A weight loss management consulting platform and its implementation method

Abstract

The application discloses a kind of fat reduction management consultation platform and its implementation method, the present application relates to fat reduction management technical field, including data collection module, basal metabolism module, body fat trend prediction module, nutrition intervention module and exercise intervention module;Data collection module is responsible for the acquisition including the basic sign data of user, body weight, body fat rate;And dynamic acquisition user exercise frequency, training type and diet record;Basal metabolism module calculates basal metabolic rate BMR (t) according to user body weight and body fat rate;According to heart rate frequency and training type, generate activity correction factor AF (t) every time;The present application constructs historical heat input and body fat change sequence by body fat trend prediction module, calls XGBoost model to predict future body fat change, combines error weighted fusion mechanism to output final body fat prediction value, and judges whether to enter stable period or fat reduction rebound trend, effectively monitors user body fat change trend, effectively helps user to understand fat reduction process.

Description

Technical Field

[0001] This invention relates to the field of weight loss management technology, specifically to a weight loss management consulting platform and its implementation method. Background Technology

[0002] With increasing health awareness and the high incidence of obesity-related chronic diseases, more and more users are seeking scientific and effective weight loss management methods. Existing weight loss management platforms mainly rely on users' self-recorded weight changes and exercise and diet logs, and use a simple calorie deficit formula of basal metabolic rate + exercise consumption - dietary calories to estimate and provide users with weight loss suggestions. Some smart wearable devices can also collect behavioral data such as steps and heart rate, and use them to estimate users' energy consumption and exercise intensity to help users develop exercise plans and dietary structures.

[0003] Existing systems mostly use static metabolic calculation formulas and fail to incorporate dynamic activity corrections based on continuous user behavior. This results in significant deviations in estimating body fat percentage trends and poor accuracy in energy consumption calculations when daily lifestyles change frequently.

[0004] To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a weight loss management consultation platform and its implementation method.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a weight loss management consultation platform and its implementation method, comprising: a data collection module, a basal metabolic rate module, a body fat trend prediction module, a nutritional intervention module, and an exercise intervention module;

[0007] The data collection module is responsible for collecting basic physical characteristics data of users, including weight and body fat percentage; and dynamically collecting users' exercise frequency, training type and diet records.

[0008] The basal metabolic module calculates the basal metabolic rate BMR(t) based on the user's weight and body fat percentage; it generates an activity correction factor AF(t) for each moment based on heart rate frequency and training type, and combines the metabolic data with the activity factor to obtain the predicted value TDEEbase(t).

[0009] The body fat trend prediction module is responsible for building a regression prediction model. The model outputs the XGBoost body fat percentage prediction sequence FatX(t+1→t+n). The XGBoost body fat percentage prediction sequence FatX(t+1→t+n) is then weighted and fused to obtain the final body fat prediction value Fatpredict(t+n).

[0010] The nutrition intervention module constructs a calorie difference sequence ΔCal(t) based on the actual intake data within consecutive days in the diet record, and constructs a nutrition behavior correction factor ΔNutri(t) for each time moment based on the energy intake deviation distribution calculated at each time moment. Based on the calorie difference sequence ΔCal(t) and the ΔNutri(t) correction factor, the target nutrition intake structure matrix Nutrimatrix(t) for time t is constructed.

[0011] The exercise intervention module is responsible for combining the user's current body fat Fat(t), target body fat percentage Ftarget, energy expenditure TDEEpredict(t), and intake Calintake(t) to calculate the daily exercise calorie deficit target Edeficit(t).

[0012] The basal metabolic module constructs the initial metabolic estimate BMR(t) at each time step based on the user's weight and body fat percentage; the initial metabolic estimate BMR(t) is calculated using the Katch-McArdle formula; the frequency of heart rate occurrence in different intervals is analyzed; training type labels are obtained to form an activity level vector; the activity level vector and heart rate frequency are input into the AF mapping model to obtain the activity correction factor AF(t) at each time step; the initial metabolic estimate BMR(t) at each time step is weighted and combined with the activity correction factor AF(t) to obtain the predicted value TDEEbase(t);

[0013] TDEEbase(t) = BMR(t) × AF(t)

[0014] Where AF(t) is the activity correction factor at time t, BMR(t) is the basal metabolic rate at time t, and TDEE(t) is the energy consumption correction term at time t.

[0015] Based on the predicted value TDEEbase(t), it is corrected by incorporating abnormal distribution factors in recent behavior to construct the TDEE prediction correction term TDEE(t) for each time point. This includes calculating the distribution of the proportion of abnormal heart rate intervals before and after each time point; the correction term TDEE(t) is calculated using an exponential function weighting, and then weighted by TDEEbase(t) to obtain the corrected total energy consumption prediction value TDEEpredict(t).

[0016] TDEEpredict(t)=TDEEbase(t)+TDEE(t)

[0017] Where TDEEpredict(t) is the corrected total energy consumption prediction, and TDEEbase(t) is the basic total energy consumption prediction.

[0018] The body fat trend prediction module constructs an input matrix from historical body fat percentage change data within a preset period and trains a regression prediction model for predicting future body fat changes. The regression prediction model is an XGBoost model based on gradient boosting trees, which is used to learn the nonlinear mapping relationship between body fat changes and calories and behavioral factors. At each time t, the regression prediction model is called, and the historical data of the most recent k days, including the user's exercise frequency, training type, and total energy consumption prediction value TDEEpredict(t), are input. The model outputs the corresponding XGBoost body fat percentage prediction sequence FatX(t+1→t+n).

[0019] The XGBoost body fat percentage prediction sequence FatX(t+1→t+n) output by the regression prediction model is fused using error weighting to obtain the final body fat prediction value Fatpredict(t+n). This includes: comparing the prediction errors at the most recent known time t-1, calculating the inverse weighting factor of each model's error, and constructing the following fusion function:

[0020] Fatpredict(t+n)=(1-α)×FatX(t+1→t+n)

[0021] Where FatX(t+1→t+n) is the XGBoost body fat percentage prediction sequence; α is the dynamic weighting factor for model prediction accuracy; Fatpredict(t+n) is the weighted fusion predicted body fat percentage; where α is the dynamic weighting factor for prediction accuracy, which is generated in real time by the inverse error function;

[0022] Based on the final predicted body fat value Fatpredict(t+n) and the user-defined target body fat percentage Ftarget, a threshold comparison is performed on the predicted sequence to identify the moment when Fatpredict(t+n)≤Ftarget is first satisfied in the predicted sequence, and this moment is output as the predicted target achievement time Thit.

[0023] If the body fat target is not met within the set n days, the abnormal body fat trend identification process will be initiated to determine whether the user has a stable period. This includes calculating the local slope β in the predicted body fat sequence. If the slope is close to zero continuously, it is determined to be a stable period. If the body fat prediction curve rises in the opposite direction under the condition of continuous calorie deficit, and the slope β is greater than the positive threshold, it is determined to be an unstable period.

[0024] The nutritional intervention module constructs a calorie difference sequence ΔCal(t) based on the actual intake data over consecutive days in the diet record, and constructs a nutritional behavior correction factor ΔNutri(t) for each time point based on the energy intake deviation distribution calculated at each time point; specifically, it includes calculating the difference between the average value of the actual calorie intake value Calintake(tk→t) and the target calorie value Caltarget(t).

[0025] Based on the calorie difference sequence ΔCal(t) and the correction factor ΔNutri(t), a target nutrient intake structure matrix Nutrimatrix(t) is constructed at time t. This matrix is ​​then decomposed into three meal target structure vectors along the time dimension. The platform's food database is invoked to match food combinations that meet the target energy and nutrient structure for each meal, constructing a dietary recommendation scheme. This scheme is selected based on the following rules: target calorie error ≤ ±5% for each meal; error for each nutrient category ≤ ±10%; and user preference / allergy / contraindication filtering tags are used. The output meal recommendation scheme includes: food name, unit weight, nutrient content, and total calories.

[0026] The exercise intervention module calculates the daily exercise calorie deficit target Edeficit(t) based on the user's set body fat percentage target Ftarget, body fat percentage Fat(t), and total energy consumption TDEEpredict(t), combined with the user's daily calorie intake Calintake(t) and training cycle information.

[0027] Edeficit(t)=TDEEpredict(t)-Calintake(t)+Δgoal(t)

[0028] Where TDEEpredict(t) is the predicted total energy consumption at time t after activity correction and abnormal behavior correction; Calintake(t) is the actual calorie intake at time t;

[0029] Δgoal(t) is the calorie deficit adjustment term required to achieve the body fat percentage target Ftarget, and is calculated using the following formula:

[0030] Δgoal(t)=7700×[Fat(t)-Ftarget]×Weight(t)

[0031] Here, 7700kcal is the energy required to theoretically reduce 1 kg of body fat; Fat(t) is the current body fat percentage; Ftarget is the body fat percentage target set by the user; and Weight(t) is the current body weight.

[0032] A method for implementing weight loss management consultation, characterized by the following:

[0033] S1. Users input basic vital signs information through the data collection module, and the system records the user's exercise frequency, training type, and diet records;

[0034] S2. The basal metabolic module generates the basal metabolic rate BMR(t) based on the user's basic vital signs information, generates the activity correction factor AF(t) by combining the user's training behavior and heart rate data, and then calculates the predicted value TDEEbase(t).

[0035] S3, the body fat trend prediction module constructs a regression prediction model. The model outputs the XGBoost body fat percentage prediction sequence FatX(t+1→t+n) and processes it. The weighted and fused body fat percentage prediction value Fatpredict(t+n) is compared with the user's target body fat to determine the time to reach the target. The slope of the predicted body fat change sequence is calculated to determine whether the user has entered a stable period.

[0036] S4. The nutrition intervention module analyzes the actual intake data within consecutive days in the diet record, calculates the calorie difference sequence ΔCal(t) and the nutrition behavior correction factor ΔNutri(t) at each time point, and constructs the target nutrition intake structure matrix Nutrimatrix(t). Based on the target nutrition intake structure matrix Nutrimatrix(t), it matches and generates a diet recommendation plan from the food database.

[0037] S5, the exercise intervention module calculates the daily exercise calorie deficit target Edeficit(t) and displays it to the user.

[0038] This invention provides a weight loss management consultation platform and its implementation method. Compared with the prior art, it has the following advantages:

[0039] This invention constructs a historical calorie intake and body fat change sequence through a body fat trend prediction module, calls the XGBoost model to predict future body fat changes, and outputs the final body fat prediction value by combining an error weighted fusion mechanism. It also determines whether the body fat has entered a stable period or a weight loss rebound trend, effectively monitoring the user's body fat change trend and helping the user understand the weight loss process.

[0040] This invention constructs a nonlinear predictive relationship between body fat changes and calorie differences through a body fat trend prediction module. Based on the XGBoost model, it outputs a body fat prediction sequence FatX(t+1→t+n) and calculates the final prediction value Fatpredict(t+n) through an error inverse fusion mechanism. Combined with the target body fat percentage Ftarget, it determines the time to reach the target Thit. This can effectively improve the accuracy of body fat prediction, thereby optimizing the individual's fat loss rhythm and the time to achieve the target.

[0041] This invention uses an exercise intervention module to comprehensively calculate the daily exercise calorie deficit target Edeficit(t) based on the body fat target, current body fat percentage, daily energy expenditure TDEEpredict(t), and actual calorie intake Calintake(t). It also calculates the total energy difference Δgoal(t) required for fat loss by combining the target body fat percentage, thereby improving the adaptability of exercise intervention, increasing training efficiency, and accelerating the rate of body fat reduction. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the principle framework of the present invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Please see Figure 1 This application provides a weight loss management consulting platform and its implementation method, including: a data collection module, a basal metabolic rate module, a body fat trend prediction module, a nutritional intervention module, and an exercise intervention module;

[0045] The data collection module is responsible for collecting basic physical characteristics data of users, including weight and body fat percentage; and dynamically collecting users' exercise frequency, training type and diet records.

[0046] The basal metabolic module calculates the basal metabolic rate BMR(t) based on the user's weight and body fat percentage; it generates an activity correction factor AF(t) for each moment based on heart rate frequency and training type, and combines the metabolic data with the activity factor to obtain the predicted value TDEEbase(t).

[0047] The body fat trend prediction module is responsible for building a regression prediction model. The model outputs the XGBoost body fat percentage prediction sequence FatX(t+1→t+n). The XGBoost body fat percentage prediction sequence FatX(t+1→t+n) is then weighted and fused to obtain the final body fat prediction value Fatpredict(t+n).

[0048] The nutrition intervention module constructs a calorie difference sequence ΔCal(t) based on the actual intake data within consecutive days in the diet record, and constructs a nutrition behavior correction factor ΔNutri(t) for each time moment based on the energy intake deviation distribution calculated at each time moment. Based on the calorie difference sequence ΔCal(t) and the ΔNutri(t) correction factor, the target nutrition intake structure matrix Nutrimatrix(t) for time t is constructed.

[0049] The exercise intervention module is responsible for combining the user's current body fat Fat(t), target body fat percentage Ftarget, energy expenditure TDEEpredict(t), and intake Calintake(t) to calculate the daily exercise calorie deficit target Edeficit(t).

[0050] The basal metabolic module constructs the initial metabolic estimate BMR(t) for each time step based on the user's weight and body fat percentage; the initial metabolic estimate BMR(t) is calculated using the Katch-McArdle formula; the frequency of heart rate occurrence in different intervals is analyzed; training type labels are obtained to form an activity level vector; the activity level vector and heart rate frequency are input into the AF mapping model to obtain the activity correction factor AF(t) for each time step; the initial metabolic estimate BMR(t) for each time step is weighted and combined with the activity correction factor AF(t) to obtain the predicted value TDEEbase(t);

[0051] TDEEbase(t) = BMR(t) × AF(t)

[0052] Where AF(t) is the activity correction factor at time t, BMR(t) is the basal metabolic rate at time t, and TDEE(t) is the energy consumption correction term at time t.

[0053] The basal metabolic rate (BMR) module constructs a dynamic estimation model for the BMR(t) based on the user's static physiological parameters, including gender, age, height, weight, and lean body mass, combined with historical health data fragments collected up to each time point. The model uses the Katch-McArdle formula:

[0054] BMR(t) = 370 + 21.6 × LBM(t)

[0055] Wherein, LBM(t) is the lean body mass at time t, calculated based on the relationship between body fat percentage and body weight. The trend of lean body mass change is inferred by using the user's initial registration information and the periodic body fat percentage measurement data provided by the smart body fat scale.

[0056] To further improve the dynamic adaptability of basal metabolic rate estimation, the system constructs a multidimensional profile distribution mapping model for static parameters, mapping each user's static parameters to the corresponding dimension in the NHANES database, and outputting the initial BMR estimate BMR(t).

[0057] The tools used include wearable devices (such as Apple Watch, Huawei Band, and Xiaomi Band) to collect heart rate and steps; and smart body fat scales (such as Withings, Yunmai Haoqing, and TANITA) to collect body fat percentage and weight.

[0058] Based on the predicted value TDEEbase(t), it is corrected by incorporating abnormal distribution factors in recent behavior to construct the TDEE prediction correction term TDEE(t) for each time point. This includes calculating the distribution of the proportion of abnormal heart rate intervals before and after each time point; the correction term TDEE(t) is calculated using an exponential function weighting, and then weighted by TDEEbase(t) to obtain the corrected total energy consumption prediction value TDEEpredict(t).

[0059] TDEEpredict(t)=TDEEbase(t)+TDEE(t)

[0060] Where TDEEpredict(t) is the corrected total energy consumption prediction, and TDEEbase(t) is the basic total energy consumption prediction.

[0061] The body fat trend prediction module constructs an input matrix from historical body fat percentage change data within a preset period and trains a regression prediction model for predicting future body fat changes. The regression prediction model is an XGBoost model based on gradient boosting trees, which is used to learn the nonlinear mapping relationship between body fat changes and calories and behavioral factors. At each time t, the regression prediction model is called, and the historical data of the most recent k days, including the user's exercise frequency, training type and total energy consumption prediction value TDEEpredict(t), are input. The model outputs the corresponding XGBoost body fat percentage prediction sequence FatX(t+1→t+n).

[0062] The XGBoost body fat percentage prediction sequence FatX(t+1→t+n) output by the regression prediction model is fused using error weighting to obtain the final body fat prediction value Fatpredict(t+n). This includes: comparing the prediction errors at the most recent known time t-1, calculating the inverse weighting factor of each model's error, and constructing the following fusion function:

[0063] Fatpredict(t+n)=(1-α)×FatX(t+1→t+n)

[0064] Where FatX(t+1→t+n) is the XGBoost body fat percentage prediction sequence; α is the dynamic weighting factor for model prediction accuracy; Fatpredict(t+n) is the weighted fusion predicted body fat percentage; where α is the dynamic weighting factor for prediction accuracy, which is generated in real time by the inverse error function;

[0065] The prediction model employs the Gradient Boosting Tree (XGBoost) algorithm to construct a nonlinear regressor, used to model the complex interaction between body fat changes and the aforementioned factors. The XGBoost model possesses strong nonlinear expressive power and efficient training performance, making it suitable for modeling small-sample, high-dimensional data. During the model training phase, a body fat change sample library generated from the platform's large dataset is used, employing MSE (mean squared error) or MAE (absolute error) as the loss function for training the target regression prediction task.

[0066] During runtime, at each time t, the system calls the trained XGBoost model, inputs the historical data of the most recent k days (including Fat(tk→t) and ΔCal(tk→t)), predicts the body fat change trend for the next n days, and outputs the predicted sequence FatX(t+1→t+n), which is used for subsequent trend judgment, target prediction and stabilization period identification.

[0067] Model training code:

[0068]

[0069]

[0070] Based on the final predicted body fat value Fatpredict(t+n) and the user-defined target body fat percentage Ftarget, a threshold comparison is performed on the predicted sequence to identify the moment when Fatpredict(t+n)≤Ftarget is first satisfied in the predicted sequence, and this moment is output as the predicted target achievement time Thit.

[0071] If the body fat target is not met within the set n days, the abnormal body fat trend identification process will be initiated to determine whether the user has a stable period. This includes calculating the local slope β in the predicted body fat sequence. If the slope is close to zero continuously, it is determined to be a stable period. If the body fat prediction curve rises in the opposite direction under the condition of continuous calorie deficit, and the slope β is greater than the positive threshold, it is determined to be an unstable period.

[0072] By constructing a sliding local linear fitting algorithm for the predicted body fat percentage sequence, slope analysis is performed on the continuous predicted body fat percentage value Fatpredict(t+1→t+n) to determine its changing trend:

[0073] Define the sliding window length as m (e.g., 3 days), and perform linear regression fitting on the body fat change values ​​within each window;

[0074] For each window, the linear fitting slope β_i is calculated to obtain the slope sequence {β1,β2,...,β_{n-m+1}};

[0075] Count the number of windows with continuous slopes close to 0 (|β_i|<ε, where ε is a set stationarity threshold);

[0076] If a slope is close to 0 for ≥k consecutive windows, it is considered a "stable period";

[0077] If the slope is significantly greater than the positive threshold β_threshold and is in a positive growth direction (β>β_threshold), and the user has a continuous heat deficit during this period (i.e., ΔCal(t)<0 and lasts for ≥p days), then it is determined to be an "unstable period".

[0078] To determine an "unstable period," the following three conditions must be met simultaneously:

[0079] The prediction slope increases significantly: within a certain window, β > β_threshold (e.g., β_threshold = 0.05);

[0080] Predicted body fat percentage shows a continuous increasing trend: Fatpredict(t+i) - Fatpredict(t+i-1) ≥ δ (e.g., ≥ 0.1%) for q days;

[0081] The user maintains a heat deficit state: ΔCal(t+i)<0 for the last q days, indicating that the user has implemented the plan but has not achieved the expected results.

[0082] Local slope calculation code:

[0083]

[0084]

[0085] The nutrition intervention module constructs a calorie difference sequence ΔCal(t) based on the actual intake data over consecutive days in the diet record, and constructs a nutritional behavior correction factor ΔNutri(t) for each time point based on the energy intake deviation distribution calculated at each time point; specifically, it includes calculating the difference between the average value of the actual calorie intake value Calintake(tk→t) and the target calorie value Caltarget(t).

[0086] The nutrition intervention module first performs statistical analysis on the user's daily calorie intake value Calintake(tk→t) within a historical consecutive number of days (such as the last 7 or 14 days), and compares it with the platform's preset calorie control target Caltarget(t) to construct a daily calorie difference sequence.

[0087] Calintake(t) is the total daily calorie intake collected by the platform through manual recording or OCR recognition of food labels / photo input interface; Caltarget(t) is the dynamic calorie control target determined by the system based on TDEEpredict(t) and the set calorie deficit Edeficit(t).

[0088] Subsequently, the module constructs a corresponding nutritional behavior correction factor ΔNutri(t) based on the fluctuation range, deviation direction, and stability of the ΔCal(t) sequence within a set time window. This factor can be generated statistically (e.g., standard deviation, deviation ratio, volatility index) or by using a behavioral scoring function to integrate calorie stability and intake deviation, and is defined as follows:

[0089] ΔNutri(t)=f(Var(ΔCal),Mean(ΔCal),Frequencyoutliers)

[0090] Here, Frequency_{outliers} represents the frequency of abnormal heat fluctuations (>±20%).

[0091] Based on the calorie difference sequence ΔCal(t) and the correction factor ΔNutri(t), a target nutrient intake structure matrix Nutrimatrix(t) is constructed at time t. This matrix is ​​then decomposed into three meal target structure vectors along the time dimension. The platform's food database is accessed to match food combinations that meet the target energy and nutrient structure for each meal, constructing a dietary recommendation plan. The recommendation plan is filtered based on the following rules: target calorie error ≤ ±5% for each meal; error for each nutrient category ≤ ±10%; and user preference / allergy / contraindication filtering tags are incorporated. The output meal recommendation plan includes: food name, unit weight, nutrient content, and total calories.

[0092] The exercise intervention module calculates the daily exercise calorie deficit target Edeficit(t) based on the user's set body fat percentage target Ftarget, body fat percentage Fat(t), and total energy expenditure TDEEpredict(t), combined with the user's daily calorie intake Calintake(t) and training cycle information.

[0093] Edeficit(t)=TDEEpredict(t)-Calintake(t)+Δgoal(t)

[0094] Where TDEEpredict(t) is the predicted total energy consumption at time t after activity correction and abnormal behavior correction; Calintake(t) is the actual calorie intake at time t;

[0095] Δgoal(t) is the calorie deficit adjustment term required to achieve the body fat percentage target Ftarget, and is calculated using the following formula:

[0096] Δgoal(t)=7700×[Fat(t)-Ftarget]×Weight(t)

[0097] Here, 7700kcal is the energy required to theoretically reduce 1 kg of body fat; Fat(t) is the current body fat percentage; Ftarget is the body fat percentage target set by the user; and Weight(t) is the current body weight.

[0098] The specific method for implementing weight loss management consulting is as follows:

[0099] S1. Users input basic vital signs information through the data collection module, and the system records the user's exercise frequency, training type, and diet records;

[0100] S2. The basal metabolic module generates the basal metabolic rate BMR(t) based on the user's basic vital signs information, generates the activity correction factor AF(t) by combining the user's training behavior and heart rate data, and then calculates the predicted value TDEEbase(t).

[0101] S3, the body fat trend prediction module constructs a regression prediction model. The model outputs the XGBoost body fat percentage prediction sequence FatX(t+1→t+n) and processes it. The weighted and fused body fat percentage prediction value Fatpredict(t+n) is compared with the user's target body fat to determine the time to reach the target. The slope of the predicted body fat change sequence is calculated to determine whether the user has entered a stable period.

[0102] S4. The nutrition intervention module analyzes the actual intake data within consecutive days in the diet record, calculates the calorie difference sequence ΔCal(t) and the nutrition behavior correction factor ΔNutri(t) at each time point, and constructs the target nutrition intake structure matrix Nutrimatrix(t). Based on the target nutrition intake structure matrix Nutrimatrix(t), it matches and generates a diet recommendation plan from the food database.

[0103] S5, the exercise intervention module calculates the daily exercise calorie deficit target Edeficit(t) and displays it to the user.

[0104] This invention constructs a nonlinear predictive relationship between body fat changes and calorie differences through a body fat trend prediction module. Based on the XGBoost model, it outputs a body fat prediction sequence FatX(t+1→t+n) and calculates the final prediction value Fatpredict(t+n) through an error inverse fusion mechanism. Combined with the target body fat percentage Ftarget, it determines the time to reach the target Thit. This can effectively improve the accuracy of body fat prediction, thereby optimizing the individual's fat loss rhythm and the time to achieve the target.

[0105] This invention uses an exercise intervention module to comprehensively calculate the daily exercise calorie deficit target Edeficit(t) based on the body fat target, current body fat percentage, daily energy expenditure TDEEpredict(t), and actual calorie intake Calintake(t). It also calculates the total energy difference Δgoal(t) required for fat loss by combining the target body fat percentage, thereby improving the adaptability of exercise intervention, increasing training efficiency, and accelerating the rate of body fat reduction.

[0106] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0107] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A fat reduction management consultation platform, characterized by, Comprise: Data collection module, basal metabolism module, body fat trend prediction module, nutrition intervention module and exercise intervention module; The data collection module is responsible for collecting user's basic physical data including weight and body fat rate; And dynamically collect user's exercise frequency, training type and diet record; The basal metabolism module calculates the basal metabolic rate BMR(t) according to the user's weight and body fat rate; According to the heart rate frequency and the training type, the activity correction factor AF(t) is generated at each time, and the metabolic data is combined with the activity factor to obtain the predicted value TDEEbase(t); The body fat trend prediction module is responsible for constructing a regression prediction model, and the model outputs the XGBoost body fat rate prediction sequence FatX(t+1→t+n), and the XGBoost body fat rate prediction sequence FatX(t+1→t+n) is error weighted and fused to obtain the final body fat prediction value Fatpredict(t+n); The nutrition intervention module constructs the calorie difference sequence ΔCal(t) according to the actual intake data in the diet record for consecutive days, and constructs the nutrition behavior correction factor ΔNutri(t) according to the energy intake deviation distribution calculated at each time, and constructs the target nutrition intake structure matrix Nutrimatrix(t) at time t based on the calorie difference sequence ΔCal(t) and the ΔNutri(t) correction factor; The exercise intervention module is responsible for calculating the daily exercise calorie deficit target Edeficit(t) by comprehensively considering the user's current body fat Fat(t), target body fat rate Ftarget, energy consumption TDEEpredict(t) and intake Calintake(t).

2. A fat management consultation platform as claimed in claim 1, wherein, The basal metabolism module constructs the initial metabolic estimate BMR(t) at each time according to the user's weight and body fat rate; The initial metabolic estimate BMR(t) is calculated using the Katch-McArdle formula; the frequency of heart rate in different intervals is analyzed; the activity level vector is constructed by obtaining the training type label; the activity level vector and the heart rate frequency are input into the AF mapping model to obtain the activity correction factor AF(t) at each time; the initial metabolic estimate BMR(t) at each time and the activity correction factor AF(t) are weighted and combined to obtain the predicted value TDEEbase(t); TDEEbase(t) = BMR(t) × AF(t) Wherein, AF(t) is the activity correction factor at time t, BMR(t) is the basal metabolic rate at time t; TDEE(t) is the energy consumption correction term at time t.

3. The fat management consultation platform according to claim 2, wherein, On the basis of the predicted value TDEEbase(t), combined with the abnormal distribution factor in the recent behavior, it is corrected to construct the TDEE prediction correction term TDEE(t) at each time, including calculating the proportion distribution of heart rate abnormal interval before and after each time; the correction term TDEE(t) is calculated by using the exponential function weighting, and the weighted TDEEbase(t) is weighted to obtain the corrected total energy consumption prediction value TDEEpredict(t): TDEEpredict(t) = TDEEbase(t) + TDEE(t) Wherein, TDEEpredict(t) is the corrected total energy expenditure prediction value, and TDEEbase(t) is the basic total energy expenditure prediction value.

4. The fat management consultation platform according to claim 1, wherein, The body fat trend prediction module constructs historical body fat rate change data in a preset period into an input matrix, trains a regression prediction model for future body fat change prediction, the regression prediction model is an XGBoost model based on gradient boosting tree, and is used to learn a nonlinear mapping relationship between body fat change and heat, and a behavior factor; at each time t, the regression prediction model is called, historical data of the last k days including user exercise frequency, training type and total energy expenditure prediction value TDEEpredict(t) are input, and the model outputs a corresponding XGBoost body fat rate prediction sequence FatX(t+1→t+n).

5. The fat management consultation platform according to claim 4, wherein, Error weighted fusion is performed on the XGBoost body fat rate prediction sequence FatX(t+1→t+n) output by the regression prediction model to obtain a final body fat prediction value Fatpredict(t+n), including: comparing the prediction error of the last known time t-1, calculating the inverse error weighting factor of each model; the following fusion function is constructed: Fatpredict(t+n)=(1-α)×FatX(t+1→t+n) Wherein, FatX(t+1→t+n) is the XGBoost body fat rate prediction sequence; α is a model prediction accuracy dynamic weight factor; Fatpredict(t+n) is the body fat rate prediction value after weighted fusion; wherein α is a prediction accuracy dynamic weight factor, which is generated in real time by an error inverse function; According to the final body fat prediction value Fatpredict(t+n) and the target body fat rate Ftarget set by the user, threshold comparison is performed on the prediction sequence, the time when Fatpredict(t+n)≤Ftarget is first met in the prediction sequence is identified, and the time is output as a prediction compliance time Thit.

6. The fat management consultation platform according to claim 5, wherein, If the body fat compliance condition cannot be met within the set n days, the body fat abnormal trend identification process is entered, whether the user has a stable period is judged, including calculating the local slope β in the predicted body fat sequence, if the continuous slope is close to zero, it is determined that it is a stable period; if the body fat prediction curve reverses and the slope β>positive threshold under the condition of continuous heat deficit, it is determined that it is an unstable period.

7. The fat management consultation platform according to claim 1, wherein, The nutrition intervention module constructs a heat difference sequence ΔCal(t) according to the actual intake data in the diet record for consecutive days, and constructs a nutrition behavior correction factor ΔNutri(t) according to the energy intake deviation distribution calculated at each time; specifically including the difference between the average value of the actual heat intake value Calintake(t-k→t) and the target heat value Caltarget(t).

8. The fat management consultation platform according to claim 7, wherein, Based on the heat difference sequence ΔCal(t) and the ΔNutri(t) correction factor, the target nutrient intake structure matrix Nutrimatrix(t) at time t is constructed, and the target nutrient intake structure matrix Nutrimatrix(t) is disassembled into three meal target structure vectors according to the time dimension; calling the platform food database, matching the food combination that meets the target energy and nutrient structure in each meal, constructing a diet recommendation scheme, the diet recommendation scheme is screened based on the following rules: the error of each meal target heat is ≤±5%; the error of each type of nutrient is ≤±10%; combined with user preference / allergy / taboo screening label; the output of three meal recommendation scheme includes: food name, unit weight, nutrient content and heat sum.

9. The fat management consultation platform according to claim 1, wherein, The exercise intervention module calculates the daily exercise heat deficit target Edeficit(t) according to the user set body fat rate target Ftarget, body fat rate Fat(t) and total energy consumption TDEEpredict(t), combined with user daily intake heat Calintake(t) and training cycle information; Edeficit(t)=TDEEpredict(t)-Calintake(t)+Δgoal(t) Wherein, TDEEpredict(t) is the total energy consumption prediction value based on activity correction and abnormal behavior correction at time t; Calintake(t) is the actual heat intake at time t; Δgoal(t) is the heat deficit adjustment item required to achieve the body fat rate target Ftarget, and the calculation formula is as follows: Δgoal(t)=7700×[Fat(t)-Ftarget]×Weight(t) Wherein, 7700kcal is the energy required to theoretically reduce 1 kg of body fat; Fat(t) is the current body fat rate; Ftarget is the user set body fat rate target; Weight(t) is the current body weight.

10. A method for implementing a weight management consultation, characterized by, As follows: S1, the user enters the basic physical sign information through the data collection module, and the system records the user's exercise frequency, training type and diet record; S2, the basal metabolic module generates the basal metabolic rate BMR(t) based on the user's basic physical sign information, generates the activity correction factor AF(t) combined with the user's training behavior and heart rate data, and then calculates the prediction value TDEEbase(t); S3, the body fat trend prediction module constructs a regression prediction model, the model outputs the XGBoost body fat rate prediction sequence FatX(t+1→t+n) and processes, compares the weighted fused body fat rate prediction value Fatpredict(t+n) with the user target body fat, judges the compliance time; the slope of the predicted body fat change sequence is calculated to judge whether the user enters the stable period; S4, the nutrition intervention module analyzes the actual intake data in the diet record for consecutive days, calculates the calorie difference sequence ΔCal(t) and the nutrition behavior correction factor ΔNutri(t) at each time, and constructs a target nutrition intake structure matrix Nutrimatrix(t), and generates a diet recommendation scheme from the food database according to the target nutrition intake structure matrix Nutrimatrix(t); S5, the exercise intervention module calculates a daily exercise calorie deficit target Edeficit(t) and displays it to the user.

Images