AI-based method, device and equipment for assisting diet conditioning management and control of chronic diseases of old people and medium

Abstract

The invention relates to an AI-based method, device and equipment for assisting diet conditioning management and control of chronic diseases of old people and a medium. The method comprises the following steps: collecting multi-dimensional user data, and preprocessing the multi-dimensional user data to obtain a user health file; based on the user health archive, calculating a nutrition health demand of a corresponding user, and obtaining a personalized nutrition target list; calculating a suitability score corresponding to the user for each food material based on the personalized nutrition target list; based on the suitability score, screening the food materials to obtain a safe food material library; based on the safe food material library, performing preliminary screening on a recipe database to obtain a candidate recipe set; constructing an optimization problem based on the personalized nutrition target list and the candidate recipe set; and solving the optimization problem to obtain a recipe scheme. By adopting the method, the complex relationship between the disease and the nutrition constraint can be accurately mined and dynamically matched, and the personalized health requirement of the patient is met.

Description

Technical Field

[0001] This invention belongs to the field of intelligent optimization decision-making, and in particular relates to a method, device, equipment and medium for the management and control of dietary conditions for elderly people with chronic diseases based on AI assistance. Background Technology

[0002] With the rapid development of artificial intelligence and digital health technologies, intelligent dietary management is gradually becoming an important direction for chronic disease prevention and control. This technology can provide patients with personalized dietary advice and achieve precise control of nutrient intake through data-driven methods. In the traditional process of dietary management for elderly patients with chronic diseases, clinical nutritionists usually rely on general guidelines and limited consultation information to develop static dietary plans for patients. While this approach has a certain degree of professionalism, its core logic is based on standardized rules and human experience. Under the traditional technical framework, nutritionists need to first assess the patient's disease type and physical indicators, then calculate the daily energy and nutrient requirements according to established nutritional rules, and finally match suitable dishes from a predefined recipe library. The entire process is highly dependent on the expert's personal experience, and once the plan is formed, it is usually kept fixed for a long time, only adjusted when the patient's indicators fluctuate significantly or during a follow-up visit. However, this traditional management method, which is based on static rules and periodic human intervention, is gradually revealing its limitations when facing the complex needs of elderly chronic disease management. The rule base has limited coverage and is difficult to effectively address nutritional goal conflicts arising from the coexistence of multiple diseases. Static rules often lack flexible solutions. The plans cannot be dynamically adapted to the user's physical condition. When the user's physiological indicators, medication, or lifestyle changes, the existing dietary plan may quickly become ineffective or even have negative effects. Traditional methods are insufficient in interpretability, making it difficult to clearly explain to users why this food is recommended instead of another, which makes it difficult to improve user trust and compliance. These factors together restrict the further development of traditional dietary management methods in terms of precision and personalization. Summary of the Invention

[0003] Therefore, it is necessary to provide an AI-assisted method, device, equipment, and medium for the dietary management and control of chronic diseases in the elderly, which can accurately discover and dynamically match the complex relationship between diseases and nutritional constraints to meet the personalized health needs of elderly patients.

[0004] Firstly, this application provides an AI-assisted dietary management method for chronic diseases in the elderly, including:

[0005] Collect multidimensional user data and preprocess it to obtain user health records; multidimensional user data includes static data, disease data, physical examination data and preference data;

[0006] Based on the user's health record, calculate the corresponding user's nutritional health needs and obtain a personalized list of nutritional goals.

[0007] Based on a personalized nutritional goal list, a suitability score is calculated for each food item for the corresponding user; and based on the suitability score, food items are selected to obtain a safe food item database.

[0008] Based on the safe food ingredient database, a preliminary screening of the recipe database was conducted to obtain a candidate recipe set;

[0009] Based on a personalized list of nutritional goals and a set of candidate recipes, an optimization problem is constructed; and the optimization problem is solved to obtain a recipe plan; the recipe plan is used to assist in dietary regulation.

[0010] Furthermore, based on a personalized list of nutritional goals and a candidate recipe set, optimization problems are constructed, including:

[0011] Based on the candidate recipe set, variables corresponding to the optimization problem are defined to obtain the decision variables;

[0012] Based on decision variables, a personalized nutrition target list, and user health records, a multi-objective function is constructed using the following formula:

[0013]

[0014] in, Multi-objective function As decision variables, For nutrition-related parameters, For disease-related parameters, For the target parameter of preference, For cost parameters, , , , These are the corresponding weighting coefficients;

[0015] Based on the personalized nutrition target list and the user's health record, constraints are generated; the constraints include at least one of the following: hard nutrition constraints, food safety constraints, diversity constraints, and rationality constraints.

[0016] By integrating decision variables, multi-objective functions, and constraints, the optimization problem is obtained.

[0017] Furthermore, the parameters of each term in the multi-objective function are obtained through the following method:

[0018] Based on a personalized list of nutrition goals, nutrition-related parameters are calculated using the following formula:

[0019]

[0020] in, For nutrition-related parameters, Let K be the importance weight of the k-th nutrient, K be the total number of nutrient types, k be the nutrient type index, i be the recipe index, N be the total number of recipes, j be the meal index, and M be the total number of meals. It is a binary variable. Let i be the content of the k-th nutrient provided by the i-th recipe at the j-th meal. This represents the user's target daily intake of the kth nutrient.

[0021] Based on the user's health record, disease-related parameters are calculated using the following formula:

[0022]

[0023] in, For disease-related parameters, Let be the severity weight of the d-th disease, d be the disease index, D be the total number of disease types, i be the recipe index, N be the total number of recipes, j be the meal index, and M be the total number of meals. It is a binary variable. For the i-th recipe, score the risk of the d-th disease at the j-th meal;

[0024] Based on the user's health profile, the preference target parameter is calculated using the following formula:

[0025]

[0026] in, Here, i represents the recipe index, N represents the total number of recipes, j represents the meal index, and M represents the total number of meals. It is a binary variable. The score is given to the degree of match between the i-th recipe and the user's taste preferences at the j-th meal.

[0027] Based on the candidate recipe set, the cost parameters are calculated using the following formula:

[0028]

[0029] in, Here, i is the recipe index, N is the total number of recipes, j is the meal index, and M is the total number of meals. It is a binary variable. The estimated cost of the i-th recipe in the j-th meal.

[0030] Furthermore, based on a personalized nutritional goal list, a suitability score is calculated for each food item in the nutritional knowledge graph for the corresponding user, including:

[0031] Based on the user's health record, foods that are clearly allergenic or unsuitable for consumption are removed from the preset full food database to obtain a preliminary set of candidate foods;

[0032] Based on a nutritional knowledge graph, we traverse each ingredient in the initial candidate ingredient set, and for each disease of the user, we query the relationship between ingredients and diseases to obtain ingredient-disease pairs.

[0033] For each pair of food-disease pairs, calculate the disease risk-reward score to obtain the food-disease risk-reward matrix;

[0034] Based on the food disease risk-benefit matrix, a suitability score is obtained between the user and the food.

[0035] Furthermore, based on the food-related disease risk-benefit matrix, a suitability score between the user and the food is obtained, including:

[0036] Based on users' health records, the following formula is used to assign importance weights to diseases:

[0037]

[0038] in, Let the importance weight of disease d be... The baseline severity of disease type d is given by [reference to a specific disease type], where D represents the total number of disease types. denoted as the normalized deviation of the key indicator for the d-th disease, where d is the disease index;

[0039] Iterate through each nutrient in the personalized nutrition goal list, calculate the contribution of food to the nutrient, and obtain the nutrition fit score.

[0040] The suitability score is obtained by weighting and summing the importance weights, the food disease risk-benefit matrix, and the nutritional compatibility score.

[0041] Furthermore, after solving the optimization problem and obtaining the recipe scheme, the process also includes:

[0042] By comparing the recipe plan with the user's health record, the key points of the recipe plan are identified, and a list of explanatory points is obtained; the key points include highlighting the advantages, explaining the trade-offs, and stating the contraindications;

[0043] For each key point in the explanation list, the nutrition knowledge graph is consulted to obtain the decision logic chain;

[0044] Based on natural language generation, the decision logic chain is transformed into natural language to obtain the corresponding recipe solution's explanation statement;

[0045] Based on the explanation statements and recipe scheme, a recipe interpretability report is generated; the recipe interpretability report is used to help understand the recipe scheme.

[0046] Furthermore, after generating a recipe interpretability report based on the explanatory statements and recipe scheme, it also includes:

[0047] Receive dietary records and subjective feelings from users, and simultaneously acquire users' health monitoring data to obtain a behavioral time series dataset;

[0048] Input the behavioral time series dataset into a pre-trained time series analysis model to obtain an abnormal pattern report;

[0049] Based on the abnormal pattern report, the nutritional knowledge graph was queried to obtain the root cause inference conclusion;

[0050] Based on the root cause inference conclusions, the diet plan is adjusted according to the predefined adjustment rule base and clinical nutrition principles to obtain a personalized adjustment strategy set; the personalized adjustment strategy set is used to update the diet plan.

[0051] Secondly, this application also provides an AI-assisted dietary management and control device for elderly people with chronic diseases, comprising:

[0052] The preprocessing module is used to collect multidimensional user data and preprocess the multidimensional user data to obtain user health records; the multidimensional user data includes static data, disease data, physical examination data and preference data;

[0053] The target module is used to calculate the nutritional and health needs of the corresponding user based on the user's health record and obtain a personalized list of nutritional goals.

[0054] The rating module is used to calculate the suitability score for each food item based on a personalized nutritional goal list for the user; and based on the suitability score, it filters food items to obtain a safe food item library.

[0055] The filtering module is used to perform preliminary filtering of the recipe database based on the safe ingredient database to obtain a candidate recipe set;

[0056] The recipe module is used to construct optimization problems based on a personalized list of nutritional goals and a set of candidate recipes; solve the optimization problems to obtain recipe solutions; and use the recipe solutions to assist in dietary regulation.

[0057] Thirdly, this application also provides a computer device including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement any step of the method provided in the first aspect of this application.

[0058] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any step of the method provided in the first aspect of this application.

[0059] The aforementioned AI-assisted dietary management and control method, device, equipment, and media for elderly patients with chronic diseases collects multidimensional user data and preprocesses it to obtain user health records. This multidimensional user data includes static data, disease data, physical examination data, and preference data. Based on the user health records, the system calculates the corresponding user's nutritional health needs, resulting in a personalized nutritional target list. Based on this personalized nutritional target list, a suitability score is calculated for each food ingredient for the corresponding user. Based on the suitability score, ingredients are screened to obtain a safe food library. Based on the safe food library, a preliminary screening of the recipe database is performed to obtain a candidate recipe set. Based on the personalized nutritional target list and the candidate recipe set, an optimization problem is constructed and solved to obtain a recipe plan. This recipe plan is used to assist in dietary regulation. It can accurately uncover and dynamically match the complex relationship between disease and nutritional constraints to meet the personalized health needs of elderly patients, construct a relationship chain between disease and nutritional needs, deeply explore the potential connection between food efficacy and disease needs, avoid rule conflicts, and dynamically adjust recommended plans. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is a schematic diagram of the process of an AI-assisted dietary management and control method for chronic diseases in the elderly, provided in an embodiment of the present invention.

[0062] Figure 2 This is a schematic diagram of the structure of an AI-assisted dietary management and control device for chronic diseases in the elderly, provided in an embodiment of the present invention. Detailed Implementation

[0063] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0064] In one embodiment, such as Figure 1 As shown, an AI-assisted dietary management method for chronic diseases in the elderly is provided. This embodiment illustrates the method's application to a terminal, but it is understood that the method can also be applied to a server, or to a system including both a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0065] Step 101: Collect multidimensional user data and preprocess the multidimensional user data to obtain user health records; multidimensional user data includes static data, disease data, physical examination data and preference data.

[0066] Multidimensional user data comprises a collection of raw user data gathered from various dimensions. This includes static data (relatively fixed basic information such as age, gender, height, weight, family history of genetic diseases, and known allergens); disease data (information related to the user's current chronic diseases, such as disease type, diagnosis date, current medication, and disease control status); physical examination data (objective indicators from recent medical examinations, such as complete blood count, lipid profiles, liver function, kidney function, and physiological parameters like blood pressure and heart rate); and preference data (personal dietary preferences and lifestyle, such as taste preferences, food restrictions, dietary habits, and cooking limitations). The user health record is a structured, digitized comprehensive archive of user health information, formed after preprocessing. It is a complete user profile integrating all static, disease, physical examination, and preference data, serving as the foundation for all subsequent calculations and decisions. The terminal collects multidimensional user data through questionnaires, importing medical reports, and connecting to health devices. It checks and processes outliers, logical errors, and missing values ​​in the data. For missing values, methods such as mean imputation, prediction based on other data, or marking as special values ​​can be used to process them. Unstructured text information is transformed into structured data; indicators from different units are unified; categorized data is encoded; and the cleaned and standardized data are linked together to form a unified and complete data view of a single user, namely the user's health record, which comprehensively and accurately describes the user's health status and individual needs.

[0067] Step 102: Based on the user's health record, calculate the corresponding user's nutritional health needs to obtain a personalized nutritional goal list.

[0068] Specifically, nutritional health needs are the quantified requirements of a user for various nutrients and energy needed to maintain health, improve disease conditions, or meet specific physiological stages. A personalized nutrition goal list is a specific, quantified list of goals, detailing the recommended daily intake of various nutrients, including macronutrients and micronutrients, as well as specific upper or lower limits for chronic diseases. Based on the user's health record, the terminal uses clinical nutrition knowledge and rules to calculate, comprehensively considering static data such as age, gender, weight, and height, to calculate basal energy expenditure and total daily energy requirements; disease data to adjust nutritional needs according to specific diseases; and physical examination data to make targeted adjustments based on abnormal indicators, translating the analyzed nutritional needs into specific numerical goals. For example, a low-protein goal is set for patients with kidney disease, and a strict upper limit for sodium intake is set for patients with hypertension; users with high blood lipids need to control their intake of saturated fat and cholesterol.

