AI personalized nutrition evaluation and guidance method based on individual continuous diet data

Abstract

The invention relates to an AI personalized nutrition evaluation and guidance method based on individual continuous diet data. The AI personalized nutrition evaluation and guidance method comprises the following steps: (1) calculating individual single meal food and nutrient intake; individual health indexes are collected, the individual daily energy demand amount is calculated, and the daily food and nutrient demand amount is determined; (2) carrying out instant personalized nutrition evaluation on food and nutrients ingested by an individual in a single meal according to a domestic public recommendation standard; (3) on the basis of individual nutrition and health data continuously recorded in the longitudinal direction of time, establishing a personalized nutrition intake-health model taking an individual as a unit, and carrying out personalized nutrition evaluation and health risk prompting; and (4) simulating the idea of nutrition guidance of professionals to create an AI nutrition guidance algorithm, and generating AI guidance. By adopting the AI personalized nutrition evaluation and guidance method based on the individual continuous diet data, a sustainable AI-assisted diet self-management solution suitable for daily life scenes is effectively provided.

Description

Technical Field

[0001] This invention relates to the field of AI service applications, and more particularly to the field of dietary nutrition guidance services, specifically referring to an AI-based personalized nutrition evaluation and guidance method based on individual continuous dietary data. Background Technology

[0002] Both domestic and international studies have shown that eating out leads to the intake of more fat and salt, lower dietary quality, and a greater risk of obesity. Overweight and obesity can cause a range of health problems, such as type 2 diabetes, cardiovascular disease, cancer, and mental illness.

[0004] In recent years, Chinese researchers have conducted studies on managing blood sugar through information technology, such as sending nutritional information via text messages or providing personalized online dietary guidance from nutritionists. However, these studies all rely on professional personnel, and there are currently no studies on methods for individual dietary management that do not depend on professional personnel. Nevertheless, with changing dietary habits and increasing health needs, the application of "AI-assisted self-management of diet" in my country could be a feasible measure to alleviate the contradiction between the prominent problem of overweight and obesity and the insufficient capacity for nutritional services.

[0005] In the past five years, with the development of AI technology, AI-generated professional opinions have emerged as a new method to guide individual dietary management. Studies show that this approach is effective in weight loss and significantly improves the long-term maintenance of weight loss results, achieving effects similar to those of receiving professional medical consultation. Due to its cost-effectiveness and application demand, some research findings have already been translated into commercial health management services in Europe and the United States.

[0006] In fact, my country already has the basic conditions to realize AI-assisted dietary self-management methods: In the past, dietary data was inaccurate due to the differences in Chinese cooking methods, but now there are standardized developments in catering recipes such as central kitchens, which can obtain more accurate nutritional data; and with the widespread use of smartphones and wireless networks, the general public is familiar with mobile applications, making it practical. Summary of the Invention

[0007] The purpose of this invention is to overcome the shortcomings of the prior art and provide an AI-based personalized nutrition evaluation and guidance method based on individual continuous dietary data.

[0008] To achieve the above objectives, the present invention provides an AI-based personalized nutrition assessment and guidance method based on individual continuous dietary data as follows: The main feature of this AI-based personalized nutrition assessment and guidance method, which is based on individual continuous dietary data, is that the method includes the following steps: (1) Calculate the individual's single meal food and nutrient intake, including dining in restaurants / canteens or non-restaurants / canteens; (2) Collect individual health index indicators and calculate individual daily energy requirements and individual daily food and nutrient requirements; (3) Based on the Chinese Dietary Guidelines and recommended intake standards of dietary nutrients corresponding to the individual's main characteristics, conduct real-time personalized nutritional evaluation of the food and nutrients consumed in a single meal; (4) Based on the experimental population data and the data information of individual longitudinal continuous records, establish a personalized nutrition intake-health effect model based on the individual, generate the current individual nutrition intake-health effect relationship, and extract individual taste and food preference feature information; (5) Based on the personalized nutrition intake-health effect model, conduct personalized nutrition evaluation and personalized health risk warning; and generate corresponding AI nutrition guidance based on the individual's determined nutrition goals, taste and food preferences, as well as the requirements of local customs and dietary habits.

[0009] Preferably, step (1) includes: (1.1) Obtain the food intake of diners in restaurants / canteens: First, collect the cooking recipe information of the restaurant kitchen, including the raw and cooked weights of the main and auxiliary ingredients and seasonings of each dish, and calculate the intake of various foods in a single serving of each dish. Then, record the individual's food order and discard information for each dish during the meal, and combine this with the above recipe information to obtain the individual's food intake for a single meal; (1.2) Obtaining food intake from meals outside the restaurant / canteen: Individuals record the various foods and their weights consumed at each meal; (1.3) Calculate individual nutrient intake per meal as follows: =η1× 2 ... n ； in, This represents the percentage of the nth dish consumed by an individual. This represents the carbohydrate content of a single serving of the nth dish, and so on for other nutrients.

[0010] Preferably, step (2) includes: (2.1) Collect basic individual information: Collect information on individual age, gender, height, weight, physical activity, and percentage of food intake at each of the three meals; (2.2) Calculate the individual's daily energy requirement as follows to determine the individual's daily food requirement: Energy requirement = Basal energy expenditure × Physical activity level; (2.3) Based on the balanced dietary patterns and food quantities set in the Dietary Guidelines for Chinese Residents for different energy requirement levels, determine the daily recommended intake of various types of food for individuals corresponding to different levels of daily energy requirement, and use it as the standard for judging whether the individual's food intake is appropriate; based on the recommended intake of each nutrient for individuals of different ages, genders, physiological conditions and labor intensity in the Dietary Reference Intakes for Chinese Residents, use it as the standard for judging whether the individual's energy and nutrient intake is appropriate.

[0011] Preferably, step (3) includes: (3.1) Evaluation of individual food intake per meal: Based on the proportion of the individual’s three meals, the daily recommended food intake in the Chinese Dietary Guidelines is converted into the recommended food intake per meal according to the proportion; and the actual intake of each individual at each meal is compared with the recommended intake. Based on the comparison results, an evaluation level is determined: exceeding / below the recommended intake range of 0%-20%, 20%-40%, and >40%, which are respectively evaluated as appropriate, excessive / insufficient, and severely excessive / severely insufficient. (3.2) Personalized nutrient intake evaluation for each meal: Based on the individual’s three meals ratio, the recommended daily intake of energy and nutrients in the “Chinese Dietary Reference Intakes” is converted into the recommended intake for each meal according to the ratio; and the actual intake of energy and nutrients for each meal is compared with the recommended intake. Based on the comparison results, an evaluation level is determined: exceeding / below the recommended intake range of 0%-20%, 20%-40%, and >40%, which are respectively evaluated as appropriate, excessive / insufficient, and severely excessive / severely insufficient.

[0012] Preferably, step (4) includes: (4.1) Establish a prior model based on experimental data: Collect dietary intake, weight, body fat and blood pressure data of the preset sample subjects within a continuous number of days, and establish a correlation model between diet and weight, body fat and blood pressure based on this data to obtain a prior probability model of nutrient intake-health effect; or generate a prior probability model of nutrient intake-health effect through public databases or literature. (4.2) After training the nutrient intake-health effect prior probability model using individual-generated time longitudinal data, a posterior model is generated: For a specific individual, the individual's dietary intake and health indicators are continuously recorded each time, and after accumulating a certain amount of time series data, the nutrient intake-health effect prior probability model is trained using the data of the specific individual; by adjusting the coefficients of the relevant independent variables in the model, a posterior model of nutrient intake-health effect at the individual level is obtained. (4.3) Based on the data continuously generated by the specific individual, dynamically adjust the posterior model of nutrition intake-health effect.

[0013] Preferably, step (5) includes: (5.1) Based on the personalized nutrition intake-health effect post-test model, the health index values ​​of weight, body fat and blood pressure after a period of time are predicted according to the nutrients currently ingested by the individual. The normal range of the Chinese population published by the official is used as the cut point to obtain the corresponding nutrition intake threshold. The nutrition intake includes, but is not limited to, nutrients such as fat and sodium, and can also be specific foods. At this time, the threshold is used as the standard to judge whether the individual's food and nutrient intake is appropriate, and personalized nutrition evaluation is carried out according to the evaluation principles provided in step (3). (5.2) Based on the indicator prediction results obtained in step (5.1), compare them with the current individual's health indicators and provide corresponding personalized health risk warnings; (5.3) Based on individual health management goals, formulate nutritional control goals, including but not limited to a 10% reduction in fat intake and a 10% reduction in sodium intake; based on the individual's dietary intake data recorded continuously over time, generate individual dietary preference characteristics and individual taste preference characteristics, and combine them with the requirements of local customs and dietary habits to generate personalized AI-based next-meal dietary selection guidance.

[0014] Preferably, step (5.3) specifically involves using a particle swarm optimization algorithm to generate suggested dish / food combinations for the next meal: (5.3.1) Initialize a swarm of particles, where the swarm size is m, each particle represents a dish / food combination, and each particle has a random position and velocity: v id ' = w × v id +c1×rand()×(p id -x id ) + c2 × Rand() × (P gd -x id ); x id ' = x id +v id； Where the velocity of particle i is v i = (v i1 , v i2 , …, v id ), d is the dimension; w is the inertia weight, c1 and c2 are acceleration constants, rand() and Rand() are two random values ​​varying in the range [0,1]; p id x represents the optimal position of each particle i in the d-th dimension; id P represents the position of each particle i in the d-th dimension; gd v is the index number of the best position experienced by all particles in the swarm. id ' represents the updated particle velocity, x id ' represents the updated particle position; for each generation, its d-th dimension (1≤d≤D) changes according to the equation above. (5.3.2) Evaluate the fitness of each particle, that is, the difference between the total target nutrients of the current particle (dish / food combination) and the target nutrient requirements. The smaller the difference, the higher the fitness. (5.3.3) For a specific particle, compare its fitness value with the best position it has ever experienced. That is, if the difference between the total target nutrient value of the particle (dish / food combination) and the target nutrient requirement is smaller, the particle is judged to be better and then it is taken as the current best position. (5.3.4) For each particle, compare its fitness value with the best position experienced by all particles in the world. If the fitness value is smaller, it is judged as better, and the index number of the best position in the world is reset for that particle. (5.3.5) Based on the calculation formula provided in step (5.3.1), establish a prior model and a posterior model of nutrient intake-health effect based on the experimental population data, continuously update the velocity and position of particles, and find better dish / food combinations. (5.3.6) Select the dish / food combination corresponding to the best position in the global order as the recommended result. This result should meet the nutritional goals, taste and food preferences, so as to obtain personalized guidance for the next meal selection; if the termination condition is not met, return to step (5.3.1).

[0015] Preferably, step (5.3) further includes correcting the AI ​​results according to the requirements of daily life customs and dietary habits, so that they conform to the empirical results of artificial nutrition guidance, wherein daily life customs and dietary habits include: each meal must have one and only one staple food; each meal must have at least one vegetable other than root vegetables; and each meal must have at least one animal food. Ultimately, we will obtain recommended food combinations for the next meal in a restaurant / canteen, or for meals outside of a restaurant / canteen, that meet nutritional goals, individual dietary preferences, daily life customs, and eating habits. Based on the characteristics of Chinese diet, reminders are generated after each of the three daily meals to encourage individuals to eat fruits and dairy products.

[0016] This invention employs an AI-based personalized nutrition assessment and guidance method based on continuous individual dietary data. Based on continuous individual dietary data and health indicators, and without relying on professional nutritionists or public health physicians, AI automatically provides personalized nutrition services. A prior logistic model is used to generate a nutrient intake-health effect model. Based on subsequently collected individual information, a Bayesian prediction model is used to form a posterior model, ultimately creating an individual model based on dynamic data over time; that is, each individual has a nutrient intake-health effect model. Clustering methods are used to extract individual tastes and food preferences from the individual dietary information dataset. Particle swarm optimization is then used to generate real-time AI-based nutritional assessments, risk warnings, and next meal guidance (i.e., next meal food selection) tailored to individual characteristics and preferences, thereby assisting users in managing their personal nutritional health. Attached Figure Description

[0017] Figure 1 This is a flowchart of the AI-based personalized nutrition assessment and guidance method based on individual continuous dietary data according to the present invention.

[0018] Figure 2 This is a schematic diagram of the menu recommendation workflow based on particle swarm optimization algorithm of the AI-based personalized nutrition evaluation and guidance method for individuals with continuous dietary data according to the present invention. Detailed Implementation

[0019] To more clearly describe the technical content of the present invention, the following description is provided in conjunction with specific embodiments.

[0020] Before describing the embodiments of the present invention in detail, it should be noted that, in the following, the terms “comprising,” “including,” or any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0021] Please see Figure 1 As shown, this AI-based personalized nutrition assessment and guidance method based on individual continuous dietary data includes the following steps: (1) Calculate the individual's single meal food and nutrient intake, including dining in restaurants / canteens or non-restaurants / canteens; (2) Collect individual health index indicators and calculate individual daily energy requirements and individual daily food and nutrient requirements; (3) Based on the Chinese Dietary Guidelines and recommended intake standards of dietary nutrients corresponding to the individual's main characteristics, conduct real-time personalized nutritional evaluation of the food and nutrients consumed in a single meal; (4) Based on the experimental population data and the data information of individual longitudinal continuous records, establish a personalized nutrition intake-health effect model based on the individual, generate the current individual nutrition intake-health effect relationship, and extract individual taste and food preference feature information; (5) Based on the personalized nutrition intake-health effect model, conduct personalized nutrition evaluation and personalized health risk warning; and based on the individual's determined nutrition goals, taste and food preferences, as well as the requirements of local customs and dietary habits, create an AI nutrition guidance algorithm to generate AI guidance by simulating the nutrition guidance of professionals.

[0022] In a preferred embodiment of the present invention, step (1) includes: (1.1) Obtain the food intake of diners in restaurants / canteens: First, collect the cooking recipe information of the restaurant kitchen, including the raw and cooked weights of the main and auxiliary ingredients and seasonings of each dish, and calculate the intake of various foods in a single serving of each dish. Then, record the individual's food order and discard information for each dish during the meal, and combine this with the above recipe information to obtain the individual's food intake for a single meal; (1.2) Obtaining food intake from meals outside the restaurant / canteen: Individuals record the various foods and their weights consumed at each meal; (1.3) Calculate the individual's single-meal energy and nutrient intake as follows: =η1× 2 ... n ； in, This represents the percentage of the nth dish consumed by an individual. This represents the carbohydrate content of a single serving of the nth dish, and so on for other nutrients.

[0023] In a preferred embodiment of the present invention, step (2) includes: (2.1) Collect basic individual information: Collect information on individual age, gender, height, weight, physical activity, and percentage of food intake at each of the three meals; (2.2) Calculate the individual's daily energy requirement as follows to determine the individual's daily food requirement: Energy requirement = Basal energy expenditure × Physical activity level; (2.3) Based on the balanced dietary patterns and food quantities set in the Dietary Guidelines for Chinese Residents for different energy requirement levels, determine the daily recommended intake of various types of food for individuals corresponding to different levels of daily energy requirement, and use it as the standard for judging whether the individual's food intake is appropriate; based on the recommended intake of each nutrient for individuals of different ages, genders, physiological conditions and labor intensity in the Dietary Reference Intakes for Chinese Residents, use it as the standard for judging whether the individual's energy and nutrient intake is appropriate.

[0024] In a preferred embodiment of the present invention, step (3) includes: (3.1) Evaluation of individual food intake per meal: Based on the proportion of the individual’s three meals, the daily recommended food intake in the Chinese Dietary Guidelines is converted into the recommended food intake per meal according to the proportion; and the actual intake of each individual at each meal is compared with the recommended intake. Based on the comparison results, an evaluation level is determined: exceeding / below the recommended intake range of 0%-20%, 20%-40%, and >40%, which are respectively evaluated as appropriate, excessive / insufficient, and severely excessive / severely insufficient. (3.2) Personalized nutrient intake evaluation for each meal: Based on the individual’s three meals ratio, the recommended daily intake of energy and nutrients in the “Chinese Dietary Reference Intakes” is converted into the recommended intake for each meal according to the ratio; and the actual intake of energy and nutrients for each meal is compared with the recommended intake. Based on the comparison results, an evaluation level is determined: exceeding / below the recommended intake range of 0%-20%, 20%-40%, and >40%, which are respectively evaluated as appropriate, excessive / insufficient, and severely excessive / severely insufficient.

[0025] In a preferred embodiment of the present invention, step (4) includes: (4.1) Establish a prior model based on experimental population data: Collect dietary intake, weight, body fat and blood pressure data of the preset sample subjects within a continuous number of days, and establish a correlation model between diet and weight, body fat and blood pressure based on this data to obtain a prior probability model of nutrient intake-health effect; or generate a prior probability model of nutrient intake-health effect through public databases or literature. (4.2) After training the nutrient intake-health effect prior probability model using individual-generated time longitudinal data, a posterior model is generated: For a specific individual, the individual's dietary intake and health indicators are continuously recorded each time, and after accumulating a certain amount of time series data, the nutrient intake-health effect prior probability model is trained using the data of the specific individual; by adjusting the coefficients of the relevant independent variables in the model, a posterior model of nutrient intake-health effect at the individual level is obtained. (4.3) Based on the data continuously generated by the specific individual, dynamically adjust the posterior model of nutrition intake-health effect.

[0026] In a preferred embodiment of the present invention, step (5) includes: (5.1) Based on the personalized nutrition intake-health effect post-test model, the health index values ​​of weight, body fat and blood pressure after a period of time are predicted according to the current individual's nutrient intake. The officially announced normal range of Chinese population is used as the cut-off point, and the corresponding nutrient intake threshold is used. The nutrient intake includes, but is not limited to, nutrients such as fat and sodium, and can also be specific foods. At this time, the threshold is used as the standard to judge whether the individual's food and nutrient intake is appropriate, and personalized nutrition evaluation is carried out according to the evaluation principles provided in step (3). (5.2) Based on the indicator prediction results obtained in step (5.1), compare them with the current individual's health indicators and provide corresponding personalized health risk warnings; (5.3) Based on individual health management goals, formulate nutritional control goals, including but not limited to a 10% reduction in fat intake and a 10% reduction in sodium intake; based on the individual's dietary intake data recorded continuously over time, generate individual dietary preference characteristics and individual taste preference characteristics, and combine them with the requirements of local customs and dietary habits to generate personalized AI-based next-meal dietary selection guidance.

[0027] Please see Figure 2 As shown, in a preferred embodiment of the present invention, step (5.3) specifically involves using a particle swarm optimization algorithm to generate a suggested combination of dishes / foods for the next meal: (5.3.1) Initialize a swarm of particles, where the swarm size is m, each particle represents a dish / food combination, and each particle has a random position and velocity: v id ' = w × v id +c1×rand()×(p id -x id ) + c2 × Rand() × (P gd -x id ); xid ' = x id +v id； Where the velocity of particle i is v i = (v i1 , v i2 , …, v id ), d is the dimension; w is the inertia weight, c1 and c2 are acceleration constants, rand() and Rand() are two random values ​​varying in the range [0,1]; p id x represents the optimal position of each particle i in the d-th dimension; id P represents the position of each particle i in the d-th dimension; gd v is the index number of the best position experienced by all particles in the swarm. id ' represents the updated particle velocity, x id ' represents the updated particle position; for each generation, its d-th dimension (1≤d≤D) changes according to the equation above; (5.3.2) Evaluate the fitness of each particle, that is, the difference between the total target nutrients of the current particle (dish / food combination) and the target nutrient requirements. The smaller the difference, the higher the fitness. (5.3.3) For a specific particle, compare its fitness value with the best position it has ever experienced. That is, if the difference between the total target nutrient value of the particle (dish / food combination) and the target nutrient requirement is smaller, the particle is judged to be better and then it is taken as the current best position. (5.3.4) For each particle, compare its fitness value with the best position experienced by all particles in the world. If the fitness value is smaller, it is judged as better, and the index number of the best position in the world is reset for that particle. (5.3.5) Based on the calculation formula provided in step (5.3.1), establish a prior model and a posterior model of nutrient intake-health effect based on the experimental population data, continuously update the velocity and position of particles, and find better dish / food combinations. (5.3.6) Select the dish / food combination corresponding to the best position in the global order as the recommended result. This result should meet the nutritional goals, taste and food preferences, so as to obtain personalized guidance for the next meal selection; if the termination condition is not met, return to step (5.3.1).

[0028] In a preferred embodiment of the present invention, step (5.3) further includes correcting the AI ​​results according to the requirements of daily life customs and dietary habits, so that they conform to the empirical results of artificial nutrition guidance. These daily life customs and dietary habits include: each meal must have one and only one staple food (rice, noodles, or root vegetables whose weight exceeds 50% of the total weight of the dish, which can be considered a staple food); each meal must have at least one vegetable other than root vegetables whose weight exceeds 50% of the total weight of the dish; and each meal must have at least one animal-based food whose weight exceeds 50% of the total weight of the dish.

[0029] Ultimately, we will obtain recommended food combinations for the next meal in a restaurant / canteen, or for meals outside of a restaurant / canteen, that meet nutritional goals, individual dietary preferences, daily life customs, and eating habits. Based on the characteristics of Chinese diet, reminders are generated after each of the three daily meals to encourage individuals to eat fruits and dairy products.

[0030] In practical applications, the specific implementation process of this solution is as follows: 1. Meal records and immediate nutritional assessment 1.1 Calculation method for individual single meal food and nutrient intake 1.1.1 Method for calculating food intake while dining in restaurants / canteens

[0031] First, the restaurant's kitchen cooking recipe information was collected, including the raw and cooked weights of the main and auxiliary ingredients and seasonings for each dish, to calculate the intake of various foods in a single serving of each dish. Individual order information and discarded food items were recorded, and combined with the above recipe information, the individual's food intake (including seasonings) for a single meal was obtained.

[0032] 1.1.2 Calculation method for food intake when dining outside restaurants / canteens Individuals should record the various foods and their weights consumed in a single meal (including main ingredients, side dishes, and seasonings; if the seasonings are unknown, they should not be recorded).

[0033] This method makes it impossible to estimate the amount of seasonings ingested.

[0034] 1.1.3 Method for calculating individual nutrient intake per meal Based on the individual's recorded food intake for each meal, and combined with the energy and various nutrients of the foods listed in the "Chinese Food Composition Table", the individual's energy and nutrient intake for that meal can be obtained.

[0035] 1.1.4 Calculation of individual nutrient intake per meal Based on an individual's recorded food intake at a single meal, and combined with the energy and various nutrients listed in the "Chinese Food Composition Table," the individual's energy and nutrient intake at that meal can be obtained: =η1× 2 ... n ； in, This represents the percentage of the nth dish consumed by an individual. This represents the carbohydrate content of a single serving of the nth dish, and so on for other nutrients.

[0036] By analogy, the intake of various nutrients for an individual meal can be obtained.

[0037] 1.2 Determining Recommended Food Intakes for Different Energy Requirements Based on the Chinese Dietary Guidelines 1.2.1 Collecting basic individual information Collect basic information such as individual age, gender, height, weight, physical activity level, and percentage of food intake at each meal.

[0038] 1.2.2 Calculate an individual's daily energy requirement Based on individual basic information, calculate the individual's daily energy requirement to determine the individual's daily food needs.

[0039] Energy requirements ( = Basic energy expenditure ( ) × physical activity level ( ).

