Nutrition health analysis and meal suggestion generation method and device based on large model

Abstract

The embodiment of the specification provides a kind of based on AI's nutritional health analysis and meal suggestion generation method and device, wherein, method includes: by Raman spectrum data and nutrition index historical data set, automatically build and optimize nutrient risk assessment model on visual machine learning platform, predict the nutrient state information of user based on new monitored Raman spectrum data by the nutrient risk assessment model;With large language model as engine, based on user basic information, historical dietary data and health monitoring data, using search enhancement generation technology dynamically searches authoritative knowledge base, generates nutritional health analysis report, and generates personalized meal suggestion according to the nutritional health analysis report, wherein, based on the health monitoring data includes: nutrient state information.

Description

Technical Field

[0001] This document relates to the field of health management technology, and in particular to a method and apparatus for nutritional health analysis and dietary recommendation generation based on a large model. Background Technology

[0002] In existing technologies, nutritional assessment methods mainly rely on blood biochemistry tests, dietary questionnaires, and doctors' experience. These technologies have the following problems: high testing costs and long cycles, making it difficult to achieve high-frequency monitoring; nutrient deficiency risk assessment relies on manual interpretation, which is highly subjective; dietary recommendations are mostly static templates, lacking individualization and dynamic adjustment capabilities; and they cannot effectively utilize Raman spectroscopy, a non-invasive, rapid, and low-cost molecular fingerprint data.

[0003] In recent years, machine learning has made breakthroughs in health data analysis, with large models performing well in knowledge retrieval and reasoning. However, there is still no system that deeply integrates the two for nutrient risk assessment and dietary recommendation generation. Summary of the Invention

[0004] The purpose of this invention is to provide a method and apparatus for nutritional health analysis and dietary recommendation generation based on a large model, in order to solve the above-mentioned problems in the prior art.

[0005] This invention provides a method for nutritional health analysis and dietary recommendation generation based on a large model, comprising: By using Raman spectroscopy data and historical datasets of nutritional indicators, a nutrient risk assessment model is automatically built and optimized on a visualization machine learning platform. Based on newly monitored Raman spectroscopy data, the nutrient risk assessment model predicts the user's nutrient status information. Using a large language model as the engine, based on user basic information, historical dietary data and health monitoring data, the system dynamically retrieves authoritative knowledge bases using retrieval enhancement generation technology, generates a nutrition and health analysis report, and generates personalized dietary recommendations based on the nutrition and health analysis report. The health monitoring data includes nutrient status information.

[0006] This invention provides a device for nutritional health analysis and dietary recommendation generation based on a large model, comprising: A prediction module is constructed to automatically build and optimize a nutrient risk assessment model on a visualization machine learning platform using Raman spectroscopy data and historical datasets of nutrient indicators. Based on newly monitored Raman spectroscopy data, the nutrient risk assessment model predicts the user's nutrient status information. The generation module is used to dynamically retrieve authoritative knowledge bases based on user basic information, historical dietary data, and health monitoring data using a large language model as the engine, and to generate a nutrition and health analysis report. Based on the nutrition and health analysis report, personalized dietary recommendations are generated. The health monitoring data includes nutrient status information.

[0007] This invention also provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, it implements the steps of the above-described method for generating nutritional health analysis and dietary recommendations based on a large model.

[0008] This invention also provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor, implements the steps of the above-described method for generating nutritional health analysis and dietary recommendations based on a large model.

[0009] The embodiments of the present invention solve the problems of strong subjectivity and template-based recommendations in traditional nutritional assessments, providing customers with accurate nutritional assessments and dietary recommendations, and can be widely applied in hospital nutrition departments, physical examination centers, maternal and infant institutions, and family health management scenarios. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart of a method for generating nutritional health analysis and dietary recommendations based on a large model, according to an embodiment of the present invention. Figure 2 This is a detailed flowchart of the method for generating nutritional health analysis and dietary recommendations based on a large model according to an embodiment of the present invention; Figure 3 This is a schematic diagram illustrating the principle of the nutritional health analysis and dietary recommendation generation method based on a large model according to an embodiment of the present invention; Figure 4 This is a schematic diagram of a large-scale model-based nutritional health analysis and dietary recommendation generation device according to an embodiment of the present invention; Figure 5 This is a schematic diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0012] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0013] Method Implementation Examples According to embodiments of the present invention, a method for nutritional health analysis and dietary recommendation generation based on a large model is provided. Figure 1 This is a flowchart of a method for generating nutritional health analysis and dietary recommendations based on a large model, as described in an embodiment of the present invention. Figure 1 As shown, the method for generating nutritional health analysis and dietary recommendations based on a large model according to an embodiment of the present invention specifically includes: Step S101 involves automatically constructing and optimizing a nutrient risk assessment model on a visualization machine learning platform using Raman spectroscopy data and historical nutrient index datasets. This nutrient risk assessment model then predicts the user's nutrient status based on newly monitored Raman spectroscopy data. Specifically, this includes: Raman spectral data covering specific wavelength band spectral intensity information are acquired using a portable or desktop Raman spectrometer; key nutritional status indicators are obtained through clinical biochemical testing, wherein the key nutritional status indicators specifically include: vitamin D, albumin, and / or ferritin; The algorithm integrates feature selection algorithms and / or classification / regression algorithms from the algorithm pool, and automatically optimizes the hyperparameters of the feature selection algorithms and / or classification / regression algorithms. It automatically builds and optimizes multiple nutrient risk assessment models for specific nutrients on a visual machine learning platform, and performs five-fold hierarchical cross-validation on the nutrient risk assessment models to select the optimal model as the final nutrient risk assessment model. The optimization methods specifically include Bayesian optimization and Hyperband early cessation strategy.

[0014] The newly monitored Raman spectral data undergoes wavelength calibration, baseline correction, smoothing, denoising, and normalization. The feature selection module corresponding to the final nutrient risk assessment model is then invoked to extract key spectral variables, which are input into the final nutrient risk assessment model. The model outputs predicted nutrient index values, maps these predicted values ​​to corresponding nutrient index levels, and outputs a structured report. The structured report specifically includes: predicted nutrient index values, nutrient index levels, confidence scores, and visualization charts.

[0015] Step S102: Using a large language model as the engine, based on user basic information, historical dietary data, and health monitoring data, a retrieval enhancement generation technology is used to dynamically retrieve authoritative knowledge bases, generate a nutrition and health analysis report, and generate personalized dietary recommendations based on the nutrition and health analysis report. The health monitoring data includes nutrient status information. Specifically, the nutrient status information includes nutrient index levels, which specifically include: normal, insufficient, deficient, and severely deficient.

[0016] As can be seen from the above technical solution, the embodiments of the present invention provide a method for nutritional health analysis and dietary recommendation generation based on machine learning and large models. First, by using Raman spectroscopy data and historical datasets of nutritional indicators, a nutrient risk assessment model is automatically constructed and optimized on a visual machine learning platform to achieve rapid and non-invasive prediction of nutrient status such as vitamin D, albumin, and ferritin. Then, using a large model as the engine, individual basic information, historical dietary data, and health monitoring data are integrated, and authoritative knowledge bases such as the "Chinese Dietary Guidelines (2022)" are invoked to generate personalized dietary recommendations.

[0017] The technical solutions described above in the embodiments of the present invention will be explained in detail below.

[0018] like Figure 2 and Figure 3 As shown, the nutritional health analysis and dietary recommendation generation method according to an embodiment of the present invention specifically includes the following processes: I. Nutrient Risk Assessment 1. Construct a nutrient risk assessment model.

[0019] A visual model for nutrient risk assessment is constructed. Based on the correlation between historical urine Raman spectroscopy data and nutritional indicators, various feature selection and classification / regression algorithms from the algorithm pool are integrated to construct multiple combinations of "feature selection-inference estimation" methods. By automatically optimizing the hyperparameters of each algorithm, a risk classification and assessment model for specific nutrients is established.

[0020] Nutrient risk assessment models can predict corresponding nutrient levels based on newly monitored spectral data and classify them according to nutrient standards as: normal, insufficient, deficient, or severely deficient.

[0021] 1) Dataset Input The input data required for model training consists of historical Raman spectroscopy datasets and corresponding nutrient index label data. Specifically: Raman spectral data: acquired using portable or benchtop Raman spectrometers, covering specific wavelength ranges (e.g., 400–1800 cm⁻¹). -1 The spectral intensity information of the spectrum contains hundreds to thousands of feature variables (wavelength points) for each spectrum.

[0022] Nutritional indicator labels: These include key nutritional status indicators such as vitamin D (serum 25(OH)D concentration, unit: nmol / L), albumin (ALB, unit: g / L), and ferritin (FER, unit: μg / L), all of which were obtained through clinical biochemical tests and used as labels for model training.

[0023] The dataset needs to be of a certain size (sample size ≥ 1000 is recommended) and cover people of different genders, ages and regions to ensure the generalization ability of the model.

[0024] 2) Built-in algorithm pool: The platform has a rich library of machine learning algorithms, which are divided into two main categories: feature selection algorithms and classification / regression algorithms. Users can freely combine them or use the system's automatic recommendation strategy.

[0025] (1) Feature selection algorithm pool: LASSO (Least Absolute Shrinkage and Selection Operator): A linear regression method based on L1 regularization that can achieve feature sparsity and automatically select important wavelength variables.

[0026] Recursive Feature Elimination (RFE): An iterative feature selection method that gradually eliminates unimportant features based on model weights. It is applicable to various base models (such as SVM and logistic regression).

[0027] LightGBM Feature Importance: Using the feature gain or split count metrics output after LightGBM model training, the Top-K important features are selected.

[0028] (2) Classification / Regression Algorithm Pool: XGBoost (eXtreme Gradient Boosting): An efficient ensemble learning algorithm based on gradient boosting decision trees, supporting regularization and automatic handling of missing values, suitable for small to medium-sized high-dimensional data.

[0029] Random Forest: An ensemble tree model based on the Bagging concept, possessing good resistance to overfitting and interpretability.

[0030] SVM (Support Vector Machine): An algorithm model based on the maximum margin principle, suitable for high-dimensional small sample scenarios, and supports various kernel functions such as RBF and polynomial.

[0031] LightGBM: Microsoft's open-source gradient boosting framework, which uses histogram algorithm and leaf-wise growth strategy, resulting in fast training speed and low memory usage.

[0032] CatBoost: A gradient boosting library developed by Yandex, specifically optimized for categorical features. It supports automatic handling of categorical variables and ordered boosting, reducing the risk of overfitting.

[0033] 3) Algorithm combination and hyperparameter optimization.

[0034] The system employs a two-stage combined strategy of "feature selection - inference estimation" to automatically generate multiple modeling paths. For example: LASSO + XGBoost: First, key spectral features are screened using LASSO, and then input into XGBoost for inference and estimation; RFE + CatBoost: The feature subset is gradually refined through RFE and then combined with CatBoost to handle potential nonlinear relationships.

[0035] To improve model performance, the system initiates an automated hyperparameter optimization process for each combination, employing the following strategies: Bayesian optimization: It builds surrogate models based on Gaussian processes (GP) or tree-structured Parzen estimators (TPE), and intelligently searches the hyperparameter space, which is more efficient than grid search and random search. Hyperband early stop strategy: Combines multi-fidelity evaluation mechanism to dynamically allocate computing resources and terminate poorly performing hyperparameter configurations in advance, significantly shortening the parameter tuning time.

[0036] The objective function is optimized by taking into account both the F1-score and AUC value on the validation set, ensuring that the model achieves a balance between precision and recall.

[0037] 4) Model evaluation and optimal model selection.

[0038] All model combinations underwent 5-fold stratified cross-validation to ensure a balanced distribution of each category (e.g., normal, insufficient, lacking, severely lacking) across the training and validation sets. Evaluation metrics included: F1-score: The harmonic mean of precision and recall, suitable for class imbalance scenarios; AUC (Area Under ROC Curve): Measures the model's ability to distinguish between positive and negative classes; the closer the value is to 1, the better the performance. Recall: Pay special attention to the ability to identify the "deficiency" and "severe deficiency" categories to avoid the risk of missing nutritional deficiencies.

[0039] The system outputs an evaluation report for each combination, including a confusion matrix, ROC curve, and feature importance ranking. Finally, the model with the highest overall score (such as weighted F1-score and AUC) is selected as the optimal model, and it is serialized and saved as a nutrient risk classification assessment model (such as .pkl or .onnx format) and deployed to the inference service module.

[0040] 2. Assess nutrient risk status.

[0041] Nutrient status classification: The nutritional index values ​​(such as vitamin D concentration) predicted by the model will be classified into the following four levels according to the authoritative standard of **Chinese Dietary Reference Intakes (DRIs 2023)**: Grading criteria (taking vitamin D as an example) and its health significance Normal ≥ 50 nmol / L, good nutritional status A level of less than 30–49 nmol / L indicates a mild deficiency risk, requiring monitoring of dietary intake. A deficiency of 12–29 nmol / L is significant; supplementation and intervention are recommended. Severe deficiency (<12 nmol / L) poses a high risk and requires medical intervention. The grading thresholds for different nutritional indicators are dynamically adjusted based on the reference values ​​for the corresponding age and gender groups in DRIs 2023, and the system supports configurable updates.

[0042] For newly acquired user Raman spectral data, perform the following procedure: Data preprocessing: wavelength calibration, baseline correction, smoothing and denoising, normalization; Feature extraction: Call the feature selection module corresponding to the optimal model to extract key spectral variables; Model Inference: Load the saved optimal inference estimation model and output the predicted values ​​of nutrient indicators; Risk grading: Map predicted values ​​to “normal / insufficient / deficient / severe deficiency” levels based on DRIs criteria; Output: Generates a structured report, including predicted values, risk levels, confidence scores, and visualizations.

[0043] II. Nutritional health analysis and dietary recommendation report generation.

[0044] 1. Nutritional and health analysis.

[0045] Using a Large Language Model (LLM) as its core engine and combining it with Retrieval-Augmented Generation (RAG) technology, the system dynamically retrieves authoritative knowledge bases (such as the full text of the "Chinese Dietary Guidelines (2022)" and national food safety standards) to provide real-time, accurate, and interpretable dietary recommendations to individual users. The system allows users to independently select target individuals and, based on their multidimensional historical data (diet, physiology, and health monitoring), complete nutritional health analysis and generate dietary recommendations.

[0046] Individual data input: Historical dietary data: Supports continuous records from 7 to 90 days, including: Ingredient name, weight, cooking method (stir-fry / boil / steam / fry), meal time; basic personnel information: age, gender, BMI; health monitoring data: blood pressure, blood sugar, blood lipids, uric acid, hemoglobin, serum vitamin D, nutrient levels inferred from urine Raman spectroscopy, etc. Knowledge base retrieval: Constructing a nutrition and health knowledge base, including: "Chinese Dietary Reference Intakes (2023 Edition)"; and the "Catalogue of Registered and Filed Health Food Products" issued by the State Administration for Market Regulation. The retrieval strategy employed is the Reranker model, which cross-encodes user queries and candidate entries to output the top-K most relevant guide entries (K=5~10).

[0047] Generate Nutrition and Health Analysis Reports from Large Models: Prompt Template Design: You are a registered dietitian in China. Please generate a nutrition and health analysis report based on the following information: - User information: {Age} years, {Gender}, {BMI}; - Average daily intake over the past 7 days: Grains {X} g, Vegetables {Y} g, Fruits {Z} g, Meat, Poultry, Fish and Eggs {M} g, Dairy Products {N} g, Soybeans and Nuts {P} g, Oil and Salt {Q} g; - Abnormal health monitoring: {serum vitamin D 18 ng / mL}, {serum albumin 38 g / L}; - Dietary guidelines retrieved: {Top-K guidelines}; Require: 1. Point out any discrepancies with the guidelines; 2. List the three most critical nutrition questions; 3. Provide specific improvement suggestions (including food substitution, portion size adjustment, and meal time optimization); 4. Use plain and easy-to-understand Chinese output, and avoid piling up technical jargon.

[0048] Output example: Key issue 1: Severe calcium deficiency Your average calcium intake over the past 7 days is 480 mg / day, which is less than 60% of the recommended intake (800 mg / day).

[0049] Recommendation: Add 1 cup of low-fat milk (250ml, containing 300mg calcium) + 1 block of firm tofu (100g, containing 138mg calcium) daily; if lactose intolerant, choose unsweetened yogurