Health management system based on artificial intelligence

Abstract

The invention relates to the technical field of health management, in particular to a health management system based on artificial intelligence. The method comprises the following steps: acquiring historical physiological indexes, diet data and environmental data; determining a current health state according to the historical physiological indexes and the diet data; determining restaurant menu constraint conditions according to the environment data and the diet data; and according to the current health state and the restaurant menu constraint condition, generating a food combination scheme and an exercise recommendation scheme, and sending the schemes to user equipment. Constructing a dynamic constraint condition of the environmental data through an API interface of a restaurant real-time menu and positioning information of user equipment; combining age and gender weight to optimize metabolism evaluation; a personalized exercise plan, a food combination scheme and an exercise recommendation scheme are made through a dynamic matching mechanism of exercise consumption and diet calories and are sent to user equipment, so that accurate recommendation and efficient execution of a health management scheme are realized.

Description

Technical Field

[0001] The present application relates to the field of health management technology, and in particular to a health management system based on artificial intelligence. Background Art

[0002] With the development of social economy and changes in lifestyle, chronic metabolic diseases have become a major threat to public health. Traditional health management methods mainly rely on manual recording of dietary data and linear analysis combined with basic physiological indicators.

[0003] However, traditional health management systems rely on users to manually input dietary information and make dietary recommendations based on static nutritional databases. This results in the problem that personalized recommendation results are unfeasible or have low execution rates, which has become a significant flaw in health management in existing technologies. Summary of the Invention

[0004] This application provides an artificial intelligence-based health management system to solve the above problems.

[0005] In a first aspect, the present application provides a health management method based on artificial intelligence, the method comprising: obtaining historical physiological indicators, dietary data and environmental data; determining the current health status based on the historical physiological indicators and the dietary data; determining the restaurant menu constraints based on the environmental data and the dietary data; generating a food combination plan and an exercise recommendation plan based on the current health status and the restaurant menu constraints and sending them to the user device.

[0006] This solution uses dual channels of biosensors and medical devices to acquire historical physiological indicators, improving the completeness of data dimensions. Using convolutional neural networks to analyze pre- and post-meal images, this solution processes soup-based foods to generate dietary data, helping to accurately calculate the oil content and purine concentration in soups and reduce calorie estimation errors. Combining GIS geographic information devices and device positioning data, it retrieves real-time menu data from nearby restaurants, accessibility indicators, and seasonal food availability to construct environmental data, helping to reduce the failure rate caused by regional supply differences. Analyzing the nonlinear relationship between changes in body fat percentage and carbohydrate intake captures the variable insulin sensitivity threshold effect, helping to automatically trigger the metabolic level correction coefficient. A dose-response surface is established to link uric acid level fluctuations with purine intake measurements, quantifying the cumulative impact of a high-purine diet on uric acid levels, namely the hysteresis effect, helping to improve the accuracy of gout risk warnings. The weight distribution of the current metabolic level feature vector is adjusted based on age and gender to output the final current health status, helping to reduce errors in calculating daily calorie requirements. Parsing a restaurant's real-time menu API to extract the list of dishes on sale and the inventory status of ingredients helps eliminate plan failures caused by sudden ingredient shortages and improve the feasibility of food combination plans. Based on the analysis of user dietary preferences in dietary data, user selection preferences are generated, helping to improve the taste compatibility of food combination plans. According to the nutritional rules of the "Dietary Guidelines for Chinese Residents," text constraints are converted into mathematical constraints, helping to improve the rate of nutritional balance. A multi-objective optimization algorithm is used to generate restaurant menu constraints based on the list of dishes on sale, the inventory status of ingredients, user selection preferences, and mathematical constraints, helping to improve the acceptance of food combination plans. The nutritional components of candidate dishes are broken down to select food combination plans that meet current health conditions, helping to balance nutritional compatibility, user preference matching, and restaurant inventory feasibility, shortening the time required to generate plans. Based on the user's exercise consumption and the amount of food combination to be consumed, exercise recommendations are generated, helping to improve calorie balance. An adaptive compression algorithm is used to match the user's device network status for transmission, ensuring the successful loading of food combination plans and exercise recommendations in weak network scenarios.

[0007] Optionally, the acquisition of historical physiological indicators, dietary data and environmental data includes: acquiring basic body information through biosensors; acquiring currently recorded physical examination records, analyzing the physical examination records, and determining advanced body information; parsing the basic body information and the advanced body information to determine historical physiological indicators; acquiring device positioning information, and determining the current location based on the device positioning information; acquiring GIS data, and determining environmental data based on the GIS data and the current location; acquiring historical dining records, parsing the historical dining records, and obtaining dietary data.

[0008] Optionally, determining the current health status based on the historical physiological indicators and the dietary data includes: analyzing the historical physiological indicators to determine changes in body fat percentage, fluctuations in uric acid levels, basal metabolic rate and exercise consumption within a preset time period; analyzing the dietary data to determine carbohydrate intake and purine intake; analyzing the changes in body fat percentage and the carbohydrate intake to determine the correlation; analyzing the fluctuations in uric acid levels and the purine intake to determine the response relationship; determining the current metabolic level based on the correlation, the response relationship, the exercise consumption and the basal metabolic rate; obtaining user information; the user information includes age and gender; determining the current health status based on the current metabolic level, the age and the gender.

[0009] Optionally, determining restaurant menu constraints based on the environmental data and the dietary data includes: determining restaurant information based on the environmental data and the GIS data; analyzing the dietary data to determine the user's dietary preferences; and determining restaurant menu constraints based on preset nutritional rules, the user's dietary preferences, and the restaurant information.

[0010] Optionally, parsing the historical dining records to obtain dietary data includes: parsing the historical dining records to determine text information and image information; analyzing the text information to determine food names, food portions, and eating times; and analyzing the image information to determine ingredient intake. Dietary data is obtained according to the food name, the food portion, the eating time and the ingredient intake.

[0011] Optionally, analyzing the picture information to determine ingredient intake includes: analyzing the picture information to determine pre-meal pictures and post-meal pictures; determining food state changes and food types based on the pre-meal pictures and post-meal pictures; determining the soup ratio, soup ingredients and food ingredients based on the food type and the pre-meal pictures; determining food intake and soup intake based on the food state changes and the soup ratio; determining ingredient intake based on the food intake, soup intake, soup ingredients and food ingredients.

[0012] Optionally, generating a food combination plan and an exercise recommendation plan based on the current health status and the restaurant menu constraints includes: determining candidate dishes based on the restaurant menu constraints; screening the candidate dishes based on the current health status to obtain a food combination plan; analyzing the ingredient intake based on the food combination plan to determine the amount to be consumed; inferring the user's daily exercise volume based on the historical physiological indicators; and determining an exercise recommendation plan based on the user's daily exercise volume and the amount to be consumed.

[0013] Optionally, the method further includes: receiving data feedback from the user device, analyzing the data feedback, and determining a solution deviation; determining a health status change and solution acceptance based on the solution deviation; and revising a recommendation strategy based on the health status change and solution acceptance.

[0014] Optionally, the recommendation strategy is modified according to the change in health status and the acceptance of the plan, including: determining food selection weights according to the acceptance of the plan; determining intake constraints according to the change in health status; determining the user's selection preference according to the food selection weights; determining optional food types according to the intake constraints; and modifying the recommendation strategy according to the selection preference and the optional food types.

[0015] In the second aspect, the present application provides a health management system based on artificial intelligence, which includes: a data acquisition module for acquiring historical physiological indicators, dietary data and environmental data; a status determination module for determining the current health status based on the historical physiological indicators and the dietary data; a condition determination module for determining the restaurant menu constraints based on the environmental data and the dietary data; a plan generation module for generating a food combination plan and an exercise recommendation plan based on the current health status and the restaurant menu constraints and sending them to the user device.

[0016] Optionally, when the data acquisition module acquires historical physiological indicators, dietary data and environmental data, it is used to: acquire basic body information through biosensors; acquire currently recorded physical examination records, analyze the physical examination records, and determine advanced body information; parse the basic body information and the advanced body information to determine historical physiological indicators; acquire device positioning information, and determine the current location based on the device positioning information; acquire GIS data, and determine environmental data based on the GIS data and the current location; acquire historical dining records, parse the historical dining records, and obtain dietary data.

[0017] Optionally, when the status determination module determines the current health status based on the historical physiological indicators and the dietary data, it is used to: analyze the historical physiological indicators to determine the changes in body fat percentage, fluctuations in uric acid levels, basal metabolic rate and exercise consumption within a preset time period; analyze the dietary data to determine the carbohydrate intake and purine intake; analyze the changes in body fat percentage and the carbohydrate intake to determine the correlation; analyze the fluctuations in uric acid levels and the purine intake to determine the response relationship; determine the current metabolic level based on the correlation, the response relationship, the exercise consumption and the basal metabolic rate; obtain user information; the user information includes age and gender; determine the current health status based on the current metabolic level, the age and the gender.

[0018] Optionally, when the condition determination module determines the restaurant menu constraints based on the environmental data and the dietary data, it is used to: determine the restaurant information based on the environmental data and the GIS data; analyze the dietary data to determine the user's dietary preferences; and determine the restaurant menu constraints based on preset nutritional rules, the user's dietary preferences and the restaurant information.

[0019] Optionally, when the data acquisition module parses the historical dining records to obtain dietary data, it is used to: parse the historical dining records to determine text information and picture information; analyze the text information to determine the food name, food portion and meal time; analyze the picture information to determine the ingredient intake; and obtain dietary data based on the food name, food portion, meal time and ingredient intake.

[0020] Optionally, when the data acquisition module analyzes the image information and determines the ingredient intake, it is used to: analyze the image information to determine the pre-meal picture and the post-meal picture; determine the food state change and the food type based on the pre-meal picture and the post-meal picture; determine the soup ratio, soup ingredients and food ingredients based on the food type and the pre-meal picture; determine the food intake and soup intake based on the food state change and the soup ratio; determine the ingredient intake based on the food intake, the soup intake, the soup ingredients and the food ingredients.

[0021] Optionally, when the plan generation module generates a food combination plan and an exercise recommendation plan based on the current health status and the restaurant menu constraints, it is used to: determine candidate dishes based on the restaurant menu constraints; screen the candidate dishes based on the current health status to obtain a food combination plan; analyze the ingredient intake based on the food combination plan to determine the amount to be consumed; infer the user's daily exercise volume based on the historical physiological indicators; and determine an exercise recommendation plan based on the user's daily exercise volume and the amount to be consumed. Optionally, the artificial intelligence-based health management system also includes a strategy modification module, which is used to: receive data feedback from the user device, analyze the data feedback, and determine plan deviations; determine health status changes and plan acceptance based on the plan deviations; and modify the recommendation strategy based on the health status changes and plan acceptance.

[0022] Optionally, when the strategy modification module modifies the recommendation strategy based on the change in health status and the acceptance of the plan, it is used to: determine the food selection weight based on the acceptance of the plan; determine the intake constraints based on the change in health status; determine the user's selection preference based on the food selection weight; determine the optional food type based on the intake constraints; and modify the recommendation strategy based on the selection preference and the optional food type. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0024] Figure 1 A schematic diagram of an application scenario provided in one embodiment of the present application; Figure 2 A flowchart of an artificial intelligence-based health management method provided in one embodiment of the present application; Figure 3 A schematic diagram of the structure of an artificial intelligence-based health management system provided in one embodiment of the present application. DETAILED DESCRIPTION

[0025] To make the purpose, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0026] In this document, the term "and / or" simply describes a relationship between related objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A exists alone, A and B exist simultaneously, or B exists alone. Furthermore, the character " / " in this document, unless otherwise specified, generally indicates an "or" relationship between the related objects.

[0027] The embodiments of the present application are described in further detail below with reference to the accompanying drawings.

[0028] Traditional health management systems rely on users to manually input dietary information and make dietary recommendations based on static nutrition databases. This results in personalized recommendation results being unfeasible or having a low execution rate, which has become a significant flaw in health management in existing technologies.

[0029] Based on this, the present application provides a health management system based on artificial intelligence, which obtains historical physiological indicators, dietary data and environmental data; determines the current health status based on historical physiological indicators and dietary data; determines the restaurant menu constraints based on environmental data and dietary data; generates food combination plans and exercise recommendation plans based on the current health status and restaurant menu constraints and sends them to the user device. Obtaining historical physiological indicators through dual channels of biosensors and medical equipment helps to improve the completeness of data dimensions; analyzing pre- and post-meal images through convolutional neural networks, performing processing on soups and generating dietary data, helps to accurately calculate the oil content and purine concentration in the soup, and reduce the error in calorie estimation; combining GIS geographic information equipment and equipment positioning data, retrieves real-time menu data, traffic accessibility indicators and seasonal food supply status of nearby restaurants, and constructs environmental data, which helps to reduce the non-executability rate caused by regional supply differences. Analyzing the nonlinear relationship between changes in body fat percentage and carbohydrate intake captures the effect of variable insulin sensitivity thresholds, helping to automatically trigger metabolic level correction coefficients. A dose-response surface is established to correlate uric acid level fluctuations with purine intake, quantifying the cumulative impact of a high-purine diet on uric acid levels, known as the hysteresis effect, helping to improve the accuracy of gout risk warnings. The weight distribution of the current metabolic level feature vector is adjusted based on age and gender to output the final current health status, helping to reduce errors in calculating daily calorie requirements. Parsing a restaurant's real-time menu API to extract the list of available dishes and ingredient inventory status helps eliminate plan failures caused by sudden ingredient shortages and improve the feasibility of food combination plans. Based on user dietary preferences analyzed from dietary data, user preference generation helps improve the taste compatibility of food combination plans. Based on the nutritional principles of the "Dietary Guidelines for Chinese Residents," textual constraints are converted into mathematical constraints, helping to improve the rate of nutritional balance achieved. A multi-objective optimization algorithm is used to generate restaurant menu constraints based on the list of available dishes and ingredient inventory status, user preference, and mathematical constraints, helping to improve the acceptability of food combination plans. Decomposing the nutritional components of candidate dishes and screening food combination plans that are suitable for the current health status helps to balance nutritional compatibility, user preference matching and restaurant inventory feasibility, and shortens the time spent on plan generation; generating exercise recommendation plans based on the user's exercise consumption and the amount of food combination to be consumed, helps to improve the calorie balance achievement rate; and transmitting through the user's device network status through an adaptive compression algorithm ensures the improvement of the loading success rate of food combination plans and exercise recommendation plans in weak network scenarios.

[0030] Figure 1This application provides a schematic diagram of an application scenario. The method provided herein is applied to health management and dietary recommendations. Specifically, the method is applied to any server, interacting with biosensors, medical devices, real-time menus, and user devices. Biosensors acquire basic physical indicators such as body fat percentage, while medical devices acquire dynamic physiological indicators such as uric acid levels. The data acquired by the biosensors and medical devices together form historical physiological indicators. Combined with dietary data uploaded by users via their devices, a convolutional neural network analyzes meal image features to accurately quantify the fat and purine content in soups. Dynamic constraints are constructed using the restaurant's real-time menu API and the user's device's location information, including environmental data such as transportation accessibility and food inventory. Nonlinear regression is used to analyze the threshold effects of body fat percentage and carbohydrate intake, optimizing metabolic assessments based on age and gender weights. Based on the "Dietary Guidelines for Chinese Residents," nutritional rules are converted into mathematical constraints, and a multi-objective optimization algorithm is used to generate menu plans that balance user preferences, nutritional compliance, and supply availability. A dynamic matching mechanism between exercise consumption and dietary calories is used to develop personalized exercise plans, food combination plans, and exercise recommendations, which are then sent to the user's device, enabling accurate recommendations and efficient execution of health management plans.

[0031] For specific implementation methods, please refer to the following embodiments.

[0032] Figure 2 This is a flowchart of an artificial intelligence-based health management method provided in one embodiment of the present application. The method of this embodiment can be applied to the server in the above scenario. Figure 2 As shown, the method includes: S201, obtaining historical physiological indicators, dietary data and environmental data; Historical physiological indicators can be physiological parameters of the user obtained by biosensors and medical devices in the past. The past period of time can be taken based on experience or manually specified.

[0033] Dietary data can be a structured record of nutritional intake.

[0034] Environmental data can be dynamic information of a dining place.

[0035] Specifically, the user's basic physical information is obtained in real time through biosensors, and medical equipment is simultaneously connected to obtain blood sugar, blood lipids, and blood pressure time series data as advanced physical information, which are then integrated into historical physiological indicators after data cleaning. At the same time, natural language processing technology is used to extract food names, food portion quantification values, and eating timestamp structured data from text records, or convolutional neural networks are used to analyze pre- and post-meal images, and processing is performed on soup-type foods: first, the liquid residue area in the plate is identified, and the volume change rate of the soup is calculated; second, based on OpenCV image segmentation technology, solid ingredients and soup components are separated; then, a component mapping is established with reference to the standard version of the "Chinese Food Composition Table", and the oil content and purine concentration in the soup are quantified; finally, dietary data containing a four-dimensional vector of food name, food portion, carbohydrate intake, and eating time is generated; then, the current location coordinates are obtained based on the user device positioning data, and the real-time menu data, traffic accessibility indicators, and seasonal food supply status of nearby restaurants are retrieved in combination with GIS geographic information equipment to construct environmental data.

[0036] S202, determining current health status based on historical physiological indicators and dietary data; The current health status can be a quantitative assessment result based on historical physiological indicators and dietary data output.

[0037] Specifically, changes in body fat percentage, fluctuations in uric acid levels, basal metabolic rate, and exercise expenditure within a preset time period are extracted from historical physiological indicators. Simultaneously, the spatiotemporal distribution of carbohydrate intake, purine intake, and fat caloric value is extracted from dietary data. A temporal attention mechanism is then used to analyze the time-lagged association between historical physiological indicators and dietary data. A spatial feature interaction layer is combined to capture multivariate dose-response relationships. Classic physiological formulas such as the Cunningham equation are then integrated to embed physical constraints on metabolic parameters. This constructs a spatiotemporal attention neural network, which inputs a matrix of historical physiological indicators and dietary data. First, in the temporal dimension, the nonlinear association between changes in body fat percentage and carbohydrate intake is analyzed to capture the rate of increase when carbohydrate intake exceeds body fat growth, i.e., the variable insulin sensitivity threshold effect. Second, in the spatial dimension, a dose-response surface is constructed for uric acid fluctuations and purine intake to quantify the cumulative impact of a high-purine diet on uric acid levels, i.e., the time-lagged effect. Finally, the Cunningham equation is used to calculate, integrating basal metabolic rate and exercise expenditure to generate a current metabolic level feature vector. Finally, the weight distribution of the current metabolic level feature vector is adjusted based on age and gender, thereby outputting the final current health status.

[0038] S203, determining restaurant menu constraints based on environmental data and dietary data; Restaurant menu constraints may refer to a set of mathematical constraints formed by nutritional rules, user selection preferences, a menu of dishes for sale, and food inventory status.

[0039] Specifically, based on environmental data: first, parse the restaurant's real-time menu API interface to extract the list of dishes on sale and the inventory status of ingredients; second, based on the analysis of user dietary preferences in dietary data, generate the user's selection preferences; then, according to the nutritional rules of the "Dietary Guidelines for Chinese Residents" and based on the dietary balance index theory in clinical nutrition, by establishing a linear equation group of nutrients and food components, the daily recommended intake in the "Dietary Guidelines for Chinese Residents" is converted into a multidimensional constraint space with boundary conditions, realizing a structured mapping of text rules to mathematical inequality systems, thereby converting text constraints into mathematical constraints; finally, through user physical fitness test data and preference questionnaires, the list of dishes on sale and the inventory status of ingredients, the user's selection preferences and mathematical constraints are combined through a multi-objective optimization algorithm to generate restaurant menu constraints.

[0040] S204: Generate a food combination plan and an exercise recommendation plan based on the current health status and restaurant menu constraints and send them to the user device.

[0041] The food combination plan can be the optimal solution set obtained by decomposing and screening the nutritional components of candidate dishes.

[0042] The exercise recommendation plan may be a gradient exercise instruction generated based on exercise consumption and the amount of food combination to be consumed.

[0043] The user equipment may be a terminal device used by the user to receive and execute a food combination plan and an exercise recommendation plan.

[0044] Specifically, a mixed integer programming algorithm is used to solve food combination plans and exercise recommendation plans within the constraints of the restaurant menu: first, the nutritional components of the candidate dishes are broken down to screen food combination plans that are in line with the current health status; second, based on the user's exercise consumption and the amount of food combination to be consumed, an exercise recommendation plan is generated; finally, the final food combination plan and exercise recommendation plan are encapsulated into a JSON data packet and transmitted through an adaptive compression algorithm that matches the user's device network status.

[0045] This solution uses dual channels of biosensors and medical devices to acquire historical physiological indicators, improving the completeness of data dimensions. Using convolutional neural networks to analyze pre- and post-meal images, this solution processes soup-based foods to generate dietary data, helping to accurately calculate the oil content and purine concentration in soups and reduce calorie estimation errors. Combining GIS geographic information devices and device positioning data, it retrieves real-time menu data from nearby restaurants, accessibility indicators, and seasonal food availability to construct environmental data, helping to reduce the failure rate caused by regional supply differences. Analyzing the nonlinear relationship between changes in body fat percentage and carbohydrate intake captures the variable insulin sensitivity threshold effect, helping to automatically trigger the metabolic level correction coefficient. A dose-response surface is established to link uric acid level fluctuations with purine intake measurements, quantifying the cumulative impact of a high-purine diet on uric acid levels, namely the hysteresis effect, helping to improve the accuracy of gout risk warnings. The weight distribution of the current metabolic level feature vector is adjusted based on age and gender to output the final current health status, helping to reduce errors in calculating daily calorie requirements. Parsing a restaurant's real-time menu API to extract the list of dishes on sale and the inventory status of ingredients helps eliminate plan failures caused by sudden ingredient shortages and improve the feasibility of food combination plans. Based on the analysis of user dietary preferences in dietary data, user selection preferences are generated, helping to improve the taste compatibility of food combination plans. According to the nutritional rules of the "Dietary Guidelines for Chinese Residents," text constraints are converted into mathematical constraints, helping to improve the rate of nutritional balance. A multi-objective optimization algorithm is used to generate restaurant menu constraints based on the list of dishes on sale, the inventory status of ingredients, user selection preferences, and mathematical constraints, helping to improve the acceptance of food combination plans. The nutritional components of candidate dishes are broken down to select food combination plans that meet current health conditions, helping to balance nutritional compatibility, user preference matching, and restaurant inventory feasibility, shortening the time required to generate plans. Based on the user's exercise consumption and the amount of food combination to be consumed, exercise recommendations are generated, helping to improve calorie balance. An adaptive compression algorithm is used to match the user's device network status for transmission, ensuring the successful loading of food combination plans and exercise recommendations in weak network scenarios.