[0040] Basal energy expenditure ( Calculation: Calculated according to the Schofield formula (As shown in Table 1 below). Subtract 10% from the calculated result for the 18-59 age group to obtain the BEE reference value for that group. For example, for a 40-year-old male weighing 80 kg, his daily... .

[0041] ; Physical Activity Level (PAL) Calculation: According to the 2013 edition of the Dietary Reference Intakes (DRIs) formulated by the Chinese Nutrition Society, the daily physical activity intensity of Chinese adults is divided into three levels: light physical activity level (PAL 1.5), moderate physical activity level (PAL 1.75), and heavy physical activity level (PAL 2.0). If the exercise is for fitness purposes, referring to planned, organized, and repetitive physical activities aimed at maintaining or improving health and enhancing physical fitness, such as moderate-intensity exercise 4-5 times a week for 30-60 minutes each time, then PAL increases by 0.3.

[0042] For example, a 40-year-old male weighing 80 kg, engaging in moderate physical activity and the above-mentioned fitness activities, has an energy requirement (EER) of = basal energy expenditure (BEE) × physical activity level (PAL) = 1790.86 × (1.75 + 0.3) = 3671.263 (kcal).

[0043] 1.2.3 Determine the individual's daily food requirements Based on the calculated daily energy requirement of the individual, and in accordance with the "Dietary Guidelines for Chinese Residents", the recommended daily intake of various types of food (oil, salt, sugar, grains, vegetables and fruits, meat and poultry, aquatic products, etc.) corresponding to different levels of daily energy requirement (such as 1800kcal, 2000kcal, 2200kcal, etc.) is determined, and the food intake is used as the standard for judging whether the individual's dietary intake is appropriate.

[0044] 1.3 Personalized Nutritional Assessment Based on Publicly Available Domestic Recommendations 1.3.1 Personalized food intake assessment for single meals The recommended daily food intake is converted into the recommended intake for each meal based on the proportions of the three meals (the default overall intake ratio for breakfast, lunch, and dinner is 20%:40%:40%, or it can be based on the percentage of food consumed by the individual).

[0045] The evaluation was conducted by comparing the intake with the recommended intake in the "Dietary Guidelines for Chinese Residents". Evaluation levels were established: exceeding / below the recommended intake range by 0%-20%, 20%-40%, and >40%, respectively rated as adequate, excessive / insufficient, and severely excessive / severely insufficient.

[0046] 1.3.2 Personalized food intake assessment for single meals The recommended daily intake of energy and nutrients is converted into the recommended intake of each meal based on the proportions of the three meals (the default overall intake ratio for breakfast, lunch and dinner is 20%:40%:40%, or it can be based on the actual percentage of food consumed by the individual).

[0047] Based on the individual's single-meal energy and nutrient intake obtained above, and combined with personal health information and the recommended intakes of each nutrient for different ages, genders, physiological conditions, and labor intensity in the "Chinese Dietary Reference Intakes," an evaluation was conducted. Evaluation levels were established: exceeding / below the recommended intake range by 0%-20%, 20%-40%, and >40%, respectively evaluated as adequate, excessive / insufficient, and severely excessive / severely insufficient.

[0048] 2. Establish a personalized nutrition intake-health model 2.1 Establishing a priori model based on experimental data To prepare a population trial, collect continuous data on dietary intake, weight, body fat, and blood pressure from 20 subjects over 10 consecutive days. Based on this experimental database, establish a correlation model between diet and weight, body fat, and blood pressure, i.e., a nutrient intake-health effect prior probability model. This involves inputting all sample data into a fixed-coefficient model based on established nutritional science principles regarding the relationship between nutrients and health effects. Specifically, health effects (weight, body fat, blood pressure, etc.) are used as dependent variables, dietary components (salt, sodium, fat, etc.) are used as primary independent variables, and basic personal information (age, gender, exercise intensity, physiological condition, etc.) are used as independent variables. This results in models such as salt / sodium-hypertension and fat-weight / body fat relationships. This model is not personalized; it represents the basic characteristics of the population to which the sample subjects belong. A Bayesian model can be used to generate prior probabilities.

[0049] 2.2 Training the posterior model using individual-generated longitudinal time data A specific individual continuously records their food (as mentioned above, recorded using food coding), energy, nutrient intake information, as well as weight, body fat percentage, blood pressure, etc., accumulating a certain amount of time-series data (data from at least 5 time points is recommended). This individual's data is then used to train the aforementioned prior probability model. Specifically, the individual's longitudinal time data is sequentially fed into the model, dynamically adjusting the coefficients of independent variables (such as salt / sodium and fat) in models related to the relationship between salt / sodium and hypertension, and fat and weight / body fat. This yields a posterior model of nutrient intake-health effects. Furthermore, this model is dynamically adjusted based on continuously provided data from the specific individual. In other words, this model is a prior probability model generated from the basic population characteristics in step 2.1 above, corrected based on the individual's longitudinal time data. At this point, the model is a personalized nutrient intake-health effects model generated based on the specific individual's own longitudinal time data, possessing personalized characteristics.

[0050] 3 Personalized AI guidance algorithms 3.1 Personalized Nutritional Assessment Once the aforementioned personalized nutrition intake-health effect model is formed, the recommended amounts from "personal immediate nutrition assessment based on the Chinese Dietary Guidelines" in step 1.3 can no longer be used to evaluate individuals. Instead, based on the posterior model, the predicted weight, body fat, blood pressure, etc., based on fat and sodium intake are compared with the normal values ​​published by the Chinese government to obtain the threshold values ​​for fat and sodium intake (which may differ from the recommended values ​​in the Chinese Dietary Guidelines listed in step 1.3 above), and the evaluation is carried out according to the evaluation principles in step 1.3.

[0051] This can be understood as a further optimized evaluation scheme for personalized nutrition evaluation. In practical applications, if users are unable to generate personalized nutrition evaluation results (e.g., due to insufficient sampling information leading to incomplete data), they can skip this step and directly use the evaluation principles in step 1.3 for evaluation.

[0052] 3.2 Personalized Health Risk Warnings Based on the personalized nutrition intake-health effect model provided in step 2.2 above, taking the sodium-hypertension model as an example, the specific value of sodium intake in an individual's current diet is used as the independent variable and substituted into the model to predict blood pressure and other values ​​after one month of continuous intake of the same amount. This predicted blood pressure value is compared with the actual measured blood pressure value uploaded by the individual, and the difference between the increase or decrease is provided as a blood pressure risk warning to the individual. Specifically, the information provided could be: "Based on the current diet, blood pressure may increase / decrease by XX mmHg after one month." Fat-weight / body fat, etc., can also be used to generate health risk warning information based on the personalized nutrition intake-health effect model in step 2.2.

[0053] 3.3 Generate AI-powered guidance for the next meal 3.3.1 Determine nutritional goals Using factors such as fat and sodium as individual dietary control targets, the AI ​​algorithm incorporates these targets as one of the criteria for generating meal suggestions for the next meal. For example, if an individual aims to reduce their dietary fat intake by 15%, the AI ​​algorithm will generate a meal combination for the next meal with a fat content 15% lower than the previously recorded dietary fat content (i.e., the baseline diet). Dietary control targets can be set by the individual.

[0054] In the AI ​​algorithm, dietary control goals can be set and achieved gradually over three months. For example, if the dietary control goal is to reduce fat intake by 15%, then the goal can be set to reduce it by 5% from the baseline in the first month, another 5% from the baseline in the second month, and another 5% from the baseline in the third month.

[0055] 3.3.2 Determining Individual Dietary Preference Characteristics AI algorithms can generate individual taste and food preference characteristics based on the types and amounts of food consumed and nutrients through clustering. When generating meal suggestions for the next meal, they prioritize recommending dishes that match the individual's taste and food preference characteristics. For example, if vinegar-based condiments (rice vinegar, white vinegar, pickled green beans, etc.) are frequently found in the diet, indicating a preference for sour flavors, dishes seasoned with vinegar will be prioritized. The extraction of taste and food preference information can be based on the personal dietary information recorded in step 1.1, combined with the coding rules for food and condiments in the "Chinese Food Composition Table," and clustered according to major food categories. The continuously recorded dietary information can be used to train the individual's posterior model and update the individual's taste and food preference characteristics.

[0056] 3.3.2.1 Defining Individual Taste Preference Characteristics Based on the most frequently occurring condiment categories in the individual's food intake information from step 1.1, individual taste preferences are defined. Frequent appearance of soy sauce in the dietary information indicates a preference for braised dishes; frequent appearance of vinegar indicates a preference for sour flavors; and frequent appearance of sugar indicates a preference for sweet flavors. Commercially available compound seasonings (those not containing information on traditional seasonings such as oil, salt, soy sauce, and vinegar in the recipe information from step 1.1.1) containing terms like "braised" or "sweet and sour" are categorized into the corresponding taste preferences.

[0057] The specific implementation method is as follows: according to the "Chinese Food Composition Table", each food is coded with a 6-digit food code. The first 2 digits represent the major category of the food, and the third digit represents the subcategory. The food is classified into the corresponding condiment category according to the major category and the subcategory. For example, the first 3 digits of the food code 201 represent soy sauce, the first 3 digits of the food code 202 represent vinegar, and the first 3 digits of the food code 181 represent sugar.

[0058] 3.3.2.2 Defining Individual Food Preference Characteristics The food categories that appear frequently in the individual's food intake information from step 1.1 are defined as the individual's food preference characteristics. For example, if green leafy vegetables, cabbage, and lettuce frequently appear in the individual's food intake information, they are classified as having a preference for leafy vegetables.

[0059] The specific implementation method is as follows: according to the "Chinese Food Composition Table", each food is coded with a 6-digit food code. The first 2 digits represent the major category of the food and the third digit represents the subcategory. The food is classified into the corresponding food according to the food code. For example, if the first 2 digits of the food code are 04 and the third digit is 7, it represents leafy vegetables.

[0060] 3.3.3 Establishing Principles for AI-Driven Next Meal Guidance Algorithms The particle swarm optimization algorithm (an optimization technique based on swarm intelligence to find the best combination of dishes that meets specific requirements) from computer science can be used as the basic algorithm to generate suggested combinations of dishes / foods for the next meal.

[0061] Based on the nutritional goals described in step 3.3.1 and the individual dietary preferences described in step 3.3.2, and a pre-set database containing information on various dishes / foods and their nutrients, the most suitable dish / food combination is selected and recommended to the individual, specifically as follows: Each particle represents a dish / food combination, and its position is binary, indicating whether a dish is selected (1 for selection, 0 for non-selection). A fitness function calculates the difference between the total nutrients of the dish / food combination represented by the particle and the target nutritional requirement. The smaller the difference, the higher the fitness. Particles update their velocity and position based on an individual nutrition-health effect model, searching for better dish / food combinations. Finally, the dish / food combination corresponding to the globally optimal position is selected as the recommendation result to meet nutritional goals and dining preferences, providing personalized guidance for the next meal.

[0062] The calculation formula is as follows: v id ' = w × v id +c1×rand()×(p id -x id ) + c2 × Rand() × (P gd -x id ); x id ' = x id +v id； Where the velocity of particle i is v i = (v i1 , v i2 , …, v id ), d is the dimension; w is the inertia weight, c1 and c2 are acceleration constants, rand() and Rand() are two random values ​​varying in the range [0,1]; p id x represents the optimal position of each particle i in the d-th dimension; id P represents the position of each particle i in the d-th dimension; gd v is the index number of the best position experienced by all particles in the swarm. id ' represents the updated particle velocity, x id ' represents the updated particle position; for each generation, its d-th dimension (1≤d≤D) changes according to the equation above; The particle swarm optimization algorithm is as follows: 1. Initialize a group of particles (group size m, each particle represents a dish / food combination), including random positions and velocities; 2. Evaluate the fitness of each particle (i.e., the difference between the total nutrients of the dish / food combination and the target nutritional requirements). 3. For each particle, compare its fitness value with the best position it has ever experienced (i.e., when the total nutrients of the dish / food combination are less than the target nutritional requirement). If it is better, then take it as the current best position. 4. For each particle, compare its fitness value with the best position it has experienced globally. If it is better, reset the index number of the best position globally. 5. Calculate the velocity and position of the changing particle using the above formula; 6. If the termination condition is not met (usually a sufficiently good fitness value or reaching a preset maximum number of generations Gmax), return to the beginning and recalculate.

