Personalized healthy diet scheme generation system based on artificial intelligence

Abstract

The invention relates to the technical field of data analysis, and particularly discloses a personalized healthy diet scheme generation system based on artificial intelligence, which is based on a food nutritional ingredient knowledge base, determines daily intake nutritional ingredient data of a user according to recent diet records of the user, and combines personal health data input by the user to generate a personalized healthy diet scheme. The deep learning algorithm is further introduced to perform joint interaction analysis on the personal health data of the user and the daily intake nutritional ingredient data so as to predict the daily required nutritional ingredients of the user, and then the daily required nutritional ingredients of the user are predicted on the basis of the obtained daily required nutritional ingredient data of the user and in combination with the diet preference information of the user. And generating a personalized healthy diet scheme according with the diet preference and the nutritional requirement of the user by means of a large language model. In this way, the problem that nutritionist resources are scarce can be effectively solved, personalized customization of diet schemes is achieved, and then efficient and accurate healthy diet suggestions can be provided for large-scale crowds.

Description

Technical Field

[0001] The present application relates to the field of data analysis technology, and more specifically, to a personalized healthy diet plan generation system based on artificial intelligence. Background Art

[0002] As people's living standards improve and their health awareness grows, a healthy diet, as the cornerstone of maintaining good health, is gaining increasing attention. A healthy diet not only prevents chronic diseases such as heart disease, diabetes, and obesity, but also improves overall quality of life, enhances mental well-being, and strengthens immunity.

[0003] Traditional healthy diet plans rely on the experience of professional nutritionists. However, this approach has numerous limitations. First, nutritionists are scarce, their services are expensive, and they struggle to meet the needs of large populations. Second, empirical dietary recommendations lack precision, fail to fully account for individual differences in dietary preferences, health status, and nutritional needs, and are difficult to track and adjust in real time.

[0004] Therefore, we look forward to a personalized healthy diet plan generation system based on artificial intelligence. Summary of the Invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a personalized healthy diet plan generation system based on artificial intelligence, which is based on a food nutrient knowledge base, determines the user's daily nutrient intake data based on the user's recent diet records, and combines the personal health data input by the user. By further introducing a deep learning algorithm, the user's personal health data and daily nutrient intake data are jointly interactively analyzed to predict the user's daily nutrient requirements. Then, based on the obtained user's daily nutrient requirement data, combined with the user's dietary preference information, a personalized healthy diet plan that meets the user's dietary preferences and nutritional needs is generated with the help of a large language model. In this way, the problem of scarce nutritionist resources can be effectively solved, and personalized customization of diet plans can be achieved, thereby providing efficient and accurate healthy diet recommendations for a large population.

[0006] Accordingly, according to one aspect of the present application, a personalized healthy diet plan generation system based on artificial intelligence is provided, which includes: A user health data acquisition module is used to receive personal health data input by the user; A user diet record acquisition module is used to acquire the user's recent diet records, which include the types of food consumed daily, the amount consumed, and the cooking method; A nutrient component analysis module, configured to perform nutrient component analysis on the diet record based on a food nutrient component knowledge base to obtain a time series of the user's daily nutrient component intake data; A nutrient requirement prediction module, configured to perform nutrient requirement prediction based on deep learning on the time series of the personal health data and the user's daily nutrient intake data to obtain the nutrient requirement data required by the user daily; User dietary preference acquisition module, used to obtain dietary preference information input by the user; A diet plan generating module is used to generate a healthy diet plan based on the diet preference information and the nutritional component data required by the user daily.

[0007] Compared with the prior art, the personalized healthy diet plan generation system based on artificial intelligence provided by this application is based on a food nutrient knowledge base, determines the user's daily nutrient intake data based on the user's recent diet records, and combines the personal health data input by the user. By further introducing a deep learning algorithm, the user's personal health data and daily nutrient intake data are jointly interactively analyzed to predict the user's daily nutrient requirements. Then, based on the obtained user's daily nutrient requirements data, combined with the user's dietary preference information, a personalized healthy diet plan that meets the user's dietary preferences and nutritional needs is generated with the help of a large language model. In this way, the problem of scarce nutritionist resources can be effectively solved, and personalized customization of diet plans can be achieved, thereby providing efficient and accurate healthy diet recommendations for large groups of people. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0009] Figure 1 This is a block diagram of a system for generating a personalized healthy diet plan based on artificial intelligence according to an embodiment of the present application.

[0010] Figure 2 This is a data flow diagram of the artificial intelligence-based personalized healthy diet plan generation system according to an embodiment of the present application.

[0011] Figure 3 This is a block diagram of a nutrient requirement prediction module in an artificial intelligence-based personalized healthy diet plan generation system according to an embodiment of the present application.

[0012] Figure 4This is a block diagram of a nutrient intake timing analysis unit in an artificial intelligence-based personalized healthy diet plan generation system according to an embodiment of the present application.

[0013] Figure 5 This is a block diagram of a feature interaction unit in an artificial intelligence-based personalized healthy diet plan generation system according to an embodiment of the present application.

[0014] Figure 6 This is a block diagram of a sparse constraint aggregation subunit in an artificial intelligence-based personalized healthy diet plan generation system according to an embodiment of the present application. DETAILED DESCRIPTION

[0015] As used in this application and the claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not intended to refer to the singular but may include the plural. Generally speaking, the terms "comprises" and "include" only indicate the inclusion of the steps and elements specifically identified, and these steps and elements do not constitute an exclusive list. A method or apparatus may also include other steps or elements.

[0016] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0017] Flowcharts are used in this application to illustrate the operations performed by the systems according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, the various steps may be processed in reverse order or simultaneously, as needed. Furthermore, other operations may be added to these processes, or one or more operations may be removed from these processes.

[0018] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0019] It is worth noting that in this application, all actions to obtain data are carried out in compliance with the relevant data protection laws and policies of the country where they are located and with the authorization given by the owner of the corresponding device.

[0020] In response to the technical problems described in the above background technology, this application proposes a personalized healthy diet plan generation system based on artificial intelligence. It is based on a food nutrient knowledge base, determines the user's daily nutrient intake data based on the user's recent diet records, and combines the personal health data input by the user. By further introducing a deep learning algorithm, the user's personal health data and daily nutrient intake data are jointly interactively analyzed to predict the user's daily nutrient requirements. Then, based on the obtained user's daily nutrient requirements data, combined with the user's dietary preference information, a personalized healthy diet plan that meets the user's dietary preferences and nutritional needs is generated with the help of a large language model. In this way, the problem of scarce nutritionist resources can be effectively solved, and personalized customization of diet plans can be achieved, thereby providing efficient and accurate healthy diet recommendations for large groups of people.

[0021] Figure 1 This is a block diagram of a system for generating a personalized healthy diet plan based on artificial intelligence according to an embodiment of the present application. Figure 2 Schematic diagram of data flow of a system for generating personalized healthy diet plans based on artificial intelligence according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the artificial intelligence-based personalized healthy diet plan generation system 100 includes: a user health data acquisition module 110, which is used to receive personal health data input by a user; a user diet record acquisition module 120, which is used to obtain the user's recent diet records, and the diet records include the types of food consumed daily, the intake amount and the cooking method; a nutrient component analysis module 130, which is used to perform nutrient component analysis on the diet records based on a food nutrient component knowledge base to obtain a time series of the user's daily nutrient component intake data; a nutrient component demand prediction module 140, which is used to perform nutrient component demand prediction based on deep learning on the time series of the personal health data and the user's daily nutrient component intake data to obtain the nutrient component data required by the user daily; a user diet preference acquisition module 150, which is used to obtain the diet preference information input by the user; and a diet plan generation module 160, which is used to generate a healthy diet plan based on the diet preference information and the nutrient component data required by the user daily.

[0022] In the aforementioned AI-based personalized healthy diet plan generation system, the user health data acquisition module 110 is used to receive personal health data input by the user. The personal health data includes height, weight, age, gender, body fat percentage, blood pressure, blood sugar, and blood lipids. It should be understood that individual health data is the foundation for developing a diet plan. In traditional nutrition, indicators such as height, weight, age, and gender can reflect the basal metabolic rate (BMR) and daily energy requirement (TDEE), while data such as body fat percentage, blood pressure, blood sugar, and blood lipids are used to reflect the user's metabolic status and potential health risks (such as hypertension and diabetes). By receiving personal health data input by the user, this application helps to build a user health profile, providing basic guidance for subsequent nutrient analysis and diet plan development.

[0023] Specifically, height and weight are key factors in assessing Body Mass Index (BMI), a metric crucial for determining whether an individual is at risk of overweight or obesity. These two parameters are not only the basis for calculating Basal Metabolic Rate (BMR), which estimates the energy required to maintain basic life activities at rest, but also the first step in understanding a person's body size and potential health risks. As the body's metabolic rate changes with age, understanding the user's actual age can help more accurately adjust energy requirement calculations. Age is also crucial in determining Total Daily Energy Expenditure (TDEE), as daily activity levels and metabolic rates vary significantly across age groups. Furthermore, gender differences also influence metabolic rate and TDEE. Men generally have higher muscle mass, which translates to a higher basal metabolic rate. Therefore, accurately recording this basic information when entering personal health data is crucial. This not only helps build a preliminary health model but also lays a solid foundation for subsequent, more in-depth health analysis.

[0024] In addition to the basic information mentioned above, body fat percentage, as an important indicator of the proportion of body fat, is directly related to an individual's cardiovascular health. A high body fat percentage may indicate a higher risk of chronic diseases such as heart disease and diabetes. Therefore, incorporating body fat percentage into health records can help the system better understand the user's body composition and adjust dietary recommendations accordingly. Body fat percentage reflects the proportion of fat tissue in the body relative to total body weight and is a more accurate description of a person's health status than weight alone. For example, two people of the same weight may have completely different health risk levels due to different body fat percentages. For those with a high body fat percentage, reducing body fat is often a key goal for improving health. Therefore, by collecting a user's body fat percentage data, the system can provide customized dietary recommendations for different body fat levels, such as increasing protein intake to promote muscle growth, choosing low-calorie-density foods to control total calorie intake, and other measures to help reduce body fat percentage.

[0025] Blood pressure is another key indicator of cardiovascular health, and its level directly affects the workload of the heart. People with hypertension need to pay special attention to their salt intake to avoid worsening their condition. By collecting users' blood pressure data, the system can identify potential risks of hypertension and propose corresponding dietary adjustment strategies. Blood pressure data not only includes systolic and diastolic blood pressure values, but also takes into account the differences in normal blood pressure ranges between individuals. For example, some people are born with low blood pressure, while for others, even a slight increase in blood pressure may mean potential health problems. Based on this, the system can recommend appropriate reductions in sodium intake and increases in potassium intake based on each user's specific blood pressure situation, and encourage the consumption of fresh vegetables and fruits and other foods rich in potassium ions to help regulate blood pressure levels and prevent the occurrence of cardiovascular disease.

[0026] Blood sugar levels reflect the body's ability to process sugar and are of great significance for preventing diabetes and its complications. Long-term high blood sugar levels can damage the functions of multiple organs, especially the eyes, kidneys, and nerves. Recording the user's blood sugar data enables the system to identify potential diabetes risks and propose corresponding dietary adjustment strategies. Blood sugar management is not limited to controlling sugar intake, but also includes balancing the proportion of carbohydrates, protein, and fat in the diet. For example, for users with high blood sugar, the system may recommend increasing dietary fiber intake and choosing low GI (glycemic index) foods, such as whole grains, beans, and non-starchy vegetables, to help stabilize blood sugar levels.

[0027] Blood lipids include components such as cholesterol and triglycerides. Excessive blood lipid levels may lead to atherosclerosis, which in turn may cause problems such as heart disease and stroke. Understanding the user's blood lipid status is particularly critical for formulating a preventive diet plan. For example, the system may recommend foods rich in unsaturated fatty acids, such as fish, nuts, and olive oil, and limit the intake of saturated fats and trans fats, such as animal fats, butter, and oils in some processed foods. In addition, increasing the intake of dietary fiber can also help lower blood cholesterol levels. By analyzing the user's blood lipid data, the system can provide specific dietary guidance, aiming to improve blood lipid status and reduce the risk of cardiovascular disease by changing the diet structure.

[0028] In the aforementioned AI-based personalized healthy diet plan generation system, the user diet record acquisition module 120 is used to obtain the user's recent diet records, which include the types of food consumed daily, the amounts consumed, and the cooking methods. It should be understood that diet records reflect the user's actual intake and can identify nutritional gaps or excesses. For example, a user may have a preference for staple foods, resulting in a high proportion of carbohydrates, or a chronic lack of dietary fiber. By recording the types of food consumed, the amounts consumed, and the cooking methods (e.g., deep-frying vs. steaming), it helps to determine whether the user's current nutritional status is balanced, and which nutrients may be excessive or insufficient. For example, a user who regularly enjoys fast food may have excessive intake of fat and salt, and insufficient intake of vegetables and fruits.

[0029] Specifically, consider adopting multiple data input methods to meet the needs and preferences of different users. On the one hand, using mobile applications as the main tool, users can manually enter the food they consume daily, the corresponding portion size, and the cooking method. This direct input method can include filling in the food name in the text box, selecting the food category, using the slider to adjust the intake size, and selecting the cooking method from the preset options (such as steaming, boiling, stir-frying, deep-frying). To improve the user experience and reduce misoperation, the application can also provide an auto-completion function. When the user starts typing the food name, the system recommends possible food options based on the existing database. This not only speeds up the input speed but also ensures the accuracy of the input.

[0030] On the other hand, considering that some users may find manual input cumbersome, image recognition technology is introduced as an auxiliary means to record dietary information. Users only need to take a photo of their meal and upload it to the system. Deep learning algorithms analyze the image content, identify the food types contained in it, and estimate the approximate intake and cooking method. This requires the system to have powerful image processing capabilities, including but not limited to object detection, classification, and segmentation technologies to accurately extract food information from complex backgrounds.

[0031] In addition to the two primary data collection methods mentioned above, we can also explore the possibility of integrating with other devices and services. For example, connecting with smart kitchen appliances to directly obtain cooking parameters and ingredient quantities from these devices; or integrating data from online food ordering platforms to understand users' dining choices. These diverse data sources can not only more comprehensively capture users' eating habits, but also provide a richer foundation for subsequent nutritional analysis.

[0032] During the specific implementation process, in order to ensure the authenticity and completeness of the diet records, an incentive mechanism can also be established to encourage users to submit data continuously and accurately. For example, a points reward system can be established, and users can obtain corresponding points for each successful diet record. After accumulating a certain number of points, they can redeem small gifts or enjoy certain privileged services. At the same time, the system should regularly provide feedback reports to users, showing the preliminary nutritional analysis results based on existing diet records, to help users realize potential problems in their diet, such as excessive carbohydrates and insufficient dietary fiber intake, so as to motivate users to actively improve their eating habits. Through such interactive methods, it not only enhances the user's sense of participation, but also promotes the effective operation of the entire system.

[0033] Developing differentiated dietary recording strategies tailored to the characteristics of different user groups is also key to improving data quality. For example, for the elderly, traditional manual input methods may be difficult due to declining eyesight or a lack of acceptance of new technologies. In this case, considering developing voice input functions, allowing users to easily record what they ate each day, how much they used, and how they cooked.

[0034] In the above-mentioned personalized healthy diet plan generation system based on artificial intelligence, the nutrient component analysis module 130 is used to perform nutrient component analysis on the diet record based on the food nutrient component knowledge base to obtain a time series of the user's daily nutrient component intake data. It should be understood that converting the user's diet record into intake nutrient component data is a prerequisite for predicting the user's nutritional needs. The food nutrient component knowledge base contains nutrient component information of various types of food, such as protein, fat, carbohydrates, vitamins and minerals. By parsing the food types, intake and cooking methods in the user's diet record and combining it with the food nutrient component knowledge base, the various nutrient component data of the user's daily intake can be queried, which helps to analyze the user's nutritional intake habits, identify existing nutritional gaps or excess problems, and provide scientific guidance for the subsequent formulation of personalized diet plans.