[0046] In some embodiments, basic body information is obtained through biosensors; currently recorded physical examination records are obtained, analyzed, and advanced body information is determined; basic body information and advanced body information are parsed to determine historical physiological indicators; device positioning information is obtained, and the current location is determined based on the device positioning information; GIS data is obtained, and environmental data is determined based on the GIS data and the current location; historical dining records are obtained, and the historical dining records are parsed to obtain dietary data.

[0047] The biosensor can be a wearable device that integrates an impedance-based body fat detection unit and an optical heart rate sensor.

[0048] The basic body information may be metabolic parameter information directly measured by a biosensor.

[0049] The current record may be a real-time data stream generated in the user's mobile terminal during the current period.

[0050] Physical examination records can be standardized health report documents issued by medical institutions.

[0051] Advanced physical information may be metabolic characteristic information obtained by performing structured analysis on physical examination records.

[0052] The device positioning information can be the original geographic coordinate data output by the GPS+Beidou dual-mode of the smart terminal.

[0053] The current location may be the user's actual geographic coordinates.

[0054] GIS data can be spatial attribute data provided by geographic information equipment.

[0055] Historical dining records can be digital records of the user's dietary behavior over a period of time in the past. This period of time can be taken based on experience or determined by thought.

[0056] Specifically, a wearable biosensor array collects basic physiological information such as the user's basal metabolic rate, changes in body fat percentage, and heart rate in real time at a constant sampling frequency. Furthermore, the device connects to the medical institution's electronic health record device and securely accesses the user's current medical examination record via the OAuth2.0 protocol. Natural language processing techniques are used to parse the report text and extract advanced physiological information such as blood uric acid concentration, fasting blood glucose level, and thyroid function indicators. Subsequently, the basic and advanced physiological information are time-aligned, and cubic spline interpolation is used to fill data gaps to generate historical physiological indicators that include characteristics such as changes in body fat percentage, fluctuations in uric acid levels, and basal metabolic rate. Furthermore, GPS and Beidou dual-mode positioning is used to obtain the user's longitude and latitude coordinates. The device's longitude and latitude coordinates are collected in real time, and geofencing technology is used to identify the user's permanent location. Discrete coordinate points are clustered into a semantically labeled current location. Finally, the device connects to a city-level GIS data platform to perform spatial buffer analysis based on the current location to extract environmental data. Subsequently, a deep residual network was used to parse the historical dining records in the user's mobile terminal, and a convolutional neural network was used to identify the ingredient composition in the dish images. The purine intake and carbohydrate intake of each meal were calculated in combination with the national food composition table. Finally, a multi-scale feature pyramid network was introduced to enhance the fat texture representation capability, and a dynamic anchor frame generation strategy was adopted in the region proposal network. The HSV color space segmentation algorithm was combined to optimize the fat area boundary detection accuracy, thereby constructing an improved R-CNN model. The improved R-CNN model was then used to identify the visible fat coverage area in the diet image to generate the final diet data.

[0057] This solution uses a wearable biosensor array to collect basic user information in real time at a constant sampling rate, dynamically capturing metabolic parameters and reducing the need for basal metabolic rate measurements. By integrating with electronic health records at medical institutions, users' current medical records can be securely accessed through the OAuth2.0 protocol to extract advanced information. This helps overcome the limitations of relying solely on basal metabolic data and accurately quantify user-specific metabolic bottlenecks. Time-aligning basic and advanced information generates historical physiological indicators, helping to capture unrecognizable nonlinear correlations. GPS and Beidou dual-mode positioning obtains the user's latitude and longitude coordinates to determine the current location, ensuring a store-level spatial resolution for restaurant accessibility assessments. Connecting to a city-level GIS data platform extracts environmental data and constructs a dynamic supply matrix with time-decay effects, improving the accuracy of restaurant menu availability predictions. A deep residual network is used to analyze historical dining records stored on users' mobile devices, improving the accuracy of ingredient recognition for undocumented dishes. An improved R-CNN model is constructed to generate final dietary data, helping to reduce errors in fat intake estimation.

[0058] In some embodiments, historical physiological indicators are analyzed to determine changes in body fat percentage, fluctuations in uric acid levels, basal metabolic rate, and exercise consumption within a preset time period; dietary data are analyzed to determine carbohydrate intake and purine intake; changes in body fat percentage and carbohydrate intake are analyzed to determine correlation; fluctuations in uric acid levels and purine intake are analyzed to determine response relationships; the current metabolic level is determined based on the correlation, response relationship, exercise consumption, and basal metabolic rate; user information is obtained; user information includes age and gender; and the current health status is determined based on the current metabolic level, age, and gender.

[0059] The body fat percentage change can be the continuous fluctuation of the user's body fat percentage relative to the initial value within a preset time period.

[0060] The fluctuation of uric acid value can be the abnormal change trajectory of uric acid concentration within a preset time period.

[0061] The basal metabolic rate may be the minimum energy consumption value required to maintain life activities at rest.

[0062] Exercise expenditure can be the total energy expenditure generated by physical activity.

[0063] Carbohydrate intake can be the total amount of carbohydrate intake parsed from the user's diet data.

[0064] Purine intake measurement can be the total amount of purine compounds accumulated based on the analysis results of the ingredients of the dish.

[0065] Correlation can be the statistical strength of the association between changes in body fat percentage and carbohydrate intake in a time series.

[0066] The response relationship can be a dynamic mathematical expression of the response of changes in purine intake to fluctuations in uric acid levels.

[0067] The current metabolic level can be a comprehensive evaluation indicator of the user's real-time metabolic status.

[0068] User information can be individual attribute parameters that affect metabolic assessment.

[0069] Specifically, based on historical physiological indicators, the team first calculated the absolute value of the first-order difference of body fat percentage between adjacent sampling points to obtain changes in body fat percentage. Secondly, a range algorithm was constructed based on the clinical laboratory reference range threshold and a time window sliding mechanism. This range algorithm was then used to chronologically intercept the uric acid test value sequence and calculate the difference between the maximum and minimum values ​​to determine uric acid level fluctuations. Then, based on resting heart rate, age, gender, weight, and height, the Harris-Benedict equation was established using the "body surface area rule" defined in the International Federation of Sports Medicine (FIMS) "Energy Metabolism Assessment Standards." Basal metabolic rate was statically calculated using the Harris-Benedict equation. Finally, a metabolic equivalent conversion model was constructed based on the standardized metabolic equivalent comparison table defined in the American College of Sports Medicine's "Guidelines for Exercise Testing and Prescription." This model then used the acceleration sensor data to generate exercise expenditure within a preset time period. Furthermore, dietary data output by the improved R-CNN model was used to establish a mapping relationship with national food composition tables to calculate carbohydrate and purine intake for a single meal. Subsequently, changes in body fat percentage were time-aligned with carbohydrate intake during the corresponding time period. Furthermore, based on the lag effect theory of dietary-metabolic response in clinical nutrition, a random forest regression algorithm was constructed. This algorithm then modeled the nonlinear relationship between the two, outputting feature importance weights and directional parameters to determine correlation. Secondly, using uric acid fluctuations as the dependent variable and purine intake as the independent variable, a lagged variable analysis method was used to determine the optimal response delay. A generalized additive model was established based on the nonlinear dose-response analysis framework defined in the "Guidelines for Modeling Environmental Exposure-Health Effects" published by the International Association for Medical Statistics. The dose-effect curves were fitted using the generalized additive model, and the purine metabolic sensitivity coefficient was calculated to determine the response relationship. A fully connected neural network was then constructed based on the multi-layer perceptron feature fusion mechanism defined in the "Technical Specifications for Metabolic Prediction Models" of the International Society for Biomedical Engineering. The correlation and response relationship were then input into a pretrained fully connected neural network. The current mean basal metabolic rate and cumulative exercise expenditure were also superimposed. The feature fusion layer then outputs the current metabolic level, including the metabolic efficiency index, energy deficit estimate, and metabolic imbalance risk probability. User information was extracted from encrypted user profiles. The user's age and gender are then extracted from the user information. Gender is encoded as a one-hot vector, normalized, and concatenated with the current metabolic level to form a multidimensional feature vector. This multidimensional feature vector is then input into a health status classifier built on the framework, which is pre-loaded with metabolic baseline thresholds for different genders and age groups. By comparing the deviation of the metabolic efficiency index from the metabolic baseline threshold and combining it with the metabolic imbalance risk probability weight, the current health status is output, including a health level label and risk type description.

[0070] This solution uses historical physiological indicators to determine changes in body fat percentage, uric acid level fluctuations, basal metabolic rate, and exercise expenditure within a preset time period. This helps quantify the dynamic characteristics of a user's metabolic regulation ability and overcomes the technical bottleneck of characterizing the insulin sensitivity threshold effect. It also helps accurately locate risk periods for abnormal purine metabolism, addressing the problem of delayed detection of metabolic indicator mutation events, eliminating baseline distortion caused by individual differences in circadian metabolic rhythms, and avoiding missed detection of non-gait movements. Determining carbohydrate intake and purine intake based on dietary data helps eliminate calorie estimation bias caused by the inability to identify dietary residues and overcomes causal misjudgment caused by data asynchrony. Determining correlations between changes in body fat percentage and carbohydrate intake helps capture overlooked insulin sensitivity threshold effects. Determining response relationships based on uric acid level fluctuations and purine intake helps identify individual metabolic hysteresis effects. Determining current metabolic levels based on correlations, response relationships, exercise expenditure, and basal metabolic rate helps eliminate the risk of misjudgment due to mutual interference and overcomes the limitations of relying solely on single data. User information is extracted from encrypted user profiles, and then the user's age and gender are extracted from this information. This eliminates metabolic rate deviations caused by physiological aging differences among peers. Combined with the encoding of hormone regulatory factors, sex-specific lipid metabolism pathways are distinguished. Current health status is determined based on current metabolic level, age, and gender, facilitating the dynamic generation of sex-differentiated health boundary conditions.

[0071] In some embodiments, restaurant information is determined based on environmental data and GIS data; dietary data is analyzed to determine user dietary preferences; and restaurant menu constraints are determined based on preset nutritional rules, user dietary preferences, and restaurant information.

[0072] Restaurant information can be real-time dynamic data sets such as the restaurant's geographic location, supply chain data, and operational status.

[0073] User dietary preferences can be a multi-dimensional feature set of user dietary tendencies.

[0074] The preset nutrition rules can be a set of computer-executable constraints established based on clinical medical guidelines and nutrition standards, which are pre-stored in the server and called when used.

[0075] Specifically, the geographic fence coordinates in the GIS data are parsed, and a list of restaurants near the user's current location is obtained through the AutoNavi Map API. The merchant's POS device is connected in real time to obtain the daily food inventory list. Combined with the meteorological elements and holiday signs in the environmental data, the LSTM network is used to predict the current trend of changes in the supply of dishes in each restaurant. Based on the road network data constructed by the node, path and spatial relationship triples contributed by the user, the coordinates of the road network intersections are abstracted into graph nodes. The XML format raw data from which the coordinate data is derived is parsed to determine the algorithm used to calculate the real-time reachable path from the user to each restaurant. The estimated arrival time is dynamically corrected in combination with the traffic flow data, and the restaurant information of candidate restaurants that match the time window is screened out. Historical order records are extracted from dietary data for time series analysis: First, users' historical dining records are converted into an ingredient frequency matrix. A time decay factor is introduced to enhance recent preferences, and a hierarchical clustering algorithm is constructed to identify clusters of frequently consumed ingredients. Second, based on the distributed semantics assumption, a Skip-gram architecture is used to maximize the co-occurrence probability of contextual dishes within a sliding window of the dish sequence, generating a Word2Vec model. The taste preference embedding space based on the Word2Vec model is then used to mine user dietary preferences using pre-defined nutritional rules. Clinical guidelines are then converted into computable constraints. Restaurant information is parsed using optical character recognition (OCR) and natural language processing (NLP) techniques to construct a dynamic available dish map and implement inventory-driven menu item status tagging. Furthermore, a Pareto frontier search algorithm is designed to simultaneously optimize three objective functions: nutritional value, user preference alignment, and restaurant operational feasibility. Finally, a taboo search mechanism is introduced to eliminate conflicting combinations, thereby generating restaurant menu constraints.

[0076] This solution identifies restaurant information based on environmental and GIS data, effectively resolving the issue of recommendation failures caused by ignoring regional accessibility, improving the geographic compatibility of the candidate restaurant set, and enhancing the practical implementation rate of the recommendation solution. Dietary data is analyzed to determine user dietary preferences, overcoming preference misjudgments caused by relying solely on statistical frequency, improving the accuracy of identifying complex preferences, and avoiding preference drift errors caused by long-term data accumulation. Restaurant menu constraints are determined based on pre-set nutritional rules, user dietary preferences, and restaurant information. This helps balance nutritional standards, user preference scores, and restaurant operational constraints, improving the overall acceptability of recommendation solutions.

[0077] In some embodiments, historical dining records are parsed to determine text information and picture information; text information is analyzed to determine food names, food portions, and meal times; picture information is analyzed to determine ingredient intake; and dietary data is obtained based on food names, food portions, meal times, and ingredient intake.

[0078] The text information may be text records related to eating behavior.

[0079] The picture information may be a visual record of the meal captured by a photographic device.

[0080] The name of the food can be a standardized meal label.

[0081] A portion size can be a quantitative measure of food mass or volume.

[0082] The eating time may be the time point when the user actually consumes food.

[0083] Ingredient intake can be a quantitative value of a nutrient element.

[0084] Specifically, based on historical dining records, an OCR engine is used to perform text recognition on paper menu photos to extract the menu's text information. Based on international metadata standards, a model is extracted and used to parse image metadata, extracting the shooting timestamp and aligning it with the user's manually entered meal time to determine the image information. A BERT-Base Chinese pre-trained model, using a Transformer encoder architecture, integrates the standard names from the "Chinese Food Composition Table" with a mapping table of local colloquial names to identify food names. Conditional random fields are used to parse modifiers to generate food portions. Relative time descriptions are mapped to absolute time intervals to determine meal times. A pre-trained model based on a diet dataset using convolutional neural network feature extraction outputs dish categories and confidence scores. Food volume is calculated through reference object detection and converted to weight using a density table. A residual network is trained to detect the area percentage of soup residue, thereby generating ingredient intake. If the discrepancy between the textual information and the fraudulent information is too high, a manual review process is triggered. A time decay factor is used to dynamically adjust the weight of historical data. Furthermore, the Chinese Food Composition Table database is accessed to establish an ingredient-nutrient mapping matrix, allowing for reverse engineering of complex dish recipes to generate the final diet data.

[0085] Through this solution, based on historical dining records, the OCR engine is called to perform text recognition on photos of paper menus, extract the text information of the menu, eliminate recognition errors caused by illegible handwriting, and solve the problem of users easily missing dish names when manually entering; by parsing image metadata, the shooting timestamp is extracted and aligned with the meal time manually entered by the user to determine the image information, which helps to eliminate the subjective bias of manual time recording. Based on the text information, the food name, food portion and meal time are determined, which helps to establish computable time series data. Based on the image information, the ingredient intake is determined, which solves the quality estimation error caused by irregular shape and reduces the error in fat intake estimation. Based on the food name, food portion, meal time and ingredient intake, dietary data is obtained, which helps to solve the problem of disconnection between metabolic indicators and nutritional elements.

[0086] In some embodiments, the image information is analyzed to determine the pre-meal picture and the post-meal picture; based on the pre-meal picture and the post-meal picture, the food state change and the food type are determined; based on the food type and the pre-meal picture, the soup ratio, soup ingredients and food ingredients are determined; based on the food state change and the soup ratio, the food intake amount and soup intake amount are determined; based on the food intake amount, soup intake amount, soup ingredients and food ingredients, the ingredient intake is determined.

[0087] The before-meal picture may be an image of a complete plate taken before the meal.

[0088] The post-meal picture may be an image of the remaining state of a plate taken after a meal is finished.

[0089] The change in food state can be the difference in the physical quantity of food in the plate before and after eating.

[0090] Food types can be standardized names determined according to the coding system of the "Chinese Food Composition Table".

[0091] The soup-water ratio may be the volume ratio of liquid soup in the container.

[0092] The composition of soup can be a quantitative indicator of the nutrients contained in the liquid part.

[0093] Food composition can be a parameter that measures the amount of nutrients in a solid food.

[0094] Food intake can be the actual weight of solid food ingested.

[0095] The soup intake amount may be the actual volume of liquid soup consumed.

[0096] Specifically, the shooting timestamp in the parsed image information is called and time-series matched with the user-entered meal time. The correspondence between the pre-meal and post-meal images is determined through a time window sliding algorithm. Furthermore, a residual network based on the architecture extracts global feature vectors for the pre-meal and pre-meal images. The rate of change of the food region is calculated using cosine similarity. An instance segmentation model built based on the target detection framework identifies the distribution of residues on the plate and constructs food state changes. Standardized food names are then identified based on the pre-meal images. For complex dishes, a graph attention network is used to analyze the spatial distribution of ingredients and determine the food type. The liquid region of the pre-meal image is segmented using HSV color space. A U-Net model constructed by combining jump connections to fuse shallow detail features with deep semantic features generates a soup area mask to calculate the soup ratio. Simultaneously, the residual network detects the sediment distribution of the soup base residue to determine the soup composition. Then, based on the food type, the Chinese food composition table is called to obtain baseline nutritional data and determine the food composition. The actual intake, or food intake, is calculated by combining the area reduction rate of the food state change. Furthermore, based on the soup-water ratio, a three-dimensional spatial mapping is established using reference objects such as standard spoons. A density conversion table is used to convert volume to weight, and the fat content is corrected based on the soup residue test results to determine the soup intake amount. Finally, based on the soup and food ingredients, an ingredient-nutrient mapping matrix is ​​established. A weighted summation of the food intake and soup intake is performed to output the ingredient intake.

[0097] Through this solution, the pictures before and after the meal are determined based on the picture information, which helps to eliminate the interference of users mistransmitting pictures. Based on the pictures before and after the meal, the changes in food status and food type are determined to avoid ingredient deviations caused by overall estimation. Based on the food type and the pictures before the meal, the soup ratio, soup ingredients and food ingredients are determined, which helps to accurately distinguish the suspended fat layer and the bottom sediment in the soup, and improve the accuracy of liquid food oil content detection. Based on the changes in food status and the soup ratio, the amount of food consumed and the amount of soup consumed are determined, which helps to improve the consistency of the estimation of metabolism-related components and reduce the error in the calorie estimation of soups. Based on the amount of food consumed, the amount of soup consumed, the soup ingredients and food ingredients, the ingredient intake is determined, which helps to improve the coverage of nutritional constraints.

[0098] In some embodiments, candidate dishes are determined based on the constraints of the restaurant menu; candidate dishes are screened based on the current health status to obtain a food combination plan; based on the food combination plan, the ingredient intake is analyzed to determine the amount to be consumed; based on historical physiological indicators, the user's daily exercise volume is estimated; and based on the user's daily exercise volume and the amount to be consumed, an exercise recommendation plan is determined.

[0099] The candidate dishes may be a set of dishes that can actually be prepared at present.

[0100] The amount to be consumed may be based on the difference between the intake of ingredients in the food combination plan selected by the user and the health goal.

[0101] The user's daily exercise volume can be the sum of the user's basal metabolic consumption and active exercise consumption.

[0102] Specifically, the system accesses the restaurant's real-time menu database to extract the available ingredients and dish preparation information for the day. Then, based on the ingredient composition in the dish preparation information and the restaurant's menu constraints, a multidimensional constraint matrix is ​​constructed. A constraint satisfaction algorithm, based on constraint satisfaction problem theory in operations research, is then used to perform a feasibility analysis on the multidimensional constraint matrix, outputting candidate dishes that meet the current restaurant's operational status. A health risk filtering rule base is then established based on the user's current health status and the purine content, fat content, and allergen markers in the food ingredients. The candidate dishes are then input into the health risk filtering rule base, and pharmacological contraindications and metabolic index thresholds are matched item by item. Exceeding the standard thresholds are then eliminated to generate a food combination plan. Furthermore, based on the food type of each dish in the food combination plan, the Chinese Food Composition Table is used to obtain nutritional data per unit weight. Based on the user's food and soup intake, a retrospective dietary assessment method in nutrition science is used to construct a personal intake habit model using the user's historical food intake data to predict the intake portion. The predicted intake portion is then multiplied by the unit nutritional data to accumulate the total ingredient intake, which is then compared to the user's health goals to calculate the remaining consumption. Based on historical physiological indicators, a time series analysis model is established by analyzing the time series data of users' historical physiological indicators using an autoregressive integral sliding average to extract the cyclical characteristics of exercise behavior. Combined with weather data, a multivariate linear regression-based exercise volume prediction function is constructed. This outputs the user's expected basal metabolic rate and active exercise consumption, which are combined into the user's daily exercise volume. Based on the expected consumption and the user's daily exercise volume, the target value for additional energy consumption is calculated. Then, a table of energy consumption for each sport is used, combined with the user's exercise preferences, to generate multiple sets of exercise duration combinations. Finally, a multi-objective optimization algorithm is used to balance exercise duration, intensity, and user tolerance, outputting a Pareto optimal solution set and generating exercise recommendations based on execution priority.

[0103] Through this solution, candidate dishes are determined based on the constraints of the restaurant menu, ensuring that the recommendation plan is strictly synchronized with the actual operating status of the restaurant, and breaking through the problem of infeasibility of recommendations caused by ignoring regional differences in food supply. Based on the current health status, candidate dishes are screened and food combination plans are determined, which helps to eliminate pharmacological contraindications and avoid misjudgment of health risks caused by the inability to handle nonlinear responses of metabolic indicators. Through the food combination plan, the intake of ingredients is analyzed and the amount to be consumed is determined, which helps to accurately quantify the nutritional contribution of food intake and soup intake, and reduce the error in single meal calorie estimation caused by the lack of soup residue detection. Based on historical physiological indicators, the user's daily exercise volume is estimated, which helps to eliminate the energy balance calculation deviation caused by ignoring the dynamic changes in individual behavior. Determining the exercise recommendation plan based on the user's daily exercise volume and the amount to be consumed helps to improve the actual acceptance rate of the exercise recommendation plan.