[0063] 3.3.4 Correcting AI results based on lifestyle and dietary habits Using all dishes on the restaurant / cafe's next meal menu as a selection pool, and in addition to the results generated by the above AI algorithm principles, further conditions related to daily life customs and eating habits are added: 1. Each meal must include, and only includes, one staple food. This staple food includes rice, noodles, and dishes primarily composed of root vegetables (root vegetables accounting for more than 50% of the total weight). The reason is that root vegetables are high in starch, or, as a type of whole grain, provide carbohydrates equivalent to staple foods. 2. Each meal must include at least one vegetable other than root vegetables; 3. Each meal must include at least one animal-based food, such as fish or poultry.

[0064] Ultimately, this results in a recommended combination of dishes for the next meal in a restaurant / canteen, or a single food combination (excluding seasonings), that aligns with nutritional goals, individual dietary preferences, lifestyle habits, and eating habits.

[0065] 3.3.5 AI Algorithm for Extra Feeding Based on the characteristics of Chinese diet, fruits and dairy products are not included in the three daily meals. If the individual's recorded information on fruits or dairy products is not obtained before 7 pm, a reminder can be generated to remind the individual to eat fruits and dairy products.

[0066] It should be understood that various parts of the present invention can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution device.

[0067] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0068] The storage media mentioned above can be read-only memory, disk, or optical disk, etc.

[0069] In the description of this specification, references to terms such as "an embodiment," "some embodiments," "example," "specific example," or "embodiment," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0070] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

[0071] This invention employs an AI-based personalized nutrition assessment and guidance method based on continuous individual dietary data. Based on continuous individual dietary data and health indicators, and without relying on professional nutritionists or public health physicians, AI automatically provides personalized nutrition services. A prior logistic model is used to generate a nutrient intake-health effect model. Based on subsequently collected individual information, a Bayesian prediction model is used to form a posterior model, ultimately creating an individual model based on dynamic data over time; that is, each individual has a nutrient intake-health effect model. Clustering methods are used to extract individual tastes and food preferences from the individual dietary information dataset. Particle swarm optimization is then used to generate real-time AI-based nutritional assessments, risk warnings, and next meal guidance (i.e., next meal food selection) tailored to individual characteristics and preferences, thereby assisting users in managing their personal nutritional health.

[0072] In this specification, the invention has been described with reference to specific embodiments thereof. However, it will be apparent that various modifications and variations can be made without departing from the spirit and scope of the invention. Therefore, the specification and drawings should be considered illustrative rather than restrictive.

Claims

1. A method for AI-based personalized nutrition assessment and guidance based on individual continuous dietary data, characterized in that, The method includes the following steps: (1) Calculate the individual's single meal food and nutrient intake, including dining in restaurants / canteens or non-restaurants / canteens; (2) Collect individual health index indicators and calculate individual daily energy requirements and individual daily food and nutrient requirements; (3) Based on the Chinese Dietary Guidelines and recommended intake standards of dietary nutrients corresponding to the individual's main characteristics, conduct real-time personalized nutritional evaluation of the food and nutrients consumed in a single meal; (4) Based on the experimental population data and the data information of individual longitudinal continuous records, establish a personalized nutrition intake-health effect model based on the individual, generate the current individual nutrition intake-health effect relationship, and extract individual taste and food preference feature information; (5) Based on the personalized nutrition intake-health effect model, conduct personalized nutrition evaluation and personalized health risk warning; and based on the individual's determined nutrition goals, taste and food preferences, as well as the requirements of local customs and dietary habits, create an AI nutrition guidance algorithm to generate AI guidance by simulating the nutrition guidance of professionals.

2. The AI-based personalized nutrition assessment and guidance method based on individual continuous dietary data according to claim 1, characterized in that, Step (1) includes: (1.1) Obtain the food intake of diners in restaurants / canteens: First, collect the cooking recipe information of the restaurant kitchen, including the raw and cooked weights of the main and auxiliary ingredients and seasonings of each dish, and calculate the intake of various foods in a single serving of each dish. Then, record the individual's food order and discard information for each dish during the meal, and combine this with the above recipe information to obtain the individual's food intake for a single meal; (1.2) Obtaining food intake from meals outside the restaurant / canteen: Individuals record the various foods and their weights consumed at each meal; (1.3) Calculate the individual's single-meal energy and nutrient intake as follows: =η1× 2 …… n ； in, This represents the percentage of the nth dish consumed by an individual. This represents the carbohydrate content of a single serving of the nth dish, and so on for other nutrients.

3. The AI-based personalized nutrition assessment and guidance method based on individual continuous dietary data according to claim 1, characterized in that, Step (2) includes: (2.1) Collect basic individual information: including individual age, gender, height, weight, physical activity, and percentage of food intake at each of the three meals; (2.2) Calculate the individual's daily energy requirement as follows to determine the individual's daily food requirement: Energy requirement = Basal energy expenditure × Physical activity level; (2.3) Based on the balanced dietary patterns and food quantities set in the Dietary Guidelines for Chinese Residents for different energy requirement levels, determine the daily recommended intake of various types of food for individuals corresponding to different levels of daily energy requirement, and use it as the standard for judging whether the individual's food intake is appropriate; based on the recommended intake of each nutrient for individuals of different ages, genders, physiological conditions and labor intensity in the Dietary Reference Intakes for Chinese Residents, use it as the standard for judging whether the individual's energy and nutrient intake is appropriate.

