personalized exercise plan recommendation method based on intelligent algorithms, including: obtaining the health data of a target user and a set of exercise types corresponding to the health data, where the health data includes current health data and expected health data; obtaining the exercise needs of the target user, where the exercise needs include at least one of exercise type preference, weekly exercise frequency, and duration of each exercise; using a multi-objective XGBoost-Pareto model to optimize and adjust the expected health data based on the exercise needs and the set of exercise types to obtain a target exercise video plan that meets the target constraint conditions, where the target constraint conditions are related to at least one of health benefits, exercise risks, and user exercise type preferences; and recommending the target exercise video plan to the target user.

[0005] At least one embodiment of the present application further provides a personalized exercise plan recommendation device based on intelligent algorithms, including: an acquisition module for obtaining the health data of a target user and a set of exercise types corresponding to the health data, where the health data includes current health data and expected health data; the acquisition module is further configured to obtain the exercise needs of the target user, where the exercise needs include at least one of exercise type preference, weekly exercise frequency, and duration of each exercise; an optimization module for using a multi-objective XGBoost-Pareto model to optimize and adjust the expected health data based on the exercise needs and the set of exercise types to obtain a target exercise video plan that meets the target constraint conditions, where the target constraint conditions are related to at least one of health benefits, exercise risks, and user exercise type preferences; and a recommendation module for recommending the target exercise video plan to the target user.

[0006] At least one embodiment of the present application further provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the above-mentioned personalized exercise plan recommendation method based on intelligent algorithms.

[0007] At least one embodiment of the present application further provides a computer-readable storage medium storing a computer program, where the computer program, when executed by a processor, implements the above-mentioned personalized exercise plan recommendation method based on intelligent algorithms.

[0008] The personalized exercise plan recommendation method based on intelligent algorithms provided by the embodiments of the present application includes: obtaining the health data of a target user and a set of exercise types corresponding to the health data; obtaining the exercise needs of the target user, where the exercise needs include at least one of exercise type preference, weekly exercise frequency, and duration of each exercise; using a multi-objective XGBoost-Pareto model to optimize and adjust the health data based on the exercise needs and the set of exercise types to obtain a target exercise video plan that meets the target constraint conditions, where the target constraint conditions are related to at least one of health benefits, exercise risks, and user exercise type preferences; and recommending the target exercise video plan to the target user. Based on a comprehensive evaluation of various health indicators of the target user, by optimizing and adjusting the health data, a target exercise video plan that can optimize the health data of the target user is determined, and this target exercise video is user-specific and more in line with the target user, so that the user can easily adhere to the exercise, effectively improving the exercise effect of the target user and enhancing the exercise experience of the user using the personalized exercise plan recommendation method based on intelligent algorithms proposed in the embodiments of the present application.

[0009] In some alternative embodiments, the multi-objective XGBoost-Pareto model includes a health benefit model, an exercise risk model, and a user compliance model. The step of using the multi-objective XGBoost-Pareto model to optimize and adjust the expected health data based on the exercise needs and the set of exercise types to obtain a target exercise video plan that meets the target constraint conditions includes: using the health benefit model to obtain an initial health benefit based on the health data; using the exercise risk model to obtain an initial exercise risk based on the set of exercise types; using the user compliance model to obtain an initial user compliance based on the exercise needs; and obtaining a target exercise video plan that meets the target constraint conditions by adjusting the initial health benefit, the initial exercise risk, and the initial user compliance. By comprehensively analyzing the initial health benefit, initial exercise risk, and user compliance determined by multiple models, a target exercise video plan that can meet the target constraint conditions is obtained, improving the user specificity and user favorability of the target exercise video plan to achieve a target exercise video plan that meets the user's exercise needs, improves the user's overall health level, and reduces the negative impact of exercise. It enhances the user's enthusiasm for exercise and improves the exercise effect and the improvement effect of health data after exercise.

[0010] In some alternative embodiments, obtaining the initial health benefits based on the health data by using the health benefits model includes: obtaining the post-exercise health data of the target user; obtaining the difference between the post-exercise health data and the health data; obtaining the health data weight for the health data; and using the health benefits model to obtain the initial health benefits based on the difference and the health data weight. By analyzing the health data of the target user before and after exercise to obtain the initial health benefits, the exercise effect of the target user after exercise can be better obtained, so as to better customize a suitable target exercise video plan for the target user based on the exercise effect of the target user subsequently.

[0011] In some alternative embodiments, obtaining the initial exercise risks based on the exercise type set by using the exercise risk model includes: for each exercise type in the exercise type set, obtaining the exercise intensity, joint load, and fatigue degree of the exercise type, so as to obtain a plurality of exercise intensities, a plurality of joint loads, and a plurality of fatigue degrees; obtaining the total exercise intensity based on the plurality of exercise intensities, obtaining the total joint load based on the plurality of joint loads, and obtaining the total fatigue degree based on the plurality of fatigue degrees; obtaining the preset exercise intensity weight, the preset joint load weight, and the preset fatigue degree weight; and using the exercise risk model to obtain the initial exercise risks based on the preset exercise intensity weight, the preset joint load weight, the preset fatigue degree weight, the total exercise intensity, the total joint load, and the total fatigue degree. By comprehensively analyzing the exercise intensity, joint load, and fatigue degree of the exercise types in the exercise type set, the negative impact of the exercise types in the exercise type set on the target user's exercise can be better determined, so as to select the exercise of the exercise type with a smaller negative impact on the target user's exercise subsequently.

[0012] In some optional embodiments, the target motion video scheme satisfying the target constraint condition is obtained by adjusting the initial health benefit, the initial motion risk and the initial user compliance, including: obtaining the target health benefit, target motion risk and target user compliance satisfying the target constraint condition by adjusting the initial health benefit, the initial motion risk and the initial user compliance; adjusting the health data based on the target health benefit to obtain target health data; adjusting the motion type set based on the target motion risk to obtain a target motion type set; adjusting the motion demand based on the target user compliance to obtain a target motion demand; determining the motion type weight corresponding to each motion type in the target motion type set based on the target health data; determining the target motion video scheme based on the motion type weight corresponding to each motion type in the target motion type set, the target motion demand, the target health benefit, the target motion risk, the target user compliance and the target health data. Through the target constraints, we can determine the target exercise video plan that can improve the health benefits of target users and make the target users' health data reach the optimal standard after exercise, while reducing the risk of target users' exercise, meeting the target users' exercise needs and enabling them to persist in exercise, thereby increasing the target users' enthusiasm for long-term exercise, effectively improving the target users' health data and improving the exercise effect.

[0013] In some optional embodiments, the method of determining the motion type weight corresponding to each motion type in the target motion type set based on the target health data includes: for each target motion type in the target motion type set, determining the initial weight and adjustment value of the target motion type based on the health data, the target health data and preset rules, thereby obtaining multiple initial weights and multiple adjustment values; determining the motion type weight corresponding to each motion type in the target motion type set based on the multiple adjustment values and the multiple initial weights. The initial weight of the target motion type can be scientifically determined through preset rules, and the health data range that needs to be improved for the target motion can be determined in combination with the target health data and health data, thereby determining the adjustment value of the target motion type that needs to be adjusted, and adjusting the initial weight based on the adjustment value, thereby determining the motion type weight of the target motion type, so as to tailor an accurate target motion video solution for the target user based on the motion type weight.

[0014] In some alternative embodiments, a Markov decision model is constructed based on the health data, the set of exercise types, the health benefits, the exercise risks, and the user's exercise type preferences; during the process of optimizing and adjusting the expected health data, a double Q-learning network is used to optimize the Markov decision model to obtain a target exercise plan that meets the target constraint conditions; based on the target exercise plan that meets the target constraint conditions, a target video plan that meets the target constraint conditions is obtained. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] One or more embodiments are exemplarily illustrated by the pictures in the corresponding drawings, and these exemplary illustrations do not constitute a limitation on the embodiments.

[0016] Figure 1 is a schematic flowchart of a personalized exercise plan recommendation method based on an intelligent algorithm provided by an embodiment of the present application;

[0017] Figure 2 is a schematic structural diagram of data acquisition provided by an embodiment of the present application;

[0018] Figure 3 is a schematic structural diagram of data processing provided by an embodiment of the present application;

[0019] Figure 4 is a schematic structural diagram of data display provided by an embodiment of the present application;

[0020] Figure 5 is a schematic structural diagram of a personalized exercise plan recommendation device based on an intelligent algorithm provided by another embodiment of the present application;

[0021] Figure 6 is a schematic structural diagram of an electronic device provided by another embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below with reference to the accompanying drawings. However, those of ordinary skill in the art can understand that in the embodiments of the present application, many technical details are proposed for the readers to better understand the present application. However, even without these technical details and various changes and modifications based on the following embodiments, the technical solutions required to be protected by the present application can be implemented. The following division of each embodiment is for convenience of description and should not constitute any limitation on the specific implementation manner of the present application. Each embodiment can be combined and cross-referenced with each other on the premise of not conflicting.

[0023] It should be noted that the acquisition or use of the data in the embodiments of this application requires user consent. Relevant data can only be obtained after the user authorizes and permits it, and the acquisition or use of the data complies with the provisions of relevant laws and regulations.

[0024] To facilitate the understanding of this solution, some terms in this solution are explained here:

[0025] BMI (Body Mass Index) is the body mass index: It is a commonly used international standard for measuring the degree of a person's fatness and thinness and whether they are healthy.

[0026] The total FMS score: refers to the total score of the "Functional Movement Screen" (abbreviated as FMS). FMS is a screening tool used to evaluate an individual's basic movement ability and movement control. It evaluates an individual's flexibility, stability, and symmetry when performing daily and sports movements by testing seven basic movement patterns.

[0027] XGBoost-Pareto is a technology that combines the XGBoost algorithm with the concept of Pareto optimization, mainly used to handle multi-objective optimization problems.

[0028] Pareto optimization: In multi-objective optimization problems, since there are often conflicts between objectives, it is difficult to find a solution that optimizes all objectives. Pareto optimization aims to find a set of non-dominated solutions. These solutions may not be optimal in some objectives, but they are non-inferior in all objectives, that is, there are no other solutions that are superior to them in all objectives.

[0029] Pareto front: During the training process, XGBoost-Pareto will maintain a Pareto front to store the currently found non-dominated solutions. As the training progresses, new non-dominated solutions may be added to the Pareto front or replace the original inferior solutions.

[0030] To facilitate the understanding of the embodiments of this application, the relevant content of the personalized exercise plan recommendation method based on intelligent algorithms is introduced here first.

[0031] With the rapid development of technology and the increasing life pressure, more and more people relieve the stress in life by swiping various videos on their mobile phones. People gradually neglect the importance of exercise, resulting in an increasing number of people who are slightly overweight or overweight. When a person's weight exceeds their standard weight and they lack exercise, they will develop more diseases, which will endanger human health. Therefore, everyone should attach importance to physical exercise to enhance physical fitness and manage weight. However, in the existing technology, most of the physical exercises carried out by people when managing their weight are following the teaching videos of online fitness bloggers. There is a lack of personalized physical exercise methods for individuals in the existing technology, making people lack personal pertinence when doing physical exercise, with poor exercise effects and difficulty in effectively enhancing physical fitness and managing weight.

[0032] To solve the above technical problems of the lack of personalized physical exercise methods for individuals, resulting in lack of personal pertinence when people do physical exercise, poor exercise effects, and difficulty in effectively enhancing physical fitness and managing weight, the present invention proposes a personalized exercise plan recommendation method based on intelligent algorithms. The following specifically describes the implementation details of the personalized exercise plan recommendation method based on intelligent algorithms in this embodiment. The following content is only the implementation details provided for easy understanding and is not necessary for implementing this solution.

[0033] Embodiment 1:

[0034] The personalized exercise plan recommendation method based on intelligent algorithms in this embodiment can be applied to an electronic device with communication, computing, and data storage capabilities. Its specific process can be as Figure 1 shown, including:

[0035] Step 101, obtain the health data of the target user and the set of exercise types corresponding to the health data. The health data includes current health data and expected health data.

[0036] Specifically, the health data includes at least one of the physiological index data, physical fitness assessment data, quality of life assessment data, and activity level assessment data of the target user.

[0037] Specifically, the physiological index data is various index data used to measure the physiological health of the target user. For example, the physiological index data may include at least one of data such as BMI, blood pressure, fasting glucose, serum creatine kinase, and lactate dehydrogenase.

[0038] Specifically, the physical fitness assessment data is various index data used to measure the body adaptation ability of the target user. For example, the physical fitness assessment data may include at least one of data such as the total FMS score, grip strength, vital capacity, sit-and-reach, and 1-minute sit-ups.

[0039] Specifically, the quality of life assessment data is used to represent the performance of the target user in all aspects of life. For example, the quality of life assessment data includes at least one of the data such as physiological function, physiological function, body pain, general health status, etc. Specifically, the quality of life assessment data of the target user can be obtained based on the SF-36 scale.

[0040] Specifically, the activity level data is used to measure the intensity and time of the target user's daily activities. For example, the activity level data can include at least one of the data such as high-intensity, medium-intensity, low-intensity activity time, walking time, sedentary time, etc.

[0041] In some examples, the aforementioned current health data is obtained by inputting user index data such as various initial physiological index data, initial living habit data, initial physical fitness assessment data, and initial quality assessment data related to the user into the feature engineering, extracting health data features from the user indexes by using the feature engineering, and performing preprocessing. The expected health data is obtained by reverse calculation based on the user's current health data and exercise goals.

[0042] Exemplarily, for how to extract health data from the user index data by using the feature engineering, refer to the code in Table 1 below:

[0043] Table 1, Code Display Table

[0044]

[0045] Specifically, the exercise type set includes multiple exercise types. The exercise types are divided based on the purpose and intensity of the exercise. The exercise types can include at least one of the data such as aerobic exercise type, strength training type, functional training, high-intensity strength training, high-intensity activity, low-intensity activity, sedentary time recommendation, etc.

[0046] In some examples, the obtaining of the health data of the target user and the exercise type set corresponding to the health data includes:

[0047] If the health data is one, the exercise type set is one; if the health data is multiple, the exercise type set is multiple. Each health data corresponds to an exercise type set, and the exercise types in the exercise type set can be one or more.

[0048] Step 102, obtain the exercise needs of the target user, where the exercise needs include at least one of exercise type preference, weekly exercise frequency, and duration of each exercise.

[0049] Specifically, the exercise type preference is used to represent the exercise types liked by the target user, and the exercise types in the exercise type preference can be one or more.

[0050] Step 103: Using the multi-objective XGBoost-Pareto model, based on the exercise requirements and the set of exercise types, optimize and adjust the health data to obtain a target exercise video plan that meets the target constraint conditions, where the target constraint conditions are related to at least one of health benefits, exercise risks, and user exercise type preferences.

[0051] Specifically, the multi-objective XGBoost-Pareto model is used to optimize multiple objectives. By defining multiple objective functions to represent multiple objectives, and using Pareto front analysis to balance the results of multiple objective functions to balance multiple objectives, the optimization of multiple objectives is achieved.

[0052] Exemplarily, by defining multiple objective functions, the objective 1 is represented by maximizing the health benefit function, which is used to obtain the health benefits of the target user; the objective 2 is represented by minimizing the exercise risk function, which is used to obtain the exercise risks of the target user; and the objective 3 is represented by maximizing the user compliance function, which is used to obtain the compliance of the target user. Among them, objective 1 is used to improve the health data of the target user so that the improved health data reaches the standard level. The standard level of health data can refer to the standard data or standard data range determined for each item of health data in the prior art. For example, objective 1 can be used to improve BMI, blood pressure, blood sugar, etc. Objective 2 is used to reduce the exercise risks of the target user, such as avoiding muscle loss and avoiding excessive joint load. Objective 3 is used to improve the satisfaction and persistence of the target user with respect to exercise.

[0053] In some examples, in the aforementioned step 103, the multi-objective XGBoost-Pareto model includes a health benefit model, an exercise risk model, and a user compliance model. The step of using the multi-objective XGBoost-Pareto model to optimize and adjust the health data based on the exercise requirements and the set of exercise types to obtain a target exercise video plan that meets the target constraint conditions includes the following steps 1031 - 1034:

[0054] Step 1031: Using the health benefit model, obtain the initial health benefits based on the health data;

[0055] In some examples, in the aforementioned step 1031, the step of using the health benefit model to obtain the initial health benefits based on the health data includes: obtaining the exercise health data after the target user exercises; obtaining the difference between the exercise health data and the health data; obtaining the health data weight for the health data; and using the health benefit model to obtain the initial health benefits based on the difference and the health data weight.