[0069] Step 103: Based on the personalized nutrition target list, calculate the suitability score for each food ingredient for the corresponding user; and based on the suitability score, filter the food ingredients to obtain a safe food ingredient library.

[0070] Specifically, the suitability score is a quantifiable score used to evaluate how suitable a particular food is for a specific user. The higher the score, the more beneficial the food is to the user's health and the lower the risk. The safe food library is a screened collection of foods containing all foods whose suitability scores reach the safe threshold. The foods in the library are considered basic options that the user can safely consume. Based on the user's health records of allergens and explicit contraindications, the terminal directly removes absolutely unsuitable foods from the full food library to obtain a preliminary candidate food set. For each food in the preliminary candidate food set, a multi-dimensional evaluation is performed to calculate its suitability score. The nutritional knowledge graph is consulted to analyze the impact of the food on each disease the user suffers from, assigning a score to each impact and weighting the sum based on the severity of the disease. The nutritional components of the food are analyzed to determine its contribution to achieving the goals in the personalized nutrition target list. A scoring threshold is set, and foods with scores higher than the threshold are included in the safe food library, while foods with scores lower than the threshold are excluded.

[0071] Step 104: Based on the safe ingredient database, perform preliminary screening of the recipe database to obtain a candidate recipe set.

[0072] The recipe database is a large database containing numerous recipes and their detailed information. Each recipe includes required ingredients, cooking methods, estimated nutritional components, and culinary origin. The candidate recipe set is a pool of recipes selected from the entire recipe database that use only ingredients from the safe ingredient library; it serves as a candidate pool for the final recipe. The terminal iterates through each recipe in the recipe database, checking if all its required ingredients are included in the safe ingredient library. If a recipe uses only ingredients from the safe ingredient library, it is retained; if any ingredients are used outside the safe ingredient library, the recipe is excluded. All retained recipes form the candidate recipe set.

[0073] Step 105: Based on the personalized nutrition target list and candidate recipe set, construct an optimization problem; solve the optimization problem to obtain a recipe plan; the recipe plan is used to assist in dietary regulation.

[0074] In this context, the optimization problem is a mathematical modeling problem. The goal is to select a set of recipes from a candidate set that optimally achieves multiple objectives while satisfying various constraints. These objectives may include maximizing nutritional goals, minimizing disease risk, maximizing taste preferences, and minimizing cost. Decision variables represent the variables chosen in the optimization problem. Constraints are the hard requirements that the recipe scheme must meet, such as nutritional constraints, food safety constraints, and diversity constraints. The recipe scheme is the specific result obtained after solving the optimization problem; it is typically a detailed meal plan specifying the recipes, portion sizes, and cooking suggestions for each meal over the next few days. Based on the candidate recipe set and planning cycle, the terminal defines all possible decision variables and establishes a comprehensive mathematical function that integrates multiple optimization objectives. The function is usually in the form of a weighted sum and needs to be minimized. The objectives include: nutritional suitability, lowest disease risk, highest preference satisfaction, and lowest cost. Each objective is preceded by a weight coefficient to adjust its importance. Based on nutritional knowledge and the actual situation of users, a series of constraint equations are defined. The constructed optimization problem is solved using linear programming, integer programming, or heuristic algorithms to find the set of decision variable values ​​that minimize the objective function value under all constraints. Based on the solution results, the recipes corresponding to the decision variables with a value of 1 are extracted and combined into the final, executable recipe scheme.

[0075] This embodiment provides an AI-assisted dietary management and control method for elderly people with chronic diseases. It collects multi-dimensional user data and preprocesses this data to obtain user health records. The multi-dimensional user data includes static data, disease data, physical examination data, and preference data. Based on the user health records, the method calculates the corresponding user's nutritional health needs, resulting in a personalized nutritional target list. Based on the personalized nutritional target list, it calculates the suitability score for each ingredient for the corresponding user. Based on the suitability score, it filters ingredients to obtain a safe ingredient library. Based on the safe ingredient library, it performs preliminary screening of the recipe database to obtain a candidate recipe set. Based on the personalized nutritional target list and the candidate recipe set, it constructs an optimization problem and solves the optimization problem to obtain a recipe plan. The recipe plan is used to assist in dietary regulation. By introducing AI (Artificial Intelligence), this method can accurately discover and dynamically match the complex relationship between disease and nutritional constraints, meeting the personalized health needs of patients; it constructs a relationship chain between disease and nutritional needs, deeply exploring the potential connection between the efficacy of ingredients and disease requirements; it avoids conflicts between disease and recipe components, and dynamically adjusts the recommended recipe plan.

[0076] In one embodiment, an optimization problem is constructed based on a personalized list of nutritional goals and a candidate recipe set, including:

[0077] Step 201: Based on the candidate recipe set, define the variables corresponding to the optimization problem to obtain the decision variables.

[0078] In this context, decision variables are the unknowns that need to be determined through calculation in the optimization problem. They directly represent the decisions to be made. In this embodiment, decision variables are binary variables, represented by... This indicates that, in other words, if =1, indicating that the i-th recipe was selected as the j-th meal; if =0 indicates that the i-th recipe is not selected as the j-th meal. Here, index i traverses all recipes in the candidate recipe set, and index j traverses all meals. The terminal enumerates all possible recipe-meal combinations based on the candidate recipe set and the planned number of meals, and creates a decision variable for each combination. The set of variables constitutes the decision space of the entire optimization problem. The goal is to find an optimal set of 0 or 1 values ​​for the variables.

[0079] Step 202: Based on decision variables, a personalized nutrition target list, and the user's health record, construct a multi-objective function using the following formula:

[0080]

[0081] in, Multi-objective function As decision variables, For nutrition-related parameters, For disease-related parameters, For the target parameter of preference, For cost parameters, , , , These are the corresponding weighting coefficients.

[0082] Specifically, the multi-objective function is a mathematical function that integrates multiple objectives into a single, minimized overall objective value. A smaller value indicates a better overall performance of the corresponding recipe plan. Nutrition-related parameters are quantitative indicators that measure the deviation of the overall nutritional status of the selected recipe plan from the personalized nutritional target list. The formula is the weighted sum of squares of the deviations of all nutrients; the squared term means that both excess and deficiency are penalized, and large deviations are significantly amplified. A smaller value indicates that the recipe plan better aligns with the nutritional goals. Disease-related parameters are quantitative indicators that measure the overall risk of the selected recipe plan to the user's existing diseases. The calculation method is to weight and sum the disease risk scores of each recipe based on the user's health record. A smaller value indicates that the recipe plan is safer for the user. Preference target parameters are quantitative indicators that measure the degree to which the selected recipe plan matches the user's taste preferences. The calculation method is to simply add the preference matching scores of the selected recipes. A larger value indicates a higher potential user satisfaction with the recipe. Cost parameters are quantitative indicators that measure the total economic cost required to implement the recipe plan. The calculation method is to add the costs of the selected recipes together. The weighting coefficients are a series of positive numbers used to adjust the importance of different objectives in the overall objective function. For example, for a user with a serious illness, the disease weight would be set higher, meaning the system would prioritize ensuring the low-risk nature of the recipes; while for a user with a tight budget, the cost weight might be set higher. For any candidate recipe, the terminal calculates the value of the corresponding parameter according to its formula, multiplies this parameter value by its respective weighting coefficient, and then combines them into the overall objective function. The preference parameter is preceded by a negative sign because preference is the goal of maximization, while the overall goal is to minimize the objective function; therefore, the negative sign is used to transform it into a quantity that needs to be minimized.

[0083] Step 203: Based on the personalized nutrition goal list and user health record, generate constraints; the constraints include at least one of the following: hard nutrition constraints, food safety constraints, diversity constraints, and rationality constraints.

[0084] Specifically, constraints are the hard requirements that the recipe scheme must meet; they represent the bottom line of the optimization problem, with no room for compromise through weight adjustments. For example, constraints may include hard nutritional constraints to ensure that the intake of certain key nutrients strictly falls within safe ranges; food safety constraints, which are logical constraints ensuring that all ingredients in the final scheme come from a safe food library (indirectly satisfied when defining the candidate recipe set, but explicitly defined again in the model to mitigate risk); diversity constraints to prevent the repeated use of the same recipes within a short period; and rationality constraints to ensure that the scheme conforms to basic dietary common sense and habits. The terminal, based on a personalized nutritional goal list and key information from the user's health record, transforms these hard requirements into mathematical inequalities or equations regarding decision variables.

[0085] Step 204: Integrate decision variables, multi-objective functions, and constraints to obtain the optimization problem.

[0086] Specifically, the optimization problem is a complete mathematical problem that integrates all the aforementioned elements. It has a standard mathematical form and aims to find a set of values ​​for decision variables that minimize the value of the multi-objective function, while satisfying all constraints. The final step formally combines the three elements—decision variables, multi-objective function, and constraints—to form a complete and solvable mathematical model.

[0087] This embodiment achieves personalized emphasis for different users by adjusting the weighting coefficients; the constraints ensure the basic safety, rationality and feasibility of the generated recipe scheme; the dietary recommendation problem is completely and accurately transformed into a mathematical optimization problem, thereby making it possible for computers to make dietary recommendations and improving the reliability of dietary management and control.

[0088] In one embodiment, the parameters of the multi-objective function are obtained by the following method:

[0089] Step 301: Based on the personalized nutrition goal list, calculate the nutrition-related parameters using the following formula:

[0090]

[0091] in, For nutrition-related parameters, Let K be the importance weight of the k-th nutrient, K be the total number of nutrient types, k be the nutrient type index, i be the recipe index, N be the total number of recipes, j be the meal index, and M be the total number of meals. It is a binary variable. Let i be the content of the k-th nutrient provided by the i-th recipe at the j-th meal. This represents the user's daily target intake of the kth nutrient.

[0092] Specifically, nutrient content is known data from a recipe database, representing the amount of nutrient k that would be provided if recipe i were consumed at meal j. The daily target intake comes from a personalized nutrition goal list, representing the target daily intake of nutrient k set for the user. Nutrient weight is a weighting coefficient indicating the importance of nutrient k to the current user; for key nutrients, the weight is set higher, meaning a more stringent requirement for the plan to meet the target for that nutrient. For a specific nutrient k, the terminal needs to calculate the total amount provided by the entire diet plan. This involves iterating through all diets and meals, but only summing up the nutrient content of the selected diets to obtain the user's total intake of nutrient k under this plan. The total intake is then compared with the target value to calculate the relative deviation. The ratio represents the degree to which the intake deviates from the target. The relative deviation is squared to ensure that the deviation is always positive, as both insufficient and excessive intake are considered undesirable. Squaring amplifies larger deviations, meaning that a plan that deviates significantly from the target will be penalized much more severely than a slight deviation, thus forcing the optimization result to be closer to the target value. The weighted squared deviations of all nutrients are summed, with the weights ensuring that important nutrients account for a larger proportion of the total deviation.

[0093] Step 302: Based on the user's health record, calculate disease-related parameters using the following formula:

[0094]

[0095] in, For disease-related parameters, Let be the severity weight of the d-th disease, d be the disease index, D be the total number of disease types, i be the recipe index, N be the total number of recipes, j be the meal index, and M be the total number of meals. It is a binary variable. The risk score for disease d is calculated for the i-th recipe at the j-th meal.

[0096] Specifically, the disease risk score is a predefined score derived from a nutritional knowledge graph, representing the degree of negative impact of diet plan i on disease d at meal j. A higher score indicates a greater risk, making it a negative indicator. The disease severity weight is a weighting coefficient set based on the severity and current control status of diseases in the user's health record. More severe and poorly controlled diseases receive a higher weight, meaning the system will focus more on avoiding diets that aggravate the disease. For the user's disease d, the terminal calculates the total risk of the entire diet plan, summing the weighted total risks of all diseases. The weighting ensures that the risk of severe diseases dominates the overall risk assessment.

[0097] Step 303: Based on the user's health profile, calculate the preference target parameter using the following formula:

[0098]

[0099] in, Here, i represents the recipe index, N represents the total number of recipes, j represents the meal index, and M represents the total number of meals. It is a binary variable. The score is given to the degree of match between the i-th recipe and the user's taste preferences at the j-th meal.

[0100] Specifically, the preference matching score is a pre-calculated score that reflects the degree to which the i-th recipe matches the user's taste preferences in the j-th meal. The higher the score, the more likely the user is to like the recipe, making it a positive indicator. The terminal sums up the preference scores of all selected recipes to obtain this parameter.

[0101] Step 304: Based on the candidate recipe set, calculate the cost parameters using the following formula:

[0102]

[0103] in, Here, i is the recipe index, N is the total number of recipes, j is the meal index, and M is the total number of meals. It is a binary variable. The estimated cost of the i-th recipe in the j-th meal.

[0104] Specifically, the estimated cost is a known data point representing the estimated total cost of preparing the i-th recipe for the j-th meal. The terminal directly sums up the costs of all selected recipes to obtain the total cost of the plan.

[0105] This embodiment quantifies the vague concept of a good recipe into a precise and calculable mathematical standard by calculating multiple parameter values, thereby improving the accuracy and reliability of dietary conditioning and management.

[0106] In one embodiment, based on a personalized nutritional goal list, a suitability score is calculated for each food item in the nutritional knowledge graph for the corresponding user, including:

[0107] Step 401: Based on the user's health record, remove foods that are clearly allergenic or unsuitable for consumption from the preset full food database to obtain a preliminary candidate food set.

[0108] The full ingredient database is a pre-defined, comprehensive collection of all potentially considered ingredients; it's a general database, not targeted at any specific user. Ingredients with known allergies or contraindications are those directly determined to be excluded based on the user's health record. The preliminary candidate ingredient set is the first level of safe ingredients after removing all explicitly contraindicated ingredients from the full ingredient database. This ensures basic safety but doesn't yet involve a detailed assessment of the health benefits or minor risks. The terminal reads the user's health record, extracts the allergen list and dietary restriction list, and compares these lists with every ingredient in the full ingredient database. Any ingredient appearing on the allergen or restriction list is directly and unconditionally eliminated, and the remaining ingredients constitute the preliminary candidate ingredient set.

[0109] Step 402: Based on the nutritional knowledge graph, traverse each ingredient in the preliminary candidate ingredient set, and for each disease of the user, query the relationship between ingredients and diseases to obtain ingredient-disease pairs.

[0110] Specifically, a nutrition knowledge graph is a structured knowledge base that stores the complex relationships between entities such as food ingredients, nutrients, diseases, and health effects in a graphical format. A food-disease pair is a combination of a specific food ingredient and a specific disease of the user. The terminal iterates through each food ingredient in the initial candidate food ingredient set, and for each food ingredient, it iterates through each disease recorded in the user's health profile. For each combination of food ingredient and disease, it queries the nutrition knowledge graph to find direct or indirect relationships.

[0111] Step 403: Calculate the disease risk-benefit score for each pair of food disease pairs to obtain the food disease risk-benefit matrix.

[0112] Specifically, the disease risk-benefit score is a quantifiable score used to evaluate whether a particular food ingredient brings risk or benefit to a user's specific disease. A negative score indicates risk, a positive score indicates benefit, and zero indicates no significant impact. The food ingredient disease risk-benefit matrix is ​​a two-dimensional table or data structure that records the risk-benefit scores of all foods in the initial candidate food ingredient set for all of the user's diseases. The rows of the matrix represent foods, the columns represent diseases, and the values ​​in the cells are the corresponding risk-benefit scores. For each food ingredient-disease pair, the terminal calculates a specific score based on the relationships in the nutritional knowledge graph, calling predefined rules or algorithm models. The calculation considers the direction and intensity of the impact. All calculated scores are then filled into the corresponding positions in the matrix to form the complete food ingredient disease risk-benefit matrix.

[0113] Step 404: Based on the food disease risk-benefit matrix, obtain the suitability score between the user and the food.

[0114] The suitability score is a comprehensive rating representing the overall suitability of a particular food for a specific user. It's a single score that aggregates the risk-benefit profile of the food for all diseases, as well as its nutritional value. The higher the score, the more suitable the food is for the user. Based on the user's health record, the terminal assigns an importance weight to each disease. The weight calculation formula considers the severity of the disease and the degree of deviation from current indicators, meaning that more severe and poorly controlled diseases have a greater weight in the total score. The system assesses the contribution of the food's nutritional components to achieving the personalized nutrition goals list. For a given food, all scores in the corresponding row of the food-disease risk-benefit matrix are multiplied by the weight of the corresponding disease, then summed, and combined with the nutritional compatibility score to obtain the final suitability score for that food.

[0115] This embodiment transforms qualitative knowledge into quantitative data, comprehensively considering the fine balance of multiple health factors for users, and generates a quantitatively comparable standard. All ingredients can be ranked based on this score, thereby improving the targeted nature of dietary conditioning.

[0116] In one embodiment, based on a food disease risk-benefit matrix, a suitability score between the user and the food is obtained, including:

[0117] Step 501: Based on the user's health record, assign importance weights to diseases using the following formula:

[0118]

[0119] in, Let the importance weight of disease d be... The baseline severity of disease type d is given by [reference to a specific disease type], where D represents the total number of disease types. Let be the normalized deviation of the key indicator for the d-th disease, where d is the disease index.

[0120] Specifically, the baseline severity is a pre-defined numerical value based on medical knowledge, representing the severity level of the d-th disease itself. It is a fixed value, independent of the user's current condition, derived from medical guidelines or expert knowledge. The normalized deviation is a dynamically calculated value based on the latest physical examination data in the user's health record, typically ranging from 0 to 1. It quantifies the degree to which the key indicator of the user's d-th disease deviates from its ideal control target; a larger deviation indicates less ideal disease control and a greater need for urgent attention. The disease importance weight is a value between 0 and 1, with the sum of the weights of all the user's diseases being 1. It integrates the inherent severity of the disease and the user's current control status, representing the proportion that the disease factor should account for when evaluating ingredients. For each disease d, the terminal calculates... , This ensures that the immediate deviation is 0, and the inherent severity of the disease still plays a role. The greater the deviation, the larger the product, indicating that the urgency and importance of the disease are amplified. The adjustment factors of all the user's diseases are added together, and the sum is used as the denominator to ensure that the sum of all weights is 1, thus achieving normalization. The adjustment factor of each disease is divided by the sum of the denominators to obtain the final importance weight of the disease.

[0121] Step 502: Iterate through each nutrient in the personalized nutrition goal list, calculate the contribution of food to the nutrient, and obtain the nutrition fit score.

[0122] The nutritional fit score is a quantitative score representing the contribution of the food's nutrients to achieving the various nutritional goals in the personalized nutrition goal list. A higher score indicates that the food is more helpful in meeting the user's nutritional needs. The terminal analyzes each nutrient listed in the personalized nutrition goal list. For each nutrient, based on its content in the food, the contribution value of the food is evaluated. For example, if the user's list requires increased calcium intake, then high-calcium foods will have a high contribution value for that nutrient. The contribution value calculation considers the nutrient's bioavailability and whether it meets the need or increases the risk of exceeding the target. The contribution values ​​of the food to all nutrients are combined to obtain an overall nutritional fit score. The aggregation process considers the importance of nutrients, assigning higher weight to deficient nutrients.

[0123] Step 503: Based on the importance weight, the food disease risk-benefit matrix, and the nutritional compatibility score, a weighted summation calculation is performed to obtain the suitability score.

[0124] The suitability score is a comprehensive quantitative score representing the overall suitability of the ingredient for the current user, balancing the ingredient's disease risk-benefit and nutritional contribution. For the ingredient currently being evaluated, the terminal extracts the risk-benefit score for each disease *d* of the user from the ingredient's disease risk-benefit matrix. Each risk-benefit score is multiplied by the importance weight of the corresponding disease, meaning that the more important a disease is to the user, the greater the influence of its corresponding risk-benefit score in the total score. All weighted scores are summed to obtain the ingredient's total score in the disease dimension. This total disease dimension score is then combined with the nutritional compatibility score using a weighted sum to obtain the suitability score.