[0104] In some embodiments, data feedback from a user device is received, the data feedback is analyzed, and a solution deviation is determined; based on the solution deviation, a health status change and solution acceptance are determined; and based on the health status change and solution acceptance, a recommendation strategy is modified.

[0105] The data feedback can be the actual execution data uploaded by the user through the device.

[0106] Scenario deviation can be the quantitative difference between the actual execution results and the recommended solution.

[0107] A change in health status may be a change in physiological status.

[0108] Solution acceptance can be the user's compliance with the recommended solution.

[0109] The recommendation strategy can be a dynamically generated personalized health management plan.

[0110] Specifically, the local database of the user's device is called to obtain real-time data such as actual diet records, recommended exercise plan execution data, and health monitoring indicators. The actual diet records are compared with the expected intake in the food combination plan to determine calorie deviation and purine excess. The exercise execution data is compared with the target duration and intensity in the recommended exercise plan to quantify the deviation in exercise duration and heart rate intensity. The time series analysis model is constructed using the stabilization processing and white noise test of historical health indicators to detect abnormal fluctuations in health monitoring indicators, thereby determining plan deviations. The weekly fluctuations in body fat percentage and the increase in uric acid value are input into the pre-trained metabolic response model based on the random forest algorithm to output changes in health status. At the same time, based on the user's actual execution data, the acceptance of the food combination plan and the exercise recommendation plan are statistically analyzed, and the plan acceptance is generated by weighted average. The recommendation strategy is revised based on changes in health status and acceptance of the plan: first, if the health status changes too much, the health risk filtering rule library is called to tighten the purine threshold and reduce the fat energy supply ratio; second, based on the plan deviation, the output value of the original heart rate signal collected by the user device after Kalman filtering is used to update the user's exercise preference weight through the reinforcement learning algorithm. When the heart rate intensity continues to decrease, the recommended exercise intensity level is lowered; then, when the acceptance of the plan is too low, the restaurant's real-time menu database is re-called, and the user's taste tendency is used as the optimization goal. The taboo search algorithm determined by clinical research data based on the pharmacological contraindication safety margin is used to generate alternative plans among the candidate dishes.

[0111] Through this solution, receiving data feedback from user devices helps eliminate the problem of delayed metabolic indicator monitoring caused by reliance on manual data entry and establishes dynamic health status tracking capabilities; analyzing data feedback, determining plan deviations, quantifying the degree of deviation between the user's actual intake and the recommended plan, overcoming errors in calorie estimation caused by soup residue, improving the accuracy of intake assessment, and eliminating the defect of ignoring differences in user behavior due to metabolic consumption. Based on plan deviations, determine changes in health status and plan acceptance, and improve the problem of insufficient plan acceptance caused by ignoring subjective preferences. Based on changes in health status and plan acceptance, revise the recommendation strategy, eliminate the problem of missing health warnings that cannot respond to the dynamic relationship between uric acid levels and purine intake, reduce the execution failure rate due to inappropriate exercise intensity, and avoid the accumulation of health risks.

[0112] In some embodiments, food selection weights are determined based on the acceptance of the plan; intake constraints are determined based on changes in health status; user selection preferences are determined based on the food selection weights; optional food types are determined based on the intake constraints; and recommendation strategies are modified based on selection preferences and optional food types.

[0113] The food selection weight may be a priority coefficient of the food in the recommendation strategy.

[0114] Intake constraints may be nutrient intake restriction rules.

[0115] Selection preferences can be a set of tendency features in users' dietary decisions.

[0116] The optional food type may be a valid food set formed after screening by the intake constraint conditions.

[0117] Specifically, based on the proposal's acceptance, the historical preference reinforcement rule is activated. Based on the user's historically high-frequency dish selections, the corresponding food selection weights are increased. Simultaneously, a taste preference mining algorithm is activated to extract user preferences from manually annotated "like or dislike" tags and restaurant menu click logs, dynamically updating food selection weights. Based on changes in health status, a nutritional taboo rule library for corresponding diseases is loaded. Constraints are encoded as linear inequalities and integrated into a recommendation strategy optimization model. This model uses taste matching and nutritional deviation as dual optimization objectives, then employs a taboo search algorithm to globally optimize within the feasible solution space of the intake constraints to generate deterministic intake constraints. Furthermore, based on the food selection weights, the user's historically high-frequency dish categories are counted. Leveraging collaborative filtering's neighborhood modeling concept, an improved weighted matrix factorization collaborative filtering algorithm is employed to mine implicit preference features and generate a taste preference vector. The historical selection frequency and taste preference vector are then combined to output the user's preference. Next, it accesses the restaurant's real-time menu database to obtain the day's available dishes and nutritional data. Based on intake constraints, it performs hard filtering to eliminate dishes that exceed the recommended intake limit. Furthermore, it uses the user's geographic location to activate regional ingredient substitution rules to determine the available food types. Finally, based on the user's preferences and available food types, it refines the recommendation strategy. First, it inputs ingredient selection weights and available food types into the taboo search algorithm, optimizing for maximizing taste matching and minimizing nutritional bias. Second, it allows for substitution of single taboo dishes or adjustments to similar ingredient combinations to generate alternatives that meet intake constraints. Finally, when inventory constraints are triggered, it prioritizes dishes with sufficient inventory.

[0118] Through this solution, the food selection weights are determined according to the acceptance of the solution, accurately representing the actual degree of user compliance with the recommended solution, and realizing dynamic adaptation of the food selection weights to the real needs of users. Intake constraints are determined according to changes in health status, realizing real-time early warning and prevention of health risks that cannot be achieved. According to the food selection weights, the user's selection preferences are determined, which not only breaks through the limitation of relying solely on the user to manually set preferences, but also improves the user's subjective acceptance of the recommendation strategy under the premise of meeting nutritional constraints. According to the intake constraints, the optional food types are determined to ensure that the candidate dishes meet both health constraints and actual supply conditions. According to the selection preferences and optional food types, the recommendation strategy is revised, which helps to achieve the difficult-to-balance personalization and safety goals simultaneously, and improves the problem of insufficient acceptance of the solution.

[0119] Figure 3 This is a structural diagram of an artificial intelligence-based health management system provided in one embodiment of the present application, such as Figure 3 As shown, the artificial intelligence-based health management system 300 of this embodiment includes: a data acquisition module 301, a state determination module 302, a condition determination module 303, and a plan generation module 304.

[0120] The data acquisition module 301 is used to obtain historical physiological indicators, dietary data and environmental data; the status determination module 302 is used to determine the current health status based on the historical physiological indicators and the dietary data; the condition determination module 303 is used to determine the restaurant menu constraints based on the environmental data and the dietary data; the plan generation module 304 is used to generate a food combination plan and an exercise recommendation plan based on the current health status and the restaurant menu constraints and send them to the user device.

[0121] Optionally, when the data acquisition module 301 acquires historical physiological indicators, dietary data and environmental data, it is used to: acquire basic body information through biosensors; acquire currently recorded physical examination records, analyze the physical examination records, and determine advanced body information; parse the basic body information and the advanced body information to determine historical physiological indicators; acquire device positioning information, and determine the current location based on the device positioning information; acquire GIS data, and determine environmental data based on the GIS data and the current location; acquire historical dining records, parse the historical dining records, and obtain dietary data.

[0122] Optionally, when the status determination module 302 determines the current health status based on the historical physiological indicators and the dietary data, it is used to: analyze the historical physiological indicators to determine the changes in body fat percentage, fluctuations in uric acid levels, basal metabolic rate and exercise consumption within a preset time period; analyze the dietary data to determine the carbohydrate intake and purine intake; analyze the changes in body fat percentage and the carbohydrate intake to determine the correlation; analyze the fluctuations in uric acid levels and the purine intake to determine the response relationship; determine the current metabolic level based on the correlation, the response relationship, the exercise consumption and the basal metabolic rate; obtain user information; the user information includes age and gender; determine the current health status based on the current metabolic level, the age and the gender.

[0123] Optionally, when the condition determination module 303 determines the restaurant menu constraints based on the environmental data and the dietary data, it is used to: determine the restaurant information based on the environmental data and the GIS data; analyze the dietary data to determine the user's dietary preferences; and determine the restaurant menu constraints based on preset nutritional rules, the user's dietary preferences and the restaurant information.

[0124] Optionally, when the data acquisition module 301 parses the historical dining records and obtains dietary data, it is used to: parse the historical dining records to determine text information and picture information; analyze the text information to determine the food name, food portion and meal time; analyze the picture information to determine the ingredient intake; and obtain dietary data based on the food name, food portion, meal time and ingredient intake.

[0125] Optionally, when the data acquisition module 301 analyzes the image information and determines the ingredient intake, it is used to: analyze the image information to determine the pre-meal picture and the post-meal picture; determine the food state change and the food type based on the pre-meal picture and the post-meal picture; determine the soup ratio, soup ingredients and food ingredients based on the food type and the pre-meal picture; determine the food intake and soup intake based on the food state change and the soup ratio; determine the ingredient intake based on the food intake, the soup intake, the soup ingredients and the food ingredients.

[0126] Optionally, when the plan generation module 304 generates a food combination plan and an exercise recommendation plan based on the current health status and the restaurant menu constraints, it is used to: determine candidate dishes based on the restaurant menu constraints; filter the candidate dishes based on the current health status to obtain a food combination plan; analyze the ingredient intake based on the food combination plan to determine the amount to be consumed; infer the user's daily exercise volume based on the historical physiological indicators; and determine an exercise recommendation plan based on the user's daily exercise volume and the amount to be consumed.

[0127] Optionally, the artificial intelligence-based health management system also includes a policy modification module 305, which is used to: receive data feedback from the user device, analyze the data feedback, and determine the solution deviation; determine the health status change and solution acceptance based on the solution deviation; and modify the recommended strategy based on the health status change and solution acceptance.

[0128] Optionally, when the strategy modification module 305 modifies the recommendation strategy according to the change in health status and the acceptance of the plan, it is used to: determine the food selection weight according to the acceptance of the plan; determine the intake constraints according to the change in health status; determine the user's selection preference according to the food selection weight; determine the optional food type according to the intake constraints; and modify the recommendation strategy according to the selection preference and the optional food type.

[0129] The system of this embodiment can be used to execute the method of any of the above embodiments. Its implementation principles and technical effects are similar and will not be described in detail here.

Claims

1. A health management method based on artificial intelligence, characterized in that: include: Obtain historical physiological indicators, dietary data, and environmental data; determining a current health status based on the historical physiological indicators and the dietary data; determining restaurant menu constraints based on the environmental data and the dietary data; Based on the current health status and the restaurant menu constraints, a food combination plan and an exercise recommendation plan are generated and sent to the user device.

2. The method according to claim 1, characterized in that The acquisition of historical physiological indicators, dietary data, and environmental data includes: Obtain basic body information through biosensors; Obtaining currently recorded physical examination records, analyzing the physical examination records, and determining advanced physical information; Analyzing the basic physical information and the advanced physical information to determine historical physiological indicators; Obtaining device positioning information, and determining the current location based on the device positioning information; Obtaining GIS data, and determining environmental data based on the GIS data and the current location; Obtain historical dining records, analyze the historical dining records, and obtain dietary data.

3. The method according to claim 1, characterized in that Determining the current health status based on the historical physiological indicators and the dietary data includes: Analyze the historical physiological indicators to determine changes in body fat percentage, fluctuations in uric acid levels, basal metabolic rate, and exercise consumption within a preset time period; Analyzing the dietary data to determine carbohydrate intake and purine intake; Analyzing the changes in body fat percentage and the carbohydrate intake to determine correlation; Analyzing the fluctuation of the uric acid level and the purine intake measurement to determine the response relationship; determining a current metabolic level according to the correlation, the response relationship, the exercise consumption, and the basal metabolic rate; Obtain user information; the user information includes age and gender; A current health status is determined based on the current metabolic level, the age, and the gender.

4. The method according to claim 2, characterized in that Determining restaurant menu constraints based on the environmental data and the dietary data includes: determining restaurant information based on the environmental data and the GIS data; Analyzing the dietary data to determine the user's dietary preferences; Restaurant menu constraints are determined based on preset nutritional rules, the user's dietary preferences, and the restaurant information.

5. The method according to claim 2, characterized in that The parsing of the historical dining records to obtain dietary data includes: Analyze the historical dining records to determine text information and picture information; Analyze the text information to determine the food name, food quantity and eating time; Analyzing the image information to determine ingredient intake; Dietary data is obtained according to the food name, the food portion, the eating time and the ingredient intake.

6. The method according to claim 5, characterized in that The analyzing the image information to determine the ingredient intake includes: Analyze the picture information to determine the before-meal picture and the after-meal picture; determining food state changes and food types based on the pre-meal picture and the post-meal picture; Determining the soup ratio, soup ingredients, and food ingredients based on the food type and the pre-eating picture; determining the amount of food intake and the amount of soup intake according to the change in the state of the food and the ratio of the soup to water; The ingredient intake is determined according to the food intake amount, the soup intake amount, the soup ingredients and the food ingredients.

7. The method according to claim 6, characterized in that Generating a food combination plan and an exercise recommendation plan based on the current health status and the restaurant menu constraints includes: Determining candidate dishes according to the restaurant menu constraints; Filter the candidate dishes according to the current health status to obtain a food combination plan; Analyzing the intake of the ingredients based on the food combination plan to determine the amount to be consumed; Estimate the user's daily exercise volume based on the historical physiological indicators; Determine an exercise recommendation plan based on the user's daily exercise volume and the amount to be consumed.

8. The method according to claim 1, characterized in that The method further comprises: receiving data feedback from the user equipment, analyzing the data feedback, and determining a solution deviation; Determine changes in health status and acceptance of the regimen based on the described regimen deviations; Modify the recommended strategy based on the changes in health status and the acceptance of the plan.

9. The method according to claim 8, characterized in that The modification of the recommended strategy according to the change in health status and the acceptability of the plan includes: Determine the weight of food choices based on the acceptability of the plan; determining intake constraints based on the health status change; determining the user's selection preference according to the food selection weight; determining optional food types according to the intake constraints; The recommendation strategy is modified according to the selection preference and the optional food types.

10. An artificial intelligence-based health management system, applied to the method according to any one of claims 1 to 9, characterized in that: include: Data acquisition module, used to obtain historical physiological indicators, dietary data and environmental data; a status determination module, configured to determine a current health status based on the historical physiological indicators and the dietary data; a condition determination module, configured to determine restaurant menu constraint conditions based on the environmental data and the dietary data; A plan generation module is used to generate a food combination plan and an exercise recommendation plan based on the current health status and the restaurant menu constraints and send them to the user device.

Images