[0056] Specifically, the health data weight is used to represent the importance degree of health data.

[0057] Specifically, the health benefit model is trained based on the aforementioned maximization health benefit function and is used to obtain the health benefit of the target user. The exercise risk model is trained with the aforementioned minimization exercise risk function and is used to obtain the exercise risk of the target user. The user compliance model is trained with the aforementioned maximization user compliance function and is used to obtain the compliance of the target user.

[0058] Among them, for the code implementation of obtaining the health benefit model, please refer to Table 2:

[0059] Table 2, Code Table for Training the Model

[0060]

[0061] Among them, the code implementation for obtaining the exercise risk model and the user compliance model is similar to that of the health benefit model. For the code implementation process of the health benefit model in Table 2, please refer to it, and details will not be elaborated here.

[0062] In some examples, obtaining the post-exercise health data of the target user includes: by obtaining the historical health data of the target user after exercise and using the historical health data as the post-exercise health data of the target user; if the historical health data of the target user after exercise is not obtained, the post-exercise health data of the same-type user of the target user can also be obtained and used as the post-exercise health data of the target user, or the post-exercise health data of the target user can be predicted through prediction. For the prediction method, please refer to the prior art and various feasible prediction means can be used for prediction, and details will not be elaborated here.

[0063] In some examples, taking BMI, blood pressure, fasting glucose, FMS score, and sedentary time as examples, it is illustrated how to use the health benefit model to obtain the initial health benefit based on the difference value and the health data weight. Specifically, please refer to the following formula:

[0064] Initial health benefit = w1×△BMI + w2×△blood pressure + w3×△fasting glucose + w4×△FMS score - w5×△sedentary time

[0065] Among them, △ is used to represent the change value of the health data of the target user before and after exercise. w1, w2, w3, w4, and w5 are the health data weights corresponding to BMI, blood pressure, fasting glucose, FMS score, and sedentary time respectively. Among them, the health data weight is used to reflect the importance of the corresponding health data.

[0066] It should be understood that during the initial optimization process, △ is the expected health data change value obtained based on the user's current health data and the expected health data.

[0067] Step 1032: Using the exercise risk model, obtain the initial exercise risk based on the exercise type set.

[0068] In some examples, in the aforementioned step 1032, the obtaining of the initial exercise risk using the exercise risk model based on the exercise type set includes: for each exercise type in the exercise type set, obtaining the exercise intensity, joint load, and fatigue degree of the exercise type, so as to obtain multiple exercise intensities, multiple joint loads, and multiple fatigue degrees; obtaining the total exercise intensity based on the multiple exercise intensities, obtaining the total joint load based on the multiple joint loads, and obtaining the total fatigue degree based on the multiple fatigue degrees; obtaining the preset exercise intensity weight, preset joint load weight, and preset fatigue degree weight; using the exercise risk model, based on the preset exercise intensity weight, the preset joint load weight, the preset fatigue degree weight, the total exercise intensity, the total joint load, and the total fatigue degree, to obtain the initial exercise risk.

[0069] In some examples, through a preset calculation method, the multiple exercise intensities can be calculated and processed to obtain the total exercise intensity, where the preset calculation method is any one of the arithmetic mean, weighted mean, geometric mean, harmonic mean, or other methods of calculating the mean in the prior art.

[0070] In some examples, the multiple joint loads can be calculated and processed through the preset calculation method to obtain the total joint load, and the multiple fatigue degrees can be calculated and processed through the preset calculation method to obtain the total fatigue degree.

[0071] In some examples, using the exercise risk model, based on the preset exercise intensity weight, the preset joint load weight, the preset fatigue degree weight, the total exercise intensity, the total joint load, and the total fatigue degree, to obtain the initial exercise risk, can refer to the following formula:

[0072] Exercise risk = w6 × total joint load + w7 × total exercise intensity + w8 × total fatigue degree

[0073] Wherein, w6 is the preset joint load weight, w7 is the preset exercise weight, and w8 is the preset fatigue degree weight.

[0074] Step 1033: Using the user compliance model, obtain the initial user compliance based on the exercise requirements.

[0075] In some examples, in the foregoing step 1033, obtaining the initial user compliance based on the exercise requirements by using the user compliance model includes:

[0076] Obtaining a preset preference weight for the exercise type preference in the exercise requirements, obtaining a preset frequency weight for the weekly exercise frequency in the exercise requirements, and obtaining a preset duration weight for the duration of each exercise in the exercise requirements;

[0077] Using the user compliance model, determining the initial user compliance based on the exercise requirements, the preset preference weight, the preset frequency weight, and the preset duration weight, where the initial user compliance is used to represent the satisfaction degree and persistence degree of the target user.

[0078] In some examples, using the user compliance model to determine the initial user compliance based on the exercise requirements, the preset preference weight, the preset frequency weight, and the preset duration weight, the following formula can be referred to:

[0079] User compliance = w9 × duration of each exercise + w10 × weekly exercise frequency + w11 × exercise type preference

[0080] Where w9 is the preset duration weight, w10 is the preset frequency weight, and w11 is the preset preference weight. The duration of each exercise is used to represent the duration of each exercise of the target user. The weekly exercise frequency is the number of times the target user exercises per week. The exercise type preference is used to represent the preference of the target user for certain exercise types (such as aerobic, strength training, etc.), which will affect whether the target user is willing to persist in the long term.

[0081] Step 1034, obtaining a target exercise video plan that meets the target constraint conditions by adjusting the initial health benefit, the initial exercise risk, and the initial user compliance.

[0082] In some examples, in the foregoing step 1034, obtaining a target exercise video plan that meets the target constraint conditions by adjusting the initial health benefit, the initial exercise risk, and the initial user compliance includes: obtaining a target health benefit, a target exercise risk, and a target user compliance that meet the target constraint conditions by adjusting the initial health benefit, the initial exercise risk, and the initial user compliance; adjusting the health data based on the target health benefit to obtain target health data; adjusting the exercise type set based on the target exercise risk to obtain a target exercise type set; adjusting the exercise demand based on the target user compliance to obtain a target exercise demand; determining the exercise type weights corresponding to each exercise type in the target exercise type set based on the target health data; and determining a target exercise video plan based on the exercise type weights corresponding to each exercise type in the target exercise type set, the target exercise demand, the target health benefit, the target exercise risk, the target user compliance, and the target health data.

[0083] In some examples, the foregoing obtaining a target health benefit, a target exercise risk, and a target user compliance that meet the target constraint conditions by adjusting the initial health benefit, the initial exercise risk, and the initial user compliance includes: obtaining an initial total utility based on the initial health benefit, the initial exercise risk, and the initial user compliance; obtaining a target total utility, and adjusting the initial health benefit, the initial exercise risk, and the initial user compliance so that the adjusted initial total utility is equal to the target total utility, and taking the initial health benefit when the adjusted initial total utility is equal to the target total utility as the target health benefit; taking the initial exercise risk when the adjusted initial total utility is equal to the target total utility as the target exercise risk; taking the initial user compliance when the adjusted initial total utility is equal to the target total utility as the target user compliance; wherein, when the initial total utility is equal to the target total utility, it is considered that the target constraint conditions are met.

[0084] It should be understood that each function (objective) aims to formulate a better personalized exercise plan for the target user. By combining the result values of multiple functions, various factors can be better integrated to formulate an excellent personalized exercise plan for the target user. In order to optimize multiple objectives more efficiently, this plan conducts model training on multiple objectives to obtain a multi-objective XGBoost-Pareto model. For example, for each function |(objective), XGBoost is used for model training to obtain prediction results under different exercise plans, and a regression model or classification model is constructed based on the relevant features of each objective in the training data to optimize the effect of each objective to the best. Exemplarily, for the initial health benefit, XGBoost will output an optimal plan by analyzing features such as different exercise intensities and frequencies. For the initial exercise risk, XGBoost will predict and minimize the possible exercise risk according to the exercise type and intensity. For user compliance, XGBoost will combine the user's historical behavior data to find an exercise plan that can improve user compliance. Finally, by adjusting the initial health benefit, initial exercise risk, and initial user compliance, an optimal solution set that meets the target total utility is obtained. The optimal solution set includes the target health benefit, target exercise risk, and target user compliance. Through XGBoost model training, the optimal solution set under each function is obtained. Each solution represents an exercise plan that achieves the optimal effect on one objective.