[0125] This embodiment generates a comprehensive, quantitative, and highly personalized assessment result. The suitability score not only tells the system which ingredients are safer, but also which ingredients offer the best overall cost-effectiveness after balancing the user's disease risks and nutritional needs, thereby improving the targeting of dietary regulation.

[0126] In one embodiment, after solving the optimization problem and obtaining the recipe scheme, the method further includes:

[0127] Step 601: Compare the recipe plan with the user's health record, identify the key points of the recipe plan, and obtain a list of explanatory points; the key points include highlighting the key points, explaining the trade-offs, and stating the contraindications.

[0128] The explanation checklist is a structured list outlining key aspects to emphasize when explaining a recipe to a user. These include: highlighting key features (designs that are particularly outstanding and significantly beneficial to the user's health); explaining trade-offs (clarifying the reasons behind seemingly contradictory or compromised design decisions); and specifying contraindications (clearly explaining why certain ingredients or dishes that the user might expect are deliberately avoided). The terminal compares the recipe generated by the optimization algorithm with the user's health profile and personalized nutrition goal list, identifying frequently occurring or intentionally added ingredients or nutrients that clearly benefit the user's primary health concerns. It analyzes the optimization problem's solution, identifying compromises made to meet higher-priority goals, confirming whether the recipe successfully avoids all user contraindications, and preparing explanations to enhance user confidence. All identified key points are categorized and organized into a clear explanation checklist, providing an outline for further in-depth explanations.

[0129] Step 602: For each point in the explanation list, query the nutrition knowledge graph to obtain the decision logic chain.

[0130] Specifically, the decision logic chain is a series of relationships based on a nutrition knowledge graph. It clearly illustrates the reasoning process from food ingredients to health effects, connecting the specific design of the solution with the user's health goals. For each point in the explanation list, the terminal performs a path query in the nutrition knowledge graph, extracting and organizing the relationships to form one or more complete causal chains, i.e., the decision logic chain. For example, for a highlight, the query path might be: food ingredient A used in the solution → contains nutrient B → has a positive impact on disease C → meets user goal D; for trade-offs and contraindications, the query path might be: food ingredient E avoided in the solution → contains risky components → will aggravate disease G → violates user constraints H.

[0131] Step 603: Based on natural language generation, the decision logic chain is converted into natural language to obtain the explanation statement of the corresponding recipe scheme.

[0132] Specifically, Natural Language Generation (NLP) is a branch of artificial intelligence that specializes in converting structured data or logic into fluent natural language text. Explanatory statements are complete and coherent textual explanations of a specific point generated by NLP technology; they are the linguistic expression of a decision-making logic chain. The terminal takes the decision-making logic chain as input to the NLP model. Based on predefined grammatical templates and a vocabulary, the model expresses each node in the chain using appropriate words and sentence structures. By adding conjunctions and adjusting word order, it combines fragmented information into a coherent and fluent statement.

[0133] Step 604: Based on the explanation statement and recipe scheme, generate a recipe interpretability report; the recipe interpretability report is used to assist in understanding the recipe scheme.

[0134] The recipe interpretability report is a comprehensive document attached to the recipe plan. It includes a plan summary and provides detailed explanations of each point in the explanation checklist, consisting of explanatory statements. The aim is to comprehensively elucidate the scientific logic and personalized considerations behind the plan. The terminal creates a report framework, inserting the recipe plan itself as the main body of the report. For key design elements within the plan, all explanatory statements are categorized and inserted into the corresponding sections of the report. The report is then formatted for clarity and readability, ultimately generating a complete recipe interpretability report.

[0135] This embodiment greatly enhances the transparency, credibility, and user acceptance of recipe plans by generating user-friendly and comprehensive documents, which is a key step in achieving personalized health management.

[0136] In one embodiment, after generating a recipe interpretability report based on the interpretation statement and recipe scheme, the method further includes:

[0137] Step 701: Receive dietary record data and subjective feeling data from user feedback, and simultaneously acquire user health monitoring data to obtain a behavioral time series dataset.

[0138] Dietary record data consists of objective records of a user's actual food intake. Subjective feeling data is a user's subjective description of their experience and feelings after eating, which may include post-meal satiety, satisfaction, presence of discomfort symptoms, energy levels, etc. Health monitoring data consists of objective physiological indicators obtained through devices or measurements. The behavioral time series dataset is a structured dataset that aligns and integrates the above three types of data in chronological order. Each data point contains a timestamp, as well as the corresponding diet, feelings, and health indicators at that moment, thus forming a sequence that can be analyzed for causal relationships. The terminal receives data through multiple channels: users manually input dietary records and subjective feelings; data from health monitoring devices is automatically synchronized through the device interface, unifying all data onto the same timeline, handling missing and outlier values, and integrating the cleaned and aligned data into a table or database containing time series data, i.e., the behavioral time series dataset.

[0139] Step 702: Input the behavioral time series dataset into the pre-trained time series analysis model to obtain an anomaly pattern report.

[0140] Specifically, the pre-trained time series analysis model is an AI model that has been trained with a large amount of historical data. It excels at identifying meaningful patterns, trends, and anomalies from time series data, including anomaly detection algorithms and pattern matching algorithms. The anomaly pattern report is a structured output of the model after analysis. The terminal inputs the behavioral time series dataset into the pre-trained model, and the model runs its algorithm to compare the current user's health data with the expected trajectory of the diet plan, identify anomaly points or periods that deviate significantly from the expected trajectory, and analyze whether these anomalies have a stable temporal correlation with specific dietary events, thus forming anomaly patterns. The model summarizes the identified patterns, the degree of deviation, the frequency of occurrence, and other information into an anomaly pattern report.

[0141] Step 703: Based on the abnormal pattern report, query the nutrition knowledge graph to obtain the root cause inference conclusion.

[0142] Specifically, the root cause inference conclusion is a speculative judgment based on the abnormal pattern report and the nutritional knowledge graph, making a prediction of the most likely cause of the abnormality. The terminal parses the abnormal pattern report, extracts key elements, and uses these key elements as a starting point to conduct in-depth queries and reasoning in the nutritional knowledge graph to construct a reasonable logical chain from food intake to the observation of the abnormality. The degree of consistency between this logical chain and the existing knowledge in the knowledge graph is evaluated, and a confidence level is given, ultimately forming the root cause inference conclusion.

[0143] Step 704: Based on the root cause inference conclusions, the diet plan is adjusted according to the predefined adjustment rule base and clinical nutrition principles to obtain a personalized adjustment strategy set; the personalized adjustment strategy set is used to update the diet plan.

[0144] The predefined adjustment rule base is a database storing corresponding adjustment rules. Clinical nutrition principles are general medical and nutritional guidelines that guide diet adjustments, ensuring the safety and scientific validity of the adjustment strategies. The personalized adjustment strategy set is a collection of specific, executable adjustment instructions. The terminal matches the root cause inference conclusions with the conditions in the adjustment rule base, identifies all applicable adjustment rules, and, in conjunction with clinical nutrition principles, verifies and refines the matched rule suggestions to generate specific and feasible adjustment strategies. All generated strategies are then integrated into the personalized adjustment strategy set.

[0145] This embodiment achieves continuous optimization of personalized health management through dynamic updates, thereby improving the reliability and accuracy of dietary regulation and control.

[0146] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0147] Based on the same inventive concept, this application also provides an AI-assisted dietary management and control device for implementing the aforementioned AI-assisted dietary management and control method for chronic diseases in the elderly. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more embodiments of the AI-assisted dietary management and control device for chronic diseases in the elderly provided below can be found in the limitations of the AI-assisted dietary management and control method for chronic diseases in the above text, and will not be repeated here.

[0148] In one exemplary embodiment, such as Figure 2 As shown, an AI-assisted dietary management and control device 800 for elderly people with chronic diseases is provided, comprising:

[0149] The preprocessing module 801 is used to collect multidimensional user data and preprocess the multidimensional user data to obtain user health records; the multidimensional user data includes static data, disease data, physical examination data and preference data;

[0150] The target module 802 is used to calculate the nutritional health needs of the corresponding user based on the user's health record and obtain a personalized nutritional target list.

[0151] The scoring module 803 is used to calculate the suitability score for each food ingredient based on a personalized nutritional target list for the user; and based on the suitability score, to filter food ingredients and obtain a safe food ingredient library.

[0152] The filtering module 804 is used to perform preliminary filtering of the recipe database based on the safe ingredient database to obtain a candidate recipe set;

[0153] The recipe module 805 is used to construct an optimization problem based on a personalized nutritional goal list and a candidate recipe set; and to solve the optimization problem to obtain a recipe plan; the recipe plan is used to assist in dietary regulation.

[0154] Furthermore, recipe module 805 is also used for:

[0155] Based on the candidate recipe set, variables corresponding to the optimization problem are defined to obtain the decision variables;

[0156] Based on decision variables, a personalized nutrition target list, and user health records, a multi-objective function is constructed using the following formula:

[0157]

[0158] in, Multi-objective function As decision variables, For nutrition-related parameters, For disease-related parameters, For the target parameter of preference, For cost parameters, , , , These are the corresponding weighting coefficients;

[0159] Based on the personalized nutrition target list and the user's health record, constraints are generated; the constraints include at least one of the following: hard nutrition constraints, food safety constraints, diversity constraints, and rationality constraints.

[0160] By integrating decision variables, multi-objective functions, and constraints, the optimization problem is obtained.

[0161] Furthermore, recipe module 805 is also used for:

[0162] Based on a personalized list of nutrition goals, nutrition-related parameters are calculated using the following formula:

[0163]

[0164] in, For nutrition-related parameters, Let K be the importance weight of the k-th nutrient, K be the total number of nutrient types, k be the nutrient type index, i be the recipe index, N be the total number of recipes, j be the meal index, and M be the total number of meals. It is a binary variable. Let i be the content of the k-th nutrient provided by the i-th recipe at the j-th meal. This represents the user's target daily intake of the kth nutrient.

[0165] Based on the user's health record, disease-related parameters are calculated using the following formula:

[0166]

[0167] in, For disease-related parameters, Let be the severity weight of the d-th disease, d be the disease index, D be the total number of disease types, i be the recipe index, N be the total number of recipes, j be the meal index, and M be the total number of meals. It is a binary variable. For the i-th recipe, score the risk of the d-th disease at the j-th meal;

[0168] Based on the user's health profile, the preference target parameter is calculated using the following formula:

[0169]

[0170] in, Here, i represents the recipe index, N represents the total number of recipes, j represents the meal index, and M represents the total number of meals. It is a binary variable. The score is given to the degree of match between the i-th recipe and the user's taste preferences at the j-th meal.

[0171] Based on the candidate recipe set, the cost parameters are calculated using the following formula:

[0172]

[0173] in, Here, i is the recipe index, N is the total number of recipes, j is the meal index, and M is the total number of meals. It is a binary variable. The estimated cost of the i-th recipe in the j-th meal.

[0174] Furthermore, the scoring module 803 is also used for:

[0175] Based on the user's health record, foods that are clearly allergenic or unsuitable for consumption are removed from the preset full food database to obtain a preliminary set of candidate foods;

[0176] Based on a nutritional knowledge graph, we traverse each ingredient in the initial candidate ingredient set, and for each disease of the user, we query the relationship between ingredients and diseases to obtain ingredient-disease pairs.

[0177] For each pair of food-disease pairs, calculate the disease risk-reward score to obtain the food-disease risk-reward matrix;

[0178] Based on the food disease risk-benefit matrix, a suitability score is obtained between the user and the food.

[0179] Furthermore, the scoring module 803 is also used for:

[0180] Based on users' health records, the following formula is used to assign importance weights to diseases:

[0181]

[0182] in, Let the importance weight of disease d be... The baseline severity of disease type d is given by [reference to a specific disease type], where D represents the total number of disease types. denoted as the normalized deviation of the key indicator for the d-th disease, where d is the disease index;

[0183] Iterate through each nutrient in the personalized nutrition goal list, calculate the contribution of food to the nutrient, and obtain the nutrition fit score.

[0184] The suitability score is obtained by weighting and summing the importance weights, the food disease risk-benefit matrix, and the nutritional compatibility score.

[0185] Furthermore, the device also includes an interpretation module for:

[0186] By comparing the recipe plan with the user's health record, the key points of the recipe plan are identified, and a list of explanatory points is obtained; the key points include highlighting the advantages, explaining the trade-offs, and stating the contraindications;

[0187] For each key point in the explanation list, the nutrition knowledge graph is consulted to obtain the decision logic chain;

[0188] Based on natural language generation, the decision logic chain is transformed into natural language to obtain the corresponding recipe solution's explanation statement;

[0189] Based on the explanation statements and recipe scheme, a recipe interpretability report is generated; the recipe interpretability report is used to help understand the recipe scheme.

[0190] Furthermore, the device also includes an update module for:

[0191] Receive dietary records and subjective feelings from users, and simultaneously acquire users' health monitoring data to obtain a behavioral time series dataset;

[0192] Input the behavioral time series dataset into a pre-trained time series analysis model to obtain an abnormal pattern report;

[0193] Based on the abnormal pattern report, the nutritional knowledge graph was queried to obtain the root cause inference conclusion;

[0194] Based on the root cause inference conclusions, the diet plan is adjusted according to the predefined adjustment rule base and clinical nutrition principles to obtain a personalized adjustment strategy set; the personalized adjustment strategy set is used to update the diet plan.

[0195] In one embodiment, a computer device is provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the AI-assisted dietary management and control method for chronic diseases in the elderly as described above.

[0196] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0197] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The components described as separate parts may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this disclosure according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0198] The above-described embodiments are merely illustrative of several implementation methods of the embodiments of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the embodiments of this application, and these modifications and improvements all fall within the protection scope of the embodiments of this application.

Claims

1. A method for managing and controlling chronic diseases in the elderly based on AI assistance, characterized in that, The method includes: Collect multidimensional user data and preprocess the multidimensional user data to obtain user health records; the multidimensional user data includes static data, disease data, physical examination data and preference data; Based on the user's health record, calculate the corresponding user's nutritional health needs to obtain a personalized nutritional target list. Based on the personalized nutrition target list, a suitability score is calculated for each food ingredient corresponding to the user; and based on the suitability score, the food ingredients are screened to obtain a safe food ingredient library; Based on the aforementioned safe food ingredient database, a preliminary screening of the recipe database is performed to obtain a candidate recipe set; Based on the personalized nutrition target list and the candidate recipe set, an optimization problem is constructed; and the optimization problem is solved to obtain a recipe scheme; the recipe scheme is used to assist in dietary regulation.

2. The method according to claim 1, characterized in that, Based on the personalized nutrition target list and the candidate recipe set, an optimization problem is constructed, including: Based on the candidate recipe set, variables corresponding to the optimization problem are defined to obtain decision variables; Based on the decision variables, the personalized nutrition target list, and the user's health record, a multi-objective function is constructed using the following formula: in, Multi-objective function As decision variables, For nutrition-related parameters, For disease-related parameters, For the target parameter of preference, For cost parameters, , , , These are the corresponding weighting coefficients; Based on the personalized nutrition target list and the user's health record, constraints are generated; the constraints include at least one of the following: hard nutrition constraints, food safety constraints, diversity constraints, and rationality constraints. By integrating the decision variables, the multi-objective function, and the constraints, the optimization problem is obtained.

3. The method according to claim 2, characterized in that, The parameters in the multi-objective function are obtained through the following method: Based on the personalized nutrition target list, the nutrition-related parameters are calculated using the following formula: in, For nutrition-related parameters, Let K be the importance weight of the k-th nutrient, K be the total number of nutrient types, k be the nutrient type index, i be the recipe index, N be the total number of recipes, j be the meal index, and M be the total number of meals. It is a binary variable. Let i be the content of the k-th nutrient provided by the i-th recipe at the j-th meal. This represents the user's target daily intake of the kth nutrient. Based on the user's health record, the disease-related parameters are calculated using the following formula: in, For disease-related parameters, Let be the severity weight of the d-th disease, d be the disease index, D be the total number of disease types, i be the recipe index, N be the total number of recipes, j be the meal index, and M be the total number of meals. It is a binary variable. For the i-th recipe, score the risk of the d-th disease at the j-th meal; Based on the user's health profile, the preference target parameter is calculated using the following formula: in, Here, i represents the recipe index, N represents the total number of recipes, j represents the meal index, and M represents the total number of meals. It is a binary variable. The score is given to the degree of match between the i-th recipe and the user's taste preferences at the j-th meal. Based on the candidate recipe set, the cost parameter is calculated using the following formula: in, Here, i is the recipe index, N is the total number of recipes, j is the meal index, and M is the total number of meals. It is a binary variable. The estimated cost of the i-th recipe in the j-th meal.

4. The method according to claim 1, characterized in that, The process of calculating a suitability score for each food item in the nutrition knowledge graph based on the personalized nutrition goal list includes: Based on the user's health record, foods that are clearly allergenic or unsuitable for consumption are removed from the preset full food database to obtain a preliminary candidate food set; Based on the nutritional knowledge graph, each ingredient in the preliminary candidate ingredient set is traversed, and for each disease of the user, the relationship between the ingredient and the disease is queried to obtain ingredient-disease pairs; For each pair of food disease pairs, calculate the disease risk-benefit score to obtain the food disease risk-benefit matrix. Based on the food ingredient disease risk-benefit matrix, the suitability score between the user and the food ingredient is obtained.

5. The method according to claim 4, characterized in that, The process of obtaining the suitability score between the user and the food based on the food's disease risk-benefit matrix includes: Based on the user's health record, the disease is assigned an importance weight using the following formula: in, Let the importance weight of disease d be... The baseline severity of disease type d is given by [reference to a specific disease type], where D represents the total number of disease types. denoted as the normalized deviation of the key indicator for the d-th disease, where d is the disease index; Iterate through each nutrient in the personalized nutrition target list, calculate the contribution of the food to the nutrient, and obtain the nutrition fit score. The suitability score is obtained by performing a weighted summation based on the importance weight, the food disease risk-benefit matrix, and the nutritional compatibility score.

6. The method according to claim 1, characterized in that, After solving the optimization problem to obtain the recipe scheme, the process further includes: By comparing the recipe plan with the user's health record, the key points of the recipe plan are identified, and a list of key points for explanation is obtained; the key points include highlighting key features, explaining trade-offs, and stating contraindications; For each of the key points in the explanation list, the nutrition knowledge graph is queried to obtain the decision logic chain; Based on natural language generation, the decision logic chain is converted into natural language to obtain the explanation statement corresponding to the recipe scheme; Based on the explanation statement and the recipe scheme, a recipe interpretability report is generated; the recipe interpretability report is used to assist in understanding the recipe scheme.

7. The method according to claim 6, characterized in that, After generating the recipe interpretability report based on the interpretation statement and the recipe scheme, the process further includes: Receive the dietary record data and subjective feeling data fed back by the user, and simultaneously acquire the user's health monitoring data to obtain a behavioral time series dataset; The behavioral time series dataset is input into a pre-trained time series analysis model to obtain an abnormal pattern report. Based on the abnormal pattern report, the nutritional knowledge graph is queried to obtain the root cause inference conclusion; Based on the root cause inference conclusions, the diet plan is adjusted according to a predefined adjustment rule base and clinical nutrition principles to obtain a personalized adjustment strategy set; the personalized adjustment strategy set is used to update the diet plan.

8. An AI-assisted dietary management and control device for elderly people with chronic diseases, characterized in that, The device includes: The preprocessing module is used to collect multidimensional user data and preprocess the multidimensional user data to obtain user health records; the multidimensional user data includes static data, disease data, physical examination data and preference data; The target module is used to calculate the nutritional health needs of the corresponding user based on the user's health record and obtain a personalized nutritional target list. The scoring module is used to calculate a suitability score for each food ingredient for the user based on the personalized nutrition target list; and to filter the food ingredients based on the suitability score to obtain a safe food ingredient library. The filtering module is used to perform preliminary filtering of the recipe database based on the safe food ingredient database to obtain a candidate recipe set; The recipe module is used to construct an optimization problem based on the personalized nutrition target list and the candidate recipe set; and to solve the optimization problem to obtain a recipe scheme; the recipe scheme is used to assist in dietary regulation.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images