4. The AI-based personalized nutrition assessment and guidance method based on individual continuous dietary data according to claim 3, characterized in that, Step (3) includes: (3.1) Evaluation of individual food intake per meal: Based on the proportion of the individual’s three meals, the daily recommended food intake in the Chinese Dietary Guidelines is converted into the recommended food intake per meal according to the proportion; and the actual intake of each individual at each meal is compared with the recommended intake. Based on the comparison results, an evaluation level is determined: exceeding / below the recommended intake range of 0%-20%, 20%-40%, and >40%, which are respectively evaluated as appropriate, excessive / insufficient, and severely excessive / severely insufficient. (3.2) Personalized nutrient intake evaluation for each meal: Based on the individual’s three meals ratio, the recommended daily intake of energy and nutrients in the “Chinese Dietary Reference Intakes” is converted into the recommended intake for each meal according to the ratio; and the actual intake of energy and nutrients for each meal is compared with the recommended intake. Based on the comparison results, an evaluation level is determined: exceeding / below the recommended intake range of 0%-20%, 20%-40%, and >40%, which are respectively evaluated as appropriate, excessive / insufficient, and severely excessive / severely insufficient.

5. The AI-based personalized nutrition assessment and guidance method based on individual continuous dietary data according to claim 4, characterized in that, Step (4) includes: (4.1) Establish a prior model based on experimental population data: Collect dietary intake, weight, body fat and blood pressure data of the preset sample subjects within a continuous number of days, and establish a correlation model between diet and weight, body fat and blood pressure based on this data to obtain a prior probability model of nutrient intake-health effect; or generate a prior probability model of nutrient intake-health effect through public databases or literature. (4.2) After training the nutrient intake-health effect prior probability model using individual-generated time longitudinal data, a posterior model is generated: For a specific individual, the individual's dietary intake and health indicators are continuously recorded each time, and after accumulating a certain amount of time series data, the nutrient intake-health effect prior probability model is trained using the data of the specific individual; by adjusting the coefficients of the relevant independent variables in the model, a posterior model of nutrient intake-health effect at the individual level is obtained. (4.3) Based on the data continuously generated by the specific individual, dynamically adjust the posterior model of nutrition intake-health effect.

6. The AI-based personalized nutrition assessment and guidance method based on individual continuous dietary data according to claim 5, characterized in that, Step (5) includes: (5.1) Based on the personalized nutrition intake-health effect post-test model, the health index values ​​of weight, body fat and blood pressure after a period of time are predicted according to the current individual's nutrient intake. The official normal range of the Chinese population is used as the cut point to obtain the corresponding nutrition intake threshold. Nutrition intake includes, but is not limited to, fat and sodium. At this time, the threshold is used as the standard to judge whether the individual's food and nutrient intake is appropriate, and personalized nutrition evaluation is carried out according to the evaluation principles provided in step (3). (5.2) Based on the health indicator prediction results obtained in step (5.1), compare them with the current individual's health indicators and provide corresponding personalized health risk warnings; (5.3) Based on individual health management goals, formulate nutritional control goals, generate individual dietary preference characteristics and individual taste preference characteristics based on individual dietary intake data recorded continuously over time, and generate personalized AI-based next-meal dietary selection guidance in combination with local customs and dietary habits.

7. The AI-based personalized nutrition assessment and guidance method based on individual continuous dietary data according to claim 6, characterized in that, Step (5.3) involves creating an AI nutrition guidance algorithm that simulates the approach of professional nutrition guidance. Specifically, it uses a particle swarm optimization algorithm to generate suggested dish / food combinations for the next meal. (5.3.1) Initialize a swarm of particles, where the swarm size is m, each particle represents a dish / food combination, and each particle has a random position and velocity: v id '=w×v id +c1×rand()×(p id -x id )+c2×Rand()×(P gd -x id ); x id '=x id +v id； Where the velocity of particle i is v i = (v i1 , v i2 , …, v id ), d is the dimension; w is the inertia weight, c1 and c2 are acceleration constants, rand() and Rand() are two random values ​​varying in the range [0,1]; p id x represents the optimal position of each particle i in the d-th dimension; id P represents the position of each particle i in the d-th dimension; gd v is the index number of the best position experienced by all particles in the swarm. id ' represents the updated particle velocity, x id ' represents the updated particle position; for each generation, its d-th dimension (1≤d≤D) changes according to the equation above. (5.3.2) Evaluate the fitness of each particle, that is, the difference between the total target nutrients of the current particle and the target nutrient requirements. The smaller the difference, the higher the fitness. (5.3.3) For a specific particle, compare its fitness value with the best position it has ever experienced. That is, if the difference between the total target nutrient value of the particle and the target nutrient requirement is smaller, the particle is judged to be better and then it is taken as the current best position. (5.3.4) For each particle, compare its fitness value with the best position experienced by all particles in the world. If the fitness value is smaller, it is judged as better, and the index number of the best position in the world is reset for that particle. (5.3.5) Based on the calculation formula provided in step (5.3.1), establish a prior model and a posterior model of nutrient intake-health effect based on the experimental population data, continuously update the velocity and position of particles, and find better dish / food combinations. (5.3.6) Select the dish / food combination corresponding to the best position in the global order as the recommended result. This result should meet the nutritional goals, taste and food preferences, so as to obtain personalized guidance for the next meal selection; if the termination condition is not met, return to step (5.3.1).

8. The AI-based personalized nutrition assessment and guidance method based on individual continuous dietary data according to claim 7, characterized in that, The step (5.3) further includes correcting the AI ​​results based on daily life customs and dietary habits to make them consistent with the experience results of professional nutrition guidance. Daily life customs and dietary habits include: each meal must have one and only one staple food; each meal must have at least one vegetable other than root vegetables; and each meal must have at least one animal food. Ultimately, we will obtain recommended food combinations for the next meal in a restaurant / canteen, or for meals outside of a restaurant / canteen, that meet nutritional goals, individual dietary preferences, daily life customs, and eating habits. Based on the characteristics of Chinese diet, reminders are generated after each of the three daily meals to encourage individuals to eat fruits and dairy products.

Images