[0085] It should be understood that there are conflicting objectives among multiple objectives. For example, increasing the exercise intensity may increase the exercise risk, that is, an increase in the initial health benefit may lead to an increase in the initial exercise risk, resulting in optimization conflicts. Therefore, Pareto front analysis is used to find an optimal balance solution. Through Pareto front analysis, the non-dominated solutions are screened out from the optimal solution set obtained from the multi-objective XGBoost-Pareto model. When the Pareto front analysis result is obtained, a balance solution is selected. Specifically, the balance solution can be obtained by the weighted method, the user preference-based selection method, or maximizing the total utility.

[0086] Among them, the weighted method is to assign a weight to each objective (for example, 40% weight to the initial health benefit, 30% weight to the initial exercise risk, and 30% weight to the initial user compliance), and weight the performance of each solution according to the weight to select the optimal balance solution.

[0087] The user preference-based selection method is that the system can obtain the user's personal preferences and select the balance solution that best meets the user's preferences according to the user's needs.

[0088] Maximizing the total utility is to define a total utility function, perform a weighted average of all objectives, and then select the solution with the maximum total utility. In this plan, maximizing the total utility is preferred.

[0089] Preferably, in some examples, obtaining the initial total utility based on the initial health benefit, the initial exercise risk, and the initial user compliance includes obtaining a health weighting coefficient for the initial health benefit, obtaining an exercise weighting coefficient for the initial exercise risk, and obtaining a compliance weighting coefficient for the initial user compliance; obtaining the initial total utility based on the health weighting coefficient, the exercise weighting coefficient, the compliance weighting coefficient, the initial health benefit, the initial exercise risk, and the initial user compliance.

[0090] Specifically, obtaining the initial total utility based on the health weighting coefficient, the exercise weighting coefficient, the compliance weighting coefficient, the initial health benefit, the initial exercise risk, and the initial user compliance can be expressed by the following formula:

[0091] Total utility = α1 × Health benefit - α2 × Exercise risk + α3 × User compliance

[0092] Where α1 is the health weighting coefficient, α2 is the exercise weighting coefficient, and α3 is the compliance weighting coefficient. α1, α2, and α3 are all weighting coefficients. Regarding the initial health benefit, the initial exercise risk, and the initial user compliance as different target values, α1, α2, and α3 are respectively used to determine the contribution degree of each target value to the target exercise video plan corresponding to the target. The selection of the weighting coefficient can be adjusted based on the needs, preferences, and health goals of the target user.

[0093] In summary, obtaining the target health index by optimizing the initial health data of the target user can improve the overall health level of the target user. For example, optimizing the initial health data such as BMI, blood pressure, fasting glucose, FMS score, sedentary time, etc. can improve the overall health level of the target user. By optimizing the initial exercise risk of the target user, selecting appropriate exercise intensity, weekly exercise frequency, and exercise type preference, the possible injuries or negative impacts during exercise can be reduced. For example, reducing joint load and setting appropriate fatigue levels. For example, for a specific type of exercise (such as strength training), it may increase the joint burden and the risk of injury, and at this time, it needs to be appropriately reduced. If the exercise intensity is too high, it may lead to over-fatigue and injury, especially in the user group with relatively weak physical fitness. Therefore, an appropriate exercise intensity needs to be set. In addition, determining an appropriate fatigue level is beneficial to avoiding overtraining. Among them, the fatigue level of the target user can be measured by evaluating the heart rate recovery time to avoid overtraining. By optimizing the initial user compliance of the target user and adjusting the plan according to the user's activity level and preference, it is possible to use an exercise plan that meets the exercise needs of the target user to increase the possibility of the target user's long-term adherence to exercise. Optimizing the target exercise video plan according to the user's exercise needs improves the user's satisfaction and persistence, and increases the possibility of the target user's long-term adherence. And the multi-objective XGBoost-Pareto model is used to optimize the weight allocation to balance the optimization of the initial health data, the initial exercise risk, and the initial user compliance, ensuring to find a balance among multiple objectives.

[0094] In some examples, determining the exercise type weights corresponding to each exercise type in the target exercise type set based on the target health data includes: for each target exercise type in the target exercise type set, determining the initial weight and the adjustment value of the target exercise type based on the health data, the target health data, and a preset rule, so as to obtain a plurality of initial weights and a plurality of adjustment values; based on the plurality of adjustment values and the plurality of initial weights, determining the exercise type weights corresponding to each exercise type in the target exercise type set.

[0095] In some examples, for each target motion type in the set of target motion types, determining the initial weight and adjustment value of the target motion type based on the health data, the target health data, and a preset rule includes: If the health data includes at least one piece of health data, and each piece of health data in the at least one piece of health data corresponds to a set of target motion types, then for each piece of health data in the at least one piece of health data, obtain the set of target motion types corresponding to the health data, for each target motion type in the set of target motion types, obtain the preset initial weight corresponding to the health data and the target motion type, obtain the preset adjustment value of the target motion type corresponding to the health data based on the preset rule, determine the initial weight based on the preset initial weight and the preset adjustment value, obtain the target health data corresponding to the health data, and obtain the adjustment value of the target motion type corresponding to the target health data based on the preset rule. Determining the motion type weight corresponding to each motion type in the set of target motion types based on the multiple adjustment values and the multiple initial weights includes: Obtain at least one initial weight and at least one adjustment value corresponding to the same target motion type in at least one set of target motion types corresponding to at least one piece of health data, calculate the at least one initial weight using the aforementioned preset calculation method to obtain the total initial weight; calculate the at least one adjustment value using the aforementioned preset calculation method to obtain the total adjustment value, sum the total initial weight and the total adjustment value, and use the sum value as the motion type weight corresponding to the target motion type, thereby obtaining the motion type weights corresponding to each target motion type in the at least one set of target motion types.

[0096] In some examples, the method of obtaining the adjustment value of the target motion type corresponding to the target health data based on the preset rule is the same as the method of obtaining the preset adjustment value of the target motion type corresponding to the health data based on the preset rule.

[0097] In some examples, the preset rules are used to assign weights of each target exercise type to each piece of health data. Specifically, the preset rules can be formulated by the expert experience method, the literature support method, the data-driven method, or other methods in the prior art. In this solution, the expert experience method is preferably used. Specifically, the expert experience method means that according to the experience of domain experts, preset initial weights are assigned to each target exercise type corresponding to each piece of health data, and preset adjustment values are assigned to each target exercise type according to different values corresponding to each piece of health data. For example, BMI and body fat percentage have a greater impact on weight management. Therefore, higher preset initial weights are assigned to each target exercise type corresponding to BMI and body fat percentage. The literature support method is used to determine the relative importance of the impact of each piece of health data on health based on referring to relevant literature or research. Thereby, preset initial weights corresponding to each target exercise type for each piece of health data are determined, and preset adjustment values are assigned to each target exercise type according to different values corresponding to each piece of health data. For example, blood pressure and blood sugar have a greater impact on cardiovascular health. Therefore, higher preset initial weights are assigned to each target exercise type corresponding to blood pressure and blood sugar. The data-driven method is used to extract weights from data using statistical methods (such as regression analysis, principal component analysis, etc.). For example, through multiple linear regression analysis, the contribution degree of each piece of health data to the health score is determined. Thereby, preset initial weights corresponding to each target exercise type for each piece of health data are determined, and preset adjustment values are assigned to each target exercise type according to different values corresponding to each piece of health data.

[0098] Exemplarily, according to the health data given by the user, an expert rule decision tree is constructed based on the expert experience method to initialize the weight assignment of the exercise plan. To more intuitively obtain the initial weights and preset adjustment values of the target exercise types corresponding to each piece of health data, refer to Table 3:

[0099] Table 3, Preset Rule Table

[0100]

[0101] Exemplarily, taking the data-driven method as an example, the preset initial weights and adjustment values of the target exercise types corresponding to each piece of health data are described, and an example is given to illustrate how to obtain the initial weights.