[0035] In the above-mentioned personalized healthy diet plan generation system based on artificial intelligence, the nutrient requirement prediction module 140 is used to perform nutrient requirement prediction based on deep learning on the time series of the personal health data and the user's daily nutrient intake data to obtain the user's daily nutrient data. Specifically, since the traditional nutrient requirement formula (such as daily protein requirement = body weight × 1.2g / kg) fails to fully consider individual differences and user dietary conditions, this application introduces a deep learning algorithm to jointly analyze the user's personal health data and daily nutrient intake data to predict the various nutrients required by the user daily, such as protein, fat, carbohydrates, vitamins and minerals. Among them, Figure 3 This is a block diagram of a nutrient requirement prediction module in an artificial intelligence-based personalized healthy diet plan generation system according to an embodiment of the present application. Figure 3 As shown, the nutrient requirement prediction module 140 includes: a nutrient intake time series analysis unit 141, which is used to perform time series analysis on the time series of the user's daily nutrient intake data to obtain a user nutrient intake time series pattern feature coding vector; an embedded coding unit 142, which is used to perform embedded processing on the personal health data to obtain a user health feature coding vector; a feature interaction unit 143, which is used to perform implicit query feature interaction based on sparse constraints on the user health feature coding vector and the user nutrient intake time series pattern feature coding vector to obtain a user nutrient requirement feature coding vector; a nutrient requirement prediction unit 144, which is used to perform nutrient requirement decoding and prediction on the user nutrient requirement feature coding vector to obtain the nutrient data required by the user daily.

[0036] Specifically, the nutrition intake time series analysis unit 141 is used to perform time series analysis on the time series of the user's daily nutrition intake data to obtain a characteristic coding vector of the user's nutrition intake time series pattern. Figure 4 This is a block diagram of a nutrient intake time series analysis unit in a personalized healthy diet plan generation system based on artificial intelligence according to an embodiment of the present application. Figure 4 As shown, the nutrition intake time series analysis unit 141 includes: a semantic embedding coding subunit 1411, which is used to perform semantic embedding coding based on the Bert model on the time series of the user's daily nutrition intake data to obtain a time series of the user's daily nutrition intake feature embedding coding vector; a time series pattern feature extraction subunit 1412, which is used to perform time series feature extraction based on the LSTM model on the time series of the user's daily nutrition intake feature embedding coding vector to obtain the user's nutrition intake time series pattern feature coding vector.

[0037] More specifically, the semantic embedding coding subunit 1411 is used to perform semantic embedding coding based on the Bert model on the time series of the user's daily nutritional intake data to obtain a time series of the user's daily nutritional intake feature embedding coding vector. Specifically, considering that traditional nutritional intake data exists in the form of structured numerical values ​​(such as "75g protein, 30g fat"), but the potential associations between numerical values ​​(such as "vitamin D deficiency may affect calcium absorption") need to be expressed through semantic-level features in order to be effectively learned by the model, and directly using the original numerical values ​​will ignore the synergistic or antagonistic effects between nutrients. Therefore, the present application further uses the Bert model to perform semantic embedding coding on the user's daily nutritional intake data to capture the potential associations between the various nutritional data and obtain a time series of the user's daily nutritional intake feature embedding coding vector. Specifically, the Bert model is based on the Transformer architecture and uses a bidirectional Transformer encoder to process the input daily nutritional data of the user. It regards the numerical value of each nutrient and its related attributes as a "word". The language representation ability learned by the Bert model on a large-scale corpus is used to map these "words" into a low-dimensional vector space, and the bidirectional Transformer structure is used to capture the contextual semantic relationship between the embedding vectors of each "word". The semantic association between each nutritional data is learned, thereby generating a feature embedding coding vector of the user's daily nutritional intake.

[0038] More specifically, the temporal pattern feature extraction subunit 1412 is configured to perform temporal feature extraction based on an LSTM model on the time series of the user's daily nutrient intake feature embedding encoding vector to obtain the user's nutrient intake temporal pattern feature encoding vector. It should be understood that, given the significant time-dependence of a user's nutrient intake, for example, a high-calorie diet for three consecutive days may trigger metabolic compensation mechanisms (e.g., a temporary increase in basal metabolic rate), while periodic insufficient intake (e.g., low-carb diet on weekdays and high-carb diet on weekends) may lead to cumulative nutritional imbalances and affect health. Therefore, to capture the dynamic characteristics of a user's nutrient intake over time, this application introduces a long short-term memory (LSTM) model to perform temporal feature extraction on the time series of the user's daily nutrient intake feature embedding encoding vector. Specifically, the LSTM model, as a special recurrent neural network (RNN), effectively handles long-term dependencies in sequence data through the introduction of a gating mechanism, capturing hidden patterns and trends in time series data. For example, it can detect a surge in saturated fat intake due to social activities at the end of each month, thereby preemptively increasing the dietary fiber recommendation to balance the situation in the prediction. In this application, the LSTM model iteratively processes the time series of the user's daily nutrient intake feature embedding coding vector, takes the nutrient feature embedding coding vector of each time step as input, and updates and transmits the cell state through the collaborative work of three key components: the forget gate, the input gate, and the output gate, so as to capture the dynamic characteristics and periodic patterns of the user's nutrient intake over time, thereby learning the temporal laws of the user's nutrient intake and generating a temporal pattern feature coding vector of the user's nutrient intake.

[0039] Specifically, in a specific example of the present application, the embedded coding unit 142 is used to: perform embedding processing on the personal health data based on a multi-layer perceptron to obtain the user health feature encoding vector. Specifically, since user health data (such as height, weight, age, BMI, and blood glucose level) is static or low-frequency updated structured data, there are modal differences with high-frequency nutritional time series data. Therefore, in order to effectively jointly analyze the two types of data, the present application further uses a multi-layer perceptron (MLP) to embed the personal health data to map the user health data into a low-dimensional vector space compatible with the nutrient intake time series pattern feature encoding vector. Specifically, the personal health data contains a variety of different types of information such as gender, body fat percentage, blood pressure, blood glucose, and blood lipids. The multi-layer perceptron can effectively realize high-order feature combination through fully connected layers and nonlinear activation functions (such as ReLU), capturing the nonlinear relationship between various health indicators. For example, the interaction between age and body fat percentage may imply the risk of metabolic syndrome, thereby affecting nutritional needs. The MLP model extracts and combines features from the individual health data through multiple fully connected layers. Each layer uses nonlinear activation functions (such as ReLU) to enhance the model's nonlinear expressiveness. This multi-layered processing maps the original heterogeneous health parameters into dense vectors in a high-dimensional feature space while preserving the nonlinear relationships between key physiological indicators (such as the negative correlation between body fat percentage and basal metabolic rate). This results in a user health feature encoding vector that fully expresses the user's health status, providing effective health status information for subsequent nutritional needs prediction.