[0102] ① The preset initial weights of each target exercise type corresponding to BMI are respectively: the preset initial weight corresponding to the aerobic exercise type = 0.5, the preset initial weight corresponding to the strength training type = 0.3, and the preset initial weight corresponding to the high-intensity training type = 0.2.

[0103] BMI > 30 (obese):

[0104] Adjustment value:

[0105] Aerobic exercise: +0.15

[0106] Strength training: -0.10

[0107] High-intensity training: No adjustment

[0108] Total adjusted weight:

[0109] 0.65 + 0.20 + 0.20 = 1.05

[0110]

[0111] BMI 25 - 30 (Overweight):

[0112] Adjustment value:

[0113] Aerobic exercise: +0.10

[0114] Strength training: +0.25

[0115] High-intensity training: No adjustment

[0116] Total adjusted weight:

[0117] 0.60 + 0.55 + 0.20 = 1.35

[0118]

[0119] BMI < 18.5 (Underweight):

[0120] Adjustment value:

[0121] Aerobic exercise: -0.05

[0122] Strength training: +0.10

[0123] High-intensity training: No adjustment

[0124] Total adjusted weight:

[0125] 0.45 + 0.40 + 0.20 = 1.05

[0126]

[0127] ② The preset initial weights of the respective target exercise types corresponding to blood pressure are: The preset initial weight corresponding to the low-intensity aerobic exercise type = 0.3, the preset initial weight corresponding to the high-intensity exercise type = 0.5, and the preset initial weight corresponding to the moderate-intensity aerobic exercise type = 0.2.

[0128] For different blood pressure ranges, the method for obtaining the initial weights can be referred to the following formula:

[0129] Blood pressure > 140 / 90 (Hypertension):

[0130] Adjustment value:

[0131] Low-intensity aerobic: +0.10

[0132] High-intensity exercise: -0.15

[0133] Moderate-intensity aerobic exercise: No adjustment

[0134] Total adjusted weight:

[0135] 0.40 + 0.35 + 0.20 = 0.95

[0136]

[0137] Blood pressure 120 - 140 / 80 - 90 (borderline high):

[0138] Adjustment value:

[0139] Low-intensity aerobic: No adjustment

[0140] High-intensity exercise: -0.05

[0141] Moderate-intensity aerobic: +0.10

[0142] Total adjusted weight:

[0143] 0.3 + 0.45 + 0.30 = 1.05

[0144]

[0145] Blood pressure <120 / 80 (normal):

[0146] Adjustment value:

[0147] Low-intensity aerobic: No adjustment

[0148] High-intensity exercise: No adjustment

[0149] Moderate-intensity aerobic: No adjustment

[0150] Total weight:

[0151] w′ 低强度 = 0.30

[0152] w′ 高强度 = 0.50

[0153] w′ 中等强度 = 0.20

[0154] ③ The preset initial weights of each target exercise type corresponding to fasting glucose are: the preset initial weight corresponding to the aerobic exercise type = 0.5, and the preset initial weight corresponding to the strength training type = 0.3.

[0155] For different fasting glucose ranges, the method for obtaining the initial weights can be seen in the following formula:

[0156] Fasting glucose > 7.0 mmol / L (diabetes risk):

[0157] Adjustment value:

[0158] Aerobic exercise: +0.15

[0159] Strength training: -0.10

[0160] Total adjusted weight:

[0161] 0.65 + 0.20 = 0.85

[0162]

[0163] Fasting glucose 5.6 - 7.0 mmol / L (prediabetes):

[0164] Adjustment value:

[0165] Aerobic exercise: +0.10

[0166] Strength training: -0.05

[0167] Total adjusted weight:

[0168] 0.60 + 0.25 = 0.85

[0169]

[0170] Fasting glucose < 5.6 mmol / L (normal):

[0171] Adjustment value:

[0172] Aerobic exercise: no adjustment

[0173] Strength training: no adjustment

[0174] Adding other exercise types: +0.2

[0175] Total adjusted weight:

[0176]

[0177] w′ 其他 = 0.20

[0178] ④ The preset initial weights for each target exercise type corresponding to FMS are: the preset initial weight for the functional training type = 0.4, and the preset initial weight for the high-intensity strength training type = 0.6.

[0179] For different FMS total score ranges, the method for obtaining the initial weights can be seen in the following formulas:

[0180] FMS total score < 14 (motor dysfunction):

[0181] Adjustment value:

[0182] Functional training: +0.20

[0183] High-intensity strength training: -0.15

[0184] Total adjusted weight:

[0185] 0.60 + 0.45 = 1.05

[0186]

[0187] FMS total score 14 - 20 (moderate functional level):

[0188] Adjustment value:

[0189] Functional training: +0.10

[0190] High-intensity strength training: -0.05

[0191] Total adjusted weight:

[0192] 0.50 + 0.55 = 1.05

[0193]

[0194] FMS total score > 20 (good functional level):

[0195] Adjustment value:

[0196] Functional training: no adjustment

[0197] High-intensity strength training: +0.10

[0198] Total adjusted weight:

[0199] 0.4 + 0.7 = 1.1

[0200]

[0201] ⑤ The preset initial weights of each target exercise type corresponding to sedentary time are: the preset initial weight of the low-intensity activity type = 0.7, and the preset initial weight of the sedentary time recommendation type = 0.3.

[0202] For different sedentary time ranges, the method for obtaining the initial weights can be seen in the following formulas:

[0203] Sedentary time > 8 hours / day:

[0204] Adjustment value:

[0205] Low-intensity activity: +0.15

[0206] Suggested sedentary time: -0.10

[0207] Total adjusted weight:

[0208] 0.85 + 0.20 = 1.05

[0209]

[0210] Sedentary time of 4 - 8 hours per day:

[0211] Adjustment value:

[0212] Low-intensity activity: +0.05

[0213] Suggested sedentary time: -0.05

[0214] Total adjusted weight:

[0215] 0.75 + 0.25 = 1.00

[0216] w′ 低强度 = 0.75

[0217] w′ 久坐 = 0.25

[0218] Sedentary time < 4 hours per day:

[0219] Adjustment value:

[0220] Low-intensity activity: No adjustment

[0221] Suggested sedentary time: No adjustment

[0222] Total weight:

[0223] w′ 低强度 = 0.70.

[0224] w′ 久坐 = 0.30.

[0225] In summary, through preset rules, relatively accurate preset initial weights can be obtained. Thus, by adjusting the weights based on the relatively accurate preset initial weights, the obtained initial weights are more accurate, and the adjustment values are also obtained based on the preset rules, which further ensures that the adjustment values of the obtained weight adjustments are relatively accurate, making the obtained initial weights and exercise type weights accurate, thereby improving the accuracy of obtaining the target exercise video plan.

[0226] In some examples, a Markov decision model is constructed based on the health data, the set of exercise types, the health benefits, the exercise risks, and the user's exercise type preferences; during the process of optimizing and adjusting the expected health data, a double Q-learning network is used to optimize the Markov decision model to obtain a target exercise plan that meets the target constraint conditions; based on the target exercise plan that meets the target constraint conditions, a target video plan that meets the target constraint conditions is obtained.

[0227] That is to say, in the embodiments of the present application, different-dimensional optimization and adjustment of the health data and the exercise type weights are realized through the XGBoost-Pareto model and the double Q-learning network, so as to obtain the index data of the maximum health benefit, the minimum exercise risk, and the maximum user compliance through the XGBoost-Pareto model, and on this basis, the double Q-learning network is used to optimize the Markov model constructed by using the expected health data, the health benefit, the exercise risk, and the user compliance, and a target exercise plan with the training effects of the maximum health benefit, the minimum exercise risk, and the maximum user compliance is obtained on the basis of meeting the index data of the maximum health benefit, the minimum exercise risk, and the maximum user compliance of the target user, and then a target exercise video plan is obtained based on the target exercise plan.

[0228] In some examples, to further optimize each objective to obtain an optimal solution set, this solution also introduces a reinforcement learning model. In reinforcement learning, the adjustment of the target motion video solution is modeled by using a Markov decision process (MDP). Specifically, in the Markov decision process (MDP) modeling process, the user's real-time physiological indicators (such as blood pressure, blood sugar, FMS score, etc.) are used as the state space (s), the motion type weights for adjusting the target motion type of the motion plan (such as the motion type weights corresponding to aerobic exercise, strength training, functional training, etc.) are used as the action space (a), and the reward calculated based on the target health benefit, target motion risk, and target user compliance is used as the reward function (r). The overestimation bias problem of traditional single Q-learning is solved by means of a double Q-learning network. For example, in traditional single Q-learning, there may be an overestimation of the return of a certain action due to incorrect action selection. The agent may choose actions with lower actual returns, resulting in low policy efficiency; or due to an unstable learning process, due to overestimation, the update of the Q value may not conform to the real return, thus affecting the stability of the entire learning process and making it difficult for the agent to converge to the optimal policy; or due to poor policy execution effect, such as overly optimistic estimation, the agent may execute a policy that does not match the actual effect and cannot effectively achieve the expected goal, etc. The stability and security of the weight adjustment cannot be guaranteed. Therefore, this solution is to ensure the stability and security of the adjustment of the motion type weights of the target motion type. To overcome the overestimation bias problem in single Q-learning, the double Q-network (Double Q-learning) introduces two independent Q-value estimations to reduce the overestimation problem. Specifically, the principle of the double Q-network is to set two independent Q-value functions. In traditional Q-learning, only one Q function is used to estimate the Q value of each state-action pair, while the double Q-network maintains two Q-value functions, usually denoted as Q1(s,a) and Q2(s,a), to separately estimate and update. When updating, one Q-network is used to select the current optimal action, while the other Q-network is used to calculate the target Q value to reduce the risk of overestimation by a single Q function.

[0229] Update formula:

[0230]

[0231] Among them, Q1(s,a) is the Q value of the first Q network in state s and action a. Q2(s′,a′) is the Q value of the second Q network in state s′ and action a′. α refers to the learning rate, which is used to control the size of the update step. γ is the discount factor, which is used to balance the importance of the current reward and the future reward. r refers to the immediate reward, which is the reward obtained when transferring from state s to state s′. s′ refers to the next state. argmaxQ1(s′,a′) refers to the action that selects the maximum Q value according to Q1 in state s′.

[0232] To understand the implementation process of the double Q network more clearly, refer to the code implementation in Table 4:

[0233] Table 4, Code implementation process of the double Q network

[0234]

[0235] In summary, by alternately using the double Q network, the overestimation of Q values can be effectively avoided. In this way, the double Q network ensures that the estimation of the target Q value in the update process is more conservative, thereby improving the stability of the learning process and enabling the intelligent agent to more accurately select the optimal action. It improves the accuracy of the motion type weight corresponding to the determined target motion type.

[0236] In some examples, using the double Q-learning method to determine a target exercise video plan based on the motion type weights corresponding to each motion type in the target motion type set, the target motion requirements, the target health benefits, the target motion risks, the target user compliance, and the target health data, including: using the double Q-learning method to determine the latest motion type weights corresponding to each motion type in each target motion type set based on the motion type weights corresponding to each motion type in the target motion type set, the target health benefits, the target motion risks, the target user compliance, and the target health data, and determining the target exercise video plan from a preset video library based on the latest motion type weights and the target motion requirements.

[0237] Specifically, recommend a personalized target exercise video plan for the target user according to the health data and target exercise requirements of the target user. Among them, the target exercise video plan includes a suitable fitness video selected for the target user from a preset video library. For example, if the personalized plan of the target user needs to increase aerobic exercise and is suitable for overweight people, then exercise videos that meet these conditions can be screened out from the video library. The system will adjust the personalized exercise plan according to the motion type weights obtained through the Markov decision process (MDP). The motion type weights reflect the priorities of different motion types, such as aerobic exercise, strength training, etc.

[0238] Exemplarily, the target motion requirements may include personal information of the target user, such as age, gender, etc. For example, in a specific state (such as the health condition of a certain user), the MDP calculates that the user needs to perform a certain type of training more (such as aerobic training). In this case, the system will generate a weight combination including different motion types. For example, assume the weight of aerobic exercise is 0.6 and the weight of strength training is 0.4. This means that the user should allocate 60% of the time to aerobic exercise and 40% of the time to strength training in the training plan at this stage. According to these weights, the system can select appropriate motion videos or programs. The difficulty of the motion program can be calculated based on the user's age to obtain their maximum heart rate. According to this maximum heart rate, determine the motion ranges of different intensities (low, medium, high intensity); based on the calculated motion intensity range (for example, low intensity corresponds to 50%-60% of the maximum heart rate), combined with the existing videos of different difficulties in the video library, select the videos that match this intensity range. For example, if the MDP recommends that the user needs high-intensity aerobic exercise and the corresponding intensity is 60%-75% of the maximum heart rate, the system will screen out suitable medium-intensity aerobic videos. The recommendation can also be adjusted according to the user's motion type preference. For example, if the user prefers low-intensity training, the system can preferentially recommend videos that meet this preference.

[0239] To understand the acquisition process of the motion videos in the target motion video program more clearly, refer to the code implementation process recorded in Table 5:

[0240] Table 5, Acquisition Code of Motion Videos

[0241]

[0242] In some examples, a preset video library can be established by creating a database or file system containing fitness videos. Each video in the video library contains a set of descriptive tags (metadata) to help match user requirements.

[0243] To understand how to create each video in the preset video library more clearly, refer to the code implementation process of creating a video illustrated in Table 6:

[0244] Table 6, Code Implementation Process of Creating a Video

[0245]

[0246] To understand each video in the preset video library intuitively, refer to Table 7:

[0247] Table 7, Preset Video Library

[0248]

[0249] Step 104: Recommend the target exercise video plan to the target user.

[0250] In some examples, the target exercise video plan includes physical condition analysis, exercise plan suggestions, and fitness video information. The physical condition analysis includes health risks (such as obesity, hypertension, pre-diabetes, etc.), physical fitness levels (such as total FMS score, grip strength, vital capacity, etc.), and activity levels (such as sedentary time, high-intensity activity time, etc.). The exercise plan suggestions include exercise types with corresponding intensities and times allocated according to exercise type weights. For example, aerobic exercise: such as brisk walking, swimming, cycling (adjust the intensity and time according to the exercise type weight allocation). The exercise plan suggestions also include functional training: such as squats, hurdle steps, straight-line lunges (adjust the difficulty according to the total FMS score). The exercise plan suggestions also include low-intensity activities: such as walking, stretching (adjust according to sedentary time and user preferences). The fitness video information includes video ID (the unique identifier of the fitness video), video title (video name), video category (select the exercise type of the video according to the needs of the target user), difficulty level (select the appropriate difficulty level according to the physical condition of the target user), duration (video duration (minutes)), goal (select the appropriate goal according to the physical condition analysis of the target user, such as building muscle, losing fat, improving flexibility, etc.), and applicable population (select a fitness video that matches the physical state or goal of the target user, such as overweight people, healthy people, the elderly, etc.).

[0251] Exemplarily, assume that the health data of the target user includes: BMI: 28 (overweight), blood pressure: 130 / 85 mmHg, fasting glucose: 6.2 mmol / L, total FMS score: 16, sedentary time: 6 hours / day. The analysis of the physical condition of the target user obtained includes BMI (28, overweight): According to the current BMI (28), it belongs to the overweight range. Overweight may increase the risk of health problems such as cardiovascular diseases and diabetes. Therefore, weight loss and improvement of body composition (reduction of body fat percentage) are the primary goals; blood pressure (130 / 85 mmHg, slightly elevated): The current blood pressure is 130 / 85 mmHg, which belongs to the slightly elevated range. Attention should be paid to controlling blood pressure to prevent it from rising to the hypertensive level. Appropriate aerobic exercise helps to improve blood pressure; fasting glucose (6.2 mmol / L, prediabetes): The fasting glucose is 6.2 mmol / L, which belongs to the prediabetes range. Attention should be paid to diet and exercise to avoid developing type 2 diabetes. Aerobic exercise and strength training help to improve insulin sensitivity; total FMS score (16, moderate functional level): The total FMS score is 16, indicating a moderate functional level. The user may have some functional deficiencies during exercise, especially in terms of stability and flexibility. Functional training is needed to improve the body's motor ability and stability; sedentary time (6 hours / day): The sedentary time is 6 hours / day. Prolonged sedentary time may have a negative impact on health and increase the risk of metabolic syndrome. It is recommended to reduce the sedentary time and increase the daily activity level. And based on the analysis of the physical condition of the target user, the target exercise needs, the set of target exercise types, and the exercise type weights corresponding to each exercise type, the corresponding exercise videos are matched to obtain the target exercise video plan. Among them, the target exercise video plan can be seen in Table 8, the target exercise video plan:

[0252] Table 8, Target Exercise Video Plan

[0253]

[0254]

[0255] In summary, the selected balanced solution is transformed into a personalized target exercise video plan, realizing the recommendation of a personalized exercise plan that suits the target user, improving the exercise effect after exercise for the target user, and enabling the user to more clearly obtain various data about themselves and more clearly understand their own physical condition, resulting in a better user experience and improving the user's sense of use.

[0256] In some examples, the data processing process of this solution is mainly handled by a data processing server. This solution also includes an information collection and classification processing server, as well as an information and data terminal display. Specifically, the information collection and classification processing server is used to collect health data and the exercise needs of users. For the content of the information collection and classification processing server, refer to Figure 2 , the exercise needs of users are collected through the basic information automatic collection module, such as collecting gender and age in the exercise needs, and the health data of users are collected through the personal physical examination data automatic collection module, such as collecting weight and body composition, chronic disease and metabolic screening, nutrition and diet health information, joke system health screening, and cardiovascular health and detection. And the physical fitness test data automatic collection module is used to collect the morphology, function and quality of the target user to obtain physical fitness assessment data, and various collected data are classified through automatic classification, and various types of data are preprocessed using data cleaning, and the cleaned data is integrated into the dynamic collection health information file through the data integration module, so as to store the health data and exercise needs of the target user collected.

[0257] The data processing server can obtain health data and exercise needs from the dynamic collection health information file in the information collection and classification server. Specifically, for the relevant content of the data processing server, refer to Figure 3 , the information processing and screening module obtains health data and exercise needs from the dynamic collection health information file, and sends the health data to the physiological index module, the physical fitness assessment module, the quality of life assessment module, and the activity level module respectively. By analyzing the physiological index module, the physical fitness assessment module, the quality of life assessment module, and the activity level module, and obtaining the corresponding exercise videos from the preset video library, a target exercise video plan is constructed. Among them, the physiological index module also analyzes the BMI of the target user based on the personal health weight atlas.

[0258] The information and data terminal display can display the target exercise video plan obtained from the data processing server, as well as various information related to the target user. Specifically, Figure 4 As shown, the terminal host obtains the target exercise video plan and various information related to the target user from the data processing server, and displays the population weight detection information list, the average body mass index level, the population physical fitness detection information list, and the average physical fitness level through the monitoring large screen. The target exercise video plan, the physiological indicators, physical fitness assessment, quality of life assessment, and activity level of the target user are displayed in the client APP.

[0259] In summary, the present application obtains the health data of the target user and the set of exercise types corresponding to the health data; obtains the exercise needs of the target user, where the exercise needs include at least one of exercise type preference, weekly exercise frequency, and duration of each exercise; uses the multi-objective XGBoost-Pareto model to optimize and adjust the health data based on the exercise needs and the set of exercise types to obtain a target exercise video plan that meets the target constraint conditions, where the target constraint conditions are related to at least one of health benefits, exercise risks, and user exercise type preferences; and recommends the target exercise video plan to the target user. Based on a comprehensive evaluation of various health indicators of the target user, by optimizing and adjusting the health data, a target exercise video plan that can optimize the health data of the target user is determined, and this target exercise video is user-specific and more in line with the target user, so that the user can easily adhere to the exercise, effectively improving the exercise effect of the target user. The effectiveness of enhancing physical fitness and managing weight is improved.

[0260] Specifically, this solution combines the DWAA-RA research framework to implement the complete process from data modeling to personalized recommendation. It uses health data modeling and weight initialization, determines the exercise type weights of the exercise types corresponding to the user's health data using expert rules, corrects the weights corresponding to the initial exercise types through multi-objective optimization processing, ensures the scientificity and personalization of the target exercise video plan, and obtains appropriate videos from the preset video library according to the user's personalized needs and the user's health data to enrich the target exercise video plan, providing the user with more information and video options, improving the user experience and the effect after the user exercises. It enables the user to better understand their own health status.

[0261] Embodiment Two:

[0262] Another embodiment of the present application relates to a personalized exercise plan recommendation device based on an intelligent algorithm. The implementation details of the personalized exercise plan recommendation device based on the intelligent algorithm in this embodiment will be specifically described below. The following content is only the implementation details provided for convenience of understanding and is not necessary for implementing this solution. The schematic diagram of the personalized exercise plan recommendation device based on the intelligent algorithm in this embodiment can be as Figure 5 shown, including an acquisition module 501, an optimization module 502, and a recommendation module 503.

[0263] The acquisition module 501 is used to acquire the health data of the target user and the set of exercise types corresponding to the health data;

[0264] The acquisition module 501 is further used to acquire the exercise needs of the target user, where the exercise needs include at least one of exercise type preference, weekly exercise frequency, and duration of each exercise;

[0265] An optimization module 502, configured to use a multi-objective XGBoost-Pareto model to optimize and adjust the health data based on the motion requirements and the set of motion types, so as to obtain a target motion video solution that meets the target constraint conditions, where the target constraint conditions are related to at least one of health benefits, motion risks, and user motion type preferences;

[0266] A recommendation module 503, configured to recommend the target motion video solution to the target user.

[0267] In some examples, the multi-objective XGBoost-Pareto model in the device includes a health benefit model, a motion risk model, and a user compliance model. When the device is configured to use the multi-objective XGBoost-Pareto model to optimize and adjust the health data based on the motion requirements and the set of motion types to obtain a target motion video solution that meets the target constraint conditions, it is specifically configured to: use the health benefit model to obtain an initial health benefit based on the health data; use the motion risk model to obtain an initial motion risk based on the set of motion types; use the user compliance model to obtain an initial user compliance based on the motion requirements; and obtain a target motion video solution that meets the target constraint conditions by adjusting the initial health benefit, the initial motion risk, and the initial user compliance.

[0268] In some examples, when the device is configured to use the health benefit model to obtain an initial health benefit based on the health data, it is specifically configured to: obtain the post-exercise health data of the target user; obtain the difference between the post-exercise health data and the health data; obtain the health data weight for the health data; and use the health benefit model to obtain the initial health benefit based on the difference and the health data weight.

[0269] In some examples, when the device is configured to use the motion risk model to obtain an initial motion risk based on the set of motion types, it is specifically configured to: for each motion type in the set of motion types, obtain the motion intensity, joint load, and fatigue degree of the motion type, so as to obtain a plurality of motion intensities, a plurality of joint loads, and a plurality of fatigue degrees; obtain the total motion intensity based on the plurality of motion intensities, obtain the total joint load based on the plurality of joint loads, and obtain the total fatigue degree based on the plurality of fatigue degrees; obtain a preset motion intensity weight, a preset joint load weight, and a preset fatigue degree weight; and use the motion risk model to obtain the initial motion risk based on the preset motion intensity weight, the preset joint load weight, the preset fatigue degree weight, the total motion intensity, the total joint load, and the total fatigue degree.

[0270] In some examples, when the device is used for the target exercise video plan that meets the target constraint conditions by adjusting the initial health benefit, the initial exercise risk, and the initial user compliance, it is specifically used for: by adjusting the initial health benefit, the initial exercise risk, and the initial user compliance, obtaining the target health benefit, the target exercise risk, and the target user compliance that meet the target constraint conditions; adjusting the health data based on the target health benefit to obtain the target health data; adjusting the set of exercise types based on the target exercise risk to obtain the target set of exercise types; adjusting the exercise requirements based on the target user compliance to obtain the target exercise requirements; determining the exercise type weights corresponding to each exercise type in the target set of exercise types based on the target health data; and determining the target exercise video plan based on the exercise type weights corresponding to each exercise type in the target set of exercise types, the target exercise requirements, the target health benefit, the target exercise risk, the target user compliance, and the target health data.

[0271] In some examples, when the device is used for determining the exercise type weights corresponding to each exercise type in the target set of exercise types based on the target health data, it is specifically used for: for each target exercise type in the target set of exercise types, determining the initial weight and the adjustment value of the target exercise type based on the health data, the target health data, and a preset rule, so as to obtain a plurality of initial weights and a plurality of adjustment values; and determining the exercise type weights corresponding to each exercise type in the target set of exercise types based on the plurality of adjustment values and the plurality of initial weights.

[0272] In some examples, a Markov decision model is constructed based on the health data, the set of exercise types, the health benefit, the exercise risk, and the user's exercise type preference; in the process of optimizing and adjusting the expected health data, the Markov decision model is optimized by using a double Q-learning network to obtain a target exercise plan that meets the target constraint conditions; and based on the target exercise plan that meets the target constraint conditions, the target video plan that meets the target constraint conditions is obtained.

[0273] It is worth mentioning that each module involved in this embodiment is a logical module. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. In addition, in order to highlight the innovative part of this application, units that are not closely related to solving the technical problems proposed in this application are not introduced in this embodiment, but this does not mean that there are no other units in this embodiment.

[0274] Embodiment III:

[0275] Another embodiment of the present application relates to an electronic device, such as Figure 6 shown, including: at least one processor 901; and a memory 902 communicatively connected to the at least one processor 901; wherein, the memory 902 stores instructions executable by the at least one processor 901, and the instructions are executed by the at least one processor 901 to enable the at least one processor 901 to execute the personalized exercise plan recommendation method based on intelligent algorithms in the above embodiments.

[0276] Among them, the memory and the processor are connected by a bus. The bus may include any number of interconnected buses and bridges, and the bus connects various circuits of one or more processors and memories together. The bus may also connect various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art, so they will not be further described herein. The bus interface provides an interface between the bus and the transceiver. The transceiver may be one element or multiple elements, such as multiple receivers and transmitters, and provides a unit for communicating with various other devices on the transmission medium. The data processed by the processor is transmitted over the wireless medium through the antenna. Further, the antenna also receives data and transmits the data to the processor.

[0277] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interface, voltage regulation, power management, and other control functions. The memory can be used to store data used by the processor when executing operations.

[0278] Embodiment 4:

[0279] Another embodiment of the present application relates to a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, the method embodiments described above are implemented.

[0280] That is, those skilled in the art can understand that all or part of the steps in implementing the above method embodiments can be completed by instructing relevant hardware through a program. The program is stored in a storage medium and includes several instructions to enable a device (which can be a single-chip microcomputer, a chip, etc.) or a processor to execute all or part of the steps of the methods described in the various embodiments of the present application. The foregoing storage media include: USB flash drives, mobile hard disks, read-only memories (ROM), random access memories (RAM), magnetic disks, or optical discs, etc., which can store program codes.

[0281] Those of ordinary skill in the art can understand that the above embodiments are specific embodiments for implementing the present application. In actual applications, various changes can be made to them in form and details without departing from the spirit and scope of the present application.

Claims

1. A personalized exercise plan recommendation method based on intelligent algorithms, characterized in that Including: Obtain the health data of the target user and the set of exercise types corresponding to the health data, where the health data includes current health data and expected health data; Obtain the exercise needs of the target user, where the exercise needs include at least one of exercise type preference, weekly exercise frequency, and duration of each exercise; Utilize the multi-objective XGBoost-Pareto model to optimize and adjust the expected health data based on the exercise needs and the set of exercise types, to obtain a target exercise video plan that meets the target constraint conditions, where the target constraint conditions are related to at least one of health benefits, exercise risks, and user exercise type preferences; Recommend the target exercise video plan to the target user.

2. The personalized exercise plan recommendation method based on intelligent algorithms according to claim 1, wherein The multi-objective XGBoost-Pareto model includes a health benefit model, an exercise risk model, and a user compliance model. The step of utilizing the multi-objective XGBoost-Pareto model to optimize and adjust the expected health data based on the exercise needs and the set of exercise types to obtain a target exercise video plan that meets the target constraint conditions includes: Utilize the health benefit model to obtain an initial health benefit based on the health data; Utilize the exercise risk model to obtain an initial exercise risk based on the set of exercise types; Utilize the user compliance model to obtain an initial user compliance based on the exercise needs; Adjust the initial health benefit, the initial exercise risk, and the initial user compliance to obtain a target exercise video plan that meets the target constraint conditions.

3. The personalized exercise plan recommendation method based on intelligent algorithms according to claim 2, characterized in that The step of utilizing the health benefit model to obtain an initial health benefit based on the health data includes: Obtain the exercise health data after the target user exercises; Obtain the difference between the exercise health data and the health data; Obtain the health data weight for the health data; Utilize the health benefit model to obtain an initial health benefit based on the difference and the health data weight.

4. The personalized exercise plan recommendation method based on intelligent algorithms according to claim 2, wherein, The step of utilizing the exercise risk model to obtain an initial exercise risk based on the set of exercise types includes: For each exercise type in the set of exercise types, obtain the exercise intensity, joint load, and fatigue degree of the exercise type, so as to obtain multiple exercise intensities, multiple joint loads, and multiple fatigue degrees; Obtain the total exercise intensity based on the multiple exercise intensities, obtain the total joint load based on the multiple joint loads, and obtain the total fatigue degree based on the multiple fatigue degrees; Obtain the preset exercise intensity weight, the preset joint load weight, and the preset fatigue degree weight; Utilize the exercise risk model to obtain an initial exercise risk based on the preset exercise intensity weight, the preset joint load weight, the preset fatigue degree weight, the total exercise intensity, the total joint load, and the total fatigue degree.

5. The personalized exercise plan recommendation method based on intelligent algorithms according to claim 2, characterized in that The step of adjusting the initial health benefit, the initial exercise risk, and the initial user compliance to obtain a target exercise video plan that meets the target constraint conditions includes: By adjusting the initial health benefit, the initial exercise risk, and the initial user compliance, a target health benefit, a target exercise risk, and a target user compliance that meet the target constraint conditions are obtained; Based on the target health benefit, the expected health data is adjusted to obtain target health data; Based on the target exercise risk, the set of exercise types is adjusted to obtain a target set of exercise types; Based on the target user compliance, the exercise requirements are adjusted to obtain target exercise requirements; Based on the target health data, the exercise type weights corresponding to each exercise type in the target set of exercise types are determined; Based on the exercise type weights corresponding to each exercise type in the target set of exercise types, the target exercise requirements, the target health benefit, the target exercise risk, the target user compliance, and the target health data, a target exercise video plan is determined.

6. The personalized exercise plan recommendation method based on intelligent algorithms according to claim 5, characterized in that The determining the exercise type weights corresponding to each exercise type in the target set of exercise types based on the target health data includes: For each target exercise type in the target set of exercise types, an initial weight and an adjustment value of the target exercise type are determined based on the health data, the target health data, and a preset rule, so as to obtain a plurality of initial weights and a plurality of adjustment values; Based on the plurality of adjustment values and the plurality of initial weights, the exercise type weights corresponding to each exercise type in the target set of exercise types are determined.

7. The personalized exercise plan recommendation method based on intelligent algorithms according to claim 1, wherein It further includes: Based on the health data, the set of exercise types, the health benefit, the exercise risk, and the user's exercise type preference, a Markov decision model is constructed; During the process of optimizing and adjusting the expected health data, the Markov decision model is optimized by using a double Q-learning network to obtain a target exercise plan that meets the target constraint conditions; Based on the target exercise plan that meets the target constraint conditions, the target video plan that meets the target constraint conditions is obtained.

8. A personalized exercise plan recommendation device based on intelligent algorithms, characterized in that, It includes: An acquisition module for acquiring the health data of a target user and the set of exercise types corresponding to the health data, where the health data includes current health data and expected health data; The acquisition module is further configured to acquire the exercise requirements of the target user, where the exercise requirements include at least one of exercise type preference, weekly exercise frequency, and duration of each exercise; An optimization module for using a multi-objective XGBoost-Pareto model to optimize and adjust the expected health data based on the exercise requirements and the set of exercise types to obtain a target exercise video plan that meets the target constraint conditions, where the target constraint conditions are related to at least one of health benefit, exercise risk, and user exercise type preference; A recommendation module for recommending the target exercise video plan to the target user.

9. An electronic device, characterized in that, It includes: At least one processor; And, A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the personalized exercise plan recommendation method based on an intelligent algorithm as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the personalized exercise plan recommendation method based on an intelligent algorithm as described in any one of claims 1 to 7.

Images