[0040] Specifically, the feature interaction unit 143 is used to perform implicit query feature interaction based on sparse constraints on the user health feature coding vector and the user nutrient intake time series pattern feature coding vector to obtain the user nutritional demand feature coding vector. It should be understood that the user health feature coding vector reflects the user's health status information, while the user nutrient intake time series pattern feature coding vector captures the dynamic characteristics of the user's nutrient intake. In order to more accurately predict the user's nutritional needs, it is necessary to further effectively fuse the two to more comprehensively reflect the user's nutritional demand status. In particular, considering that directly performing feature splicing and fusion on the user health feature coding vector and the user nutrient intake time series pattern feature coding vector will lead to noise interference between irrelevant features (such as the accidental correlation between the user's height and vitamin C intake last week). In this regard, this application proposes a latent query feature interaction method based on sparse constraints, which optimizes the feature interaction process by constructing the interactive relationship between user health features and user nutrient intake time series features in the latent query space, and introducing sparse constraints to optimize the feature interaction process, retaining only key feature interactions, such as "recent high sodium intake + history of hypertension" needs to trigger low sodium recommendations, while ignoring irrelevant combinations (such as "vitamin C intake + body fat percentage"), thereby improving the accuracy and efficiency of feature interaction, reducing noise interference, and obtaining a more representative user nutritional demand feature encoding vector. Among them, Figure 5 FIG. 1 is a block diagram of a feature interaction unit in a personalized healthy diet plan generation system based on artificial intelligence according to an embodiment of the present application. Figure 5 As shown, the feature interaction unit 143 includes: a local feature interaction sub-unit 1431, which is used to perform feature interaction encoding on the user's nutrient intake time series pattern feature coding vector and the user's health feature coding vector based on a local implicit query space to obtain a set of user nutrient intake-health status implicit query space joint interaction coding matrices; a sparse constraint aggregation sub-unit 1432, which is used to perform sparse constraint-based adaptive aggregation on the set of user nutrient intake-health status implicit query space joint interaction coding matrices to obtain the user's nutritional demand feature coding vector.

[0041] More specifically, in a specific example of the present application, the local feature interaction subunit 1431 is used to perform feature phase space reconstruction based on one-dimensional convolution coding on the user's nutrient intake time series pattern feature coding vector and the user's health feature coding vector to obtain a set of user nutrient intake local time series pattern feature coding vectors and a set of user health status local feature coding vectors, which can be expressed as follows: in, represents the characteristic coding vector of the user's nutrient intake temporal pattern, represents the user health feature encoding vector, is a one-dimensional convolutional coding network, A set of feature encoding vectors representing the local temporal pattern of the user’s nutrient intake, 、 、 and Respectively represent the first, second, and third in the set of characteristic coding vectors of the local temporal pattern of the user's nutrient intake and The feature encoding vector of the local temporal pattern of nutrient intake of each user, The number of vectors in the set of the local temporal pattern feature encoding vectors of the user's nutrient intake, Represents the set of local feature encoding vectors of the user's health status, 、 、 and Respectively represent the first, second, and third in the set of local feature coding vectors of the user's health status and The local feature encoding vector of a user's health status.

[0042] That is, by introducing the phase space reconstruction technology of one-dimensional convolutional coding, the user's nutrient intake time series pattern feature coding vector and the user's health feature coding vector are abstracted into potential trajectories, and the sliding window mechanism is used to extract local feature fragments, which not only retains the dynamic dependency relationship between continuous time steps in the time series data, but also explores the nonlinear coupling effect of health features and nutrient intake in local time and space, constructs a more discriminative set of user nutrient intake local time series pattern feature coding vectors and a set of user health status local feature coding vectors, and enhances the model's robustness to noise through local feature redundancy. At the same time, it breaks through the limitations of traditional linear fusion, enabling the model to identify the fine-grained interaction pattern between user eating behavior and health status, significantly improving the accuracy of nutritional demand prediction.

[0043] More specifically, in a specific example of the present application, the local feature interaction subunit 1431 is further used to calculate the implicit query space joint coding matrix between any set of user nutrient intake local time series pattern feature coding vectors and user health status local feature coding vectors in the set of the user nutrient intake local time series pattern feature coding vectors and the set of the user health status local feature coding vectors to obtain the set of the user nutrient intake-health status implicit query space joint interaction coding matrices, which is expressed as follows: in, represents the transpose of the matrix, represents the matrix multiplication operation, and Represent the user's nutrient intake feature weight matrix and the user's health status feature weight matrix respectively, is the characteristic scale scaling factor, express and Joint interaction encoding matrix between user nutrient intake and health status implicit query space.

[0044] That is, by mapping the local temporal pattern feature coding vector of the user's nutrient intake and the local feature coding vector of the user's health status to the same implicit query space, and using the implicit query space joint coding matrix to calculate their nonlinear interaction relationship, it is possible to break through the limitations of the traditional fixed interaction space, enabling the model to automatically mine the semantic associations between cross-dimensional features, build a semantic interaction bridge, deeply integrate discrete temporal nutrient intake data with continuous health status information, screen key interaction patterns through sparse constraints, and generate the user nutrient intake-health status implicit query space joint interaction coding matrix.

[0045] Figure 6 FIG. 1 is a block diagram of a sparse constraint aggregation subunit in a personalized healthy diet plan generation system based on artificial intelligence according to an embodiment of the present application. Figure 6 As shown, the sparse constraint aggregation subunit 1432 includes: an interactive connection redundancy reduction secondary subunit 14321, which is used to perform interactive connection redundancy reduction on each user nutrient intake-health status implicit query space joint interactive coding matrix in the set of user nutrient intake-health status implicit query space joint interactive coding matrices to obtain a set of optimized user nutrient intake-health status implicit query space joint interactive coding matrices; a sparse constraint factor calculation secondary subunit 14322, which is used to calculate the sparse constraint factors of each optimized user nutrient intake-health status implicit query space joint interactive coding matrix in the set of optimized user nutrient intake-health status implicit query space joint interactive coding matrices to obtain a set of user nutrient intake-health status implicit interaction space sparse constraint factors; an adaptive aggregation coding secondary subunit 14323, which is used to perform adaptive aggregation coding on the set of optimized user nutrient intake-health status implicit query space joint interactive coding matrices based on the set of user nutrient intake-health status implicit interaction space sparse constraint factors to obtain the user nutritional demand feature coding vector.

[0046] In particular, in a preferred example of the present application, the interactive connection redundancy reduction secondary subunit 14321 is expressed as follows: in, for The eigenvalues, for The eigenvalues, Represents the exponential function operation with e as the base, represents the user's nutrient intake-health status joint interaction phase matrix, Represents the user's nutrient intake-health status joint interaction phase matrix The eigenvalues ​​of the position, represents the inverse matrix, Represents the optimized joint interaction coding matrix of the user's nutrient intake-health status implicit query space.

[0047] Specifically, by introducing a phase matrix and its inverse matrix to optimize the mapping of the joint interaction encoding matrix of the user's nutrient intake and health status implicit query space while preserving the norm, the model can interpret the phase difference between the periodic physiological rhythms inherent in user data and discrete dietary events, and reconstruct the connection weights of the joint interaction encoding matrix of the user's nutrient intake and health status implicit query space accordingly. This process transforms feature associations, originally achieved through multiple nonlinear interactions, into linear combination representations based on flat manifolds. Essentially, it uses the trivial connection mechanism in algebraic topology to isometrically map the global structure of local nonlinear interactions onto a low-dimensional manifold, thereby eliminating the cross-resonance effects caused by different user groups during the phase space reconstruction process. In terms of computational efficiency, redundancy reduction reduces the computational complexity of feature interactions from exponential to polynomial levels. In terms of model robustness, by filtering out abnormal communication paths caused by noisy data or abnormal dietary records, the risk of overfitting caused by excessive parameterization is avoided. In terms of application value, the optimized user nutrient intake-health status implicit query space joint interaction coding matrix can more accurately capture the dynamic balance point between the user's health status and dietary needs, thereby realizing continuous tracking and predictive intervention of the user's health trajectory.

[0048] In a specific example of the present application, the sparsity constraint factor calculation secondary subunit 14322 is expressed as follows: in, is the feature association strength measurement function, It means to calculate the square of the Frobenius norm of the matrix. express The corresponding sparse constraint factor of the implicit interaction space between user nutrient intake and health status.

[0049] Specifically, by calculating the sparsity constraint factor, the model can quantify the sparsity of the element distribution in each optimized user's nutrient intake-health status implicit query space joint interaction encoding matrix, identifying the implicit query relationships that play a decisive role in nutritional needs prediction. This allows the model to prioritize retaining highly discriminative feature interaction patterns in the subsequent aggregation process while suppressing redundant connections with low responsiveness. Thus, the generated sparse constraint factor for the user's nutrient intake-health status implicit interaction space helps compress invalid interaction dimensions in the subsequent local interaction feature fusion process, significantly reducing the model parameter size.

[0050] In a specific example of the present application, the adaptive aggregate coding secondary subunit 14323 is used to perform a normalization process based on the softmax function on the set of sparse constraint factors of the user's nutrient intake-health status implicit interaction space to obtain a set of normalized sparse constraint factors of the user's nutrient intake-health status implicit interaction space, which is expressed as follows: in, represents the normalized exponential function, express The corresponding normalized sparse constraint factor of the implicit interaction space between user nutrient intake and health status.

[0051] That is, by introducing softmax normalization, the discrete sparse constraint values ​​are mapped into a probability distribution form, making the contribution of different interaction dimensions comparable. At the same time, the dominance of high-responsiveness interaction patterns is strengthened, and the adaptive calibration of feature interaction weights is achieved. The driving effect of significant interactions on the final prediction is amplified by exponential probability, and the interference of low-responsiveness features is suppressed, thereby improving the model's accuracy in capturing the dynamic relationship between individualized nutritional needs and health status.

[0052] In a specific example of the present application, the adaptive aggregation coding secondary subunit 14323 is further used to: use the set of sparse constraint factors of the normalized user nutrient intake-health status implicit interaction space as the weight distribution, and perform weighted aggregation on the set of the optimized user nutrient intake-health status implicit query space joint interaction coding matrices to obtain the user nutrient intake-health status implicit interaction feature sparsity fusion matrix, which is expressed as follows: in, Represents the sparse fusion matrix of implicit interaction features between user nutrient intake and health status.

[0053] Specifically, by introducing a sparse constraint factor in the normalized user nutrient intake-health status implicit interaction space as a dynamic weight, an adaptive aggregation mechanism based on interaction importance was constructed, which enabled highly significant interactions to dominate the feature space while simultaneously suppressing the noise amplification effect of low-responsive interactions through weight decay. This weighted fusion approach achieved secondary optimization of the feature space, guiding the model to focus on the most explanatory interaction pathways, significantly improving the model's sensitivity to individualized nutritional imbalance patterns and amplifying weak but stable correlation signals.

[0054] In a specific example of the present application, the adaptive aggregate coding secondary subunit 14323 is further used to reshape the user's nutrient intake-health status implicit interaction feature sparsity fusion matrix to obtain the user's nutritional demand feature coding vector, which is expressed as follows: in, represents the feature shape reshaping, Represents the user's nutritional demand feature encoding vector.

[0055] In other words, by reshaping the features, a mapping transformation from a complex feature space to a structured encoding vector is achieved, preserving the topological relationships of key interaction patterns while satisfying the format requirements of subsequent decoding units for fixed-dimensional input. This structured transformation constructs a semantic compression channel in the feature space, condensing the nutritional gap association patterns, temporal fluctuation patterns, and health status response mechanisms scattered across the multidimensional matrix into a more explanatory feature encoding vector for user nutritional needs.

[0056] Specifically, the nutrient requirement prediction unit 144 is used to perform nutrient requirement decoding prediction on the user nutrient requirement feature coding vector to obtain the nutrient component data required by the user daily. Specifically, in order to convert the user nutrient requirement feature coding vector into specific nutrient component data required by the user daily, the present application adopts a decoder model to decode the user nutrient requirement feature coding vector. In an embodiment of the present application, the decoder adopts a fully connected neural network architecture, takes the user nutrient requirement feature coding vector as input, learns the complex mapping relationship between the user's health status, nutrient intake and nutrient requirement through multiple layers of fully connected layers, gradually maps the nutrient requirement feature coding vector back to the nutrient component data space, and outputs a detailed list containing multiple nutrients and their corresponding requirements, which serves as a guide for the user's daily nutrient intake, helping them improve their health and prevent the occurrence of diseases.

[0057] In the above-mentioned personalized healthy diet plan generation system based on artificial intelligence, the user diet preference acquisition module 150 is used to obtain the diet preference information input by the user. It should be understood that this application takes into account that even if the diet plan formulated is scientific and reasonable at the nutritional level, it is difficult for the user to adhere to it for a long time if it does not conform to the user's taste and eating habits. Therefore, this application further obtains the user's diet preference information to ensure that the user's favorite food and cooking methods are incorporated under the premise of meeting nutritional needs, thereby improving the acceptability and compliance of the diet plan and ensuring that the user can follow the diet plan for a long time to achieve the purpose of improving health.

[0058] In the aforementioned AI-based personalized healthy diet plan generation system, the diet plan generation module 160 is configured to generate a healthy diet plan based on the dietary preference information and the user's daily nutritional data. In a specific example of the present application, the diet plan generation module 160 is configured to input the dietary preference information and the user's daily nutritional data into a personalized diet plan generator based on a large language model to generate a healthy diet plan, which includes food types, intake amounts, and cooking methods. It should be understood that large language models, trained on large amounts of text data, possess powerful text generation capabilities and can generate logically clear, natural, and compliant text based on given information. In the present application, the dietary preference information and the user's daily nutritional data are organized into structured text according to a preset format and input into a large language model (e.g., the GPT series). Leveraging the language patterns and knowledge learned during training, a specific healthy diet plan including food types, intake amounts, and cooking methods can be quickly and efficiently generated, providing users with intuitive and actionable dietary guidance, helping them plan their diet rationally and achieve nutritional balance and health goals.

[0059] In summary, the artificial intelligence-based personalized healthy diet plan generation system based on the embodiment of the present application is explained, which is based on the food nutrient knowledge base, determines the user's daily nutrient intake data based on the user's recent diet records, and combines the personal health data input by the user. By further introducing a deep learning algorithm, the user's personal health data and daily nutrient intake data are jointly interactively analyzed to predict the user's daily nutrient requirements. Then, based on the obtained user's daily nutrient requirement data, combined with the user's dietary preference information, a large language model is used to generate a personalized healthy diet plan that meets the user's dietary preferences and nutritional needs. In this way, the problem of scarce nutritionist resources can be effectively solved, and personalized customization of diet plans can be achieved, thereby providing efficient and accurate healthy diet recommendations for large groups of people.

[0060] The basic principles of the present invention have been described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, and effects mentioned in the present invention are merely illustrative and non-limiting, and should not be construed as necessarily possessed by each embodiment of the present invention. Furthermore, the specific details of the above embodiments are provided for illustrative purposes and to facilitate understanding, and are not intended to be limiting. These details do not necessarily limit the present invention to being implemented using these specific details.

[0061] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, please refer to the relevant description of other embodiments. In the several embodiments provided by the present invention, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiment described above is only schematic. For example, the unit division is only a logical function division, and there may be other division methods in actual implementation. The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the scheme of this embodiment.

[0062] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be encompassed therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0063] In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units stated in the system claims can also be implemented by one unit through software or hardware.

[0064] Finally, it should be noted that the above description has been provided for purposes of illustration and description. Furthermore, the above embodiments are intended only to illustrate the technical solutions of the present invention and are not intended to be limiting. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art will appreciate that the technical solutions of the present invention may be modified or replaced with equivalents without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A personalized healthy diet plan generation system based on artificial intelligence, characterized by: include: A user health data acquisition module is used to receive personal health data input by the user; A user diet record acquisition module is used to acquire the user's recent diet records, which include the types of food consumed daily, the amount consumed, and the cooking method; A nutrient component analysis module, configured to perform nutrient component analysis on the diet record based on a food nutrient component knowledge base to obtain a time series of the user's daily nutrient component intake data; A nutrient requirement prediction module, configured to perform nutrient requirement prediction based on deep learning on the time series of the personal health data and the user's daily nutrient intake data to obtain the nutrient requirement data required by the user daily; User dietary preference acquisition module, used to obtain dietary preference information input by the user; A diet plan generating module is used to generate a healthy diet plan based on the diet preference information and the nutritional component data required by the user daily.

2. The artificial intelligence-based personalized healthy diet plan generation system according to claim 1, characterized in that: The personal health data includes height, weight, age, gender, body fat percentage, blood pressure, blood sugar, and blood lipids.

3. The artificial intelligence-based personalized healthy diet plan generation system according to claim 2, characterized in that: The nutrient requirement prediction module includes: a nutrient intake time series analysis unit, configured to perform time series analysis on the time series of the user's daily nutrient intake data to obtain a characteristic coding vector of the user's nutrient intake time series pattern; An embedding coding unit, configured to perform embedding processing on the personal health data to obtain a user health feature coding vector; a feature interaction unit, configured to perform implicit query feature interaction based on sparse constraints on the user health feature coding vector and the user nutrient component intake time series pattern feature coding vector to obtain the user nutritional demand feature coding vector; The nutrient requirement prediction unit is used to perform nutrient requirement decoding prediction on the user's nutrient requirement feature coding vector to obtain the nutrient requirement data required by the user on a daily basis.

4. The artificial intelligence-based personalized healthy diet plan generation system according to claim 3, characterized in that: The nutrition intake time series analysis unit comprises: A semantic embedding coding subunit, configured to perform semantic embedding coding based on a Bert model on the time series of the user's daily nutrient intake data to obtain a time series of the user's daily nutrient intake feature embedding coding vector; The time series pattern feature extraction subunit is used to perform time series feature extraction based on the LSTM model on the time series of the user's daily nutrient component intake feature embedded coding vector to obtain the user's nutrient component intake time series pattern feature coding vector.

5. The artificial intelligence-based personalized healthy diet plan generation system according to claim 4, characterized in that: The embedded coding unit is used to: The personal health data is embedded in a multi-layer perceptron to obtain the user health feature encoding vector.

6. The artificial intelligence-based personalized healthy diet plan generation system according to claim 5, characterized in that: The feature interaction unit includes: a local feature interaction subunit, configured to perform feature interaction coding on the user's nutrient intake temporal pattern feature coding vector and the user's health feature coding vector based on a local implicit query space to obtain a set of user nutrient intake-health status implicit query space joint interaction coding matrices; The sparse constraint aggregation subunit is used to perform sparse constraint-based adaptive aggregation on the set of user nutrient intake-health status implicit query space joint interaction coding matrices to obtain the user nutritional demand feature coding vector.

7. The artificial intelligence-based personalized healthy diet plan generation system according to claim 6, characterized in that: The local feature interaction subunit is used to: Performing feature phase space reconstruction based on one-dimensional convolution coding on the user's nutrient intake time series pattern feature coding vector and the user's health feature coding vector to obtain a set of user's nutrient intake local time series pattern feature coding vectors and a set of user's health status local feature coding vectors; Calculate the implicit query space joint coding matrix between any set of user nutrient intake local time series pattern feature coding vectors and user health status local feature coding vectors in the set of user nutrient intake local time series pattern feature coding vectors and the set of user health status local feature coding vectors to obtain the set of user nutrient intake-health status implicit query space joint interaction coding matrices.

8. The artificial intelligence-based personalized healthy diet plan generation system according to claim 7, characterized in that: The sparse constraint aggregation subunit includes: an interactive connection redundancy reduction secondary subunit, configured to perform interactive connection redundancy reduction on each user nutrient intake-health status implicit query space joint interactive coding matrix in the set of user nutrient intake-health status implicit query space joint interactive coding matrices to obtain an optimized set of user nutrient intake-health status implicit query space joint interactive coding matrices; A sparse constraint factor calculation secondary subunit is used to calculate the sparse constraint factor of each optimized user nutrient intake-health status implicit query space joint interaction coding matrix in the set of optimized user nutrient intake-health status implicit query space joint interaction coding matrices to obtain a set of user nutrient intake-health status implicit interaction space sparse constraint factors; The adaptive aggregation coding secondary sub-unit is used to perform adaptive aggregation coding on the set of optimized user nutrient intake-health status implicit query space joint interaction coding matrices based on the set of sparse constraint factors of the user nutrient intake-health status implicit interaction space to obtain the user nutritional demand feature coding vector.

9. The artificial intelligence-based personalized healthy diet plan generation system according to claim 8, characterized in that: The adaptive aggregation coding secondary sub-unit is configured to: performing a normalization process based on a softmax function on the set of sparse constraint factors of the user's nutrient intake-health status implicit interaction space to obtain a set of normalized sparse constraint factors of the user's nutrient intake-health status implicit interaction space; Using the set of normalized user nutrient intake-health status implicit interaction space sparse constraint factors as weight distribution, weighted aggregation is performed on the set of optimized user nutrient intake-health status implicit query space joint interaction encoding matrices to obtain a user nutrient intake-health status implicit interaction feature sparsity fusion matrix; The user's nutritional component intake-health status implicit interaction feature sparsity fusion matrix is ​​reshaped to obtain the user's nutritional demand feature encoding vector.

10. The artificial intelligence-based personalized healthy diet plan generation system according to claim 9, characterized in that: The diet plan generating module is used to: The dietary preference information and the user's daily nutritional data are input into a personalized diet plan generator based on a large language model to obtain a healthy diet plan, which includes food types, intake amounts, and cooking methods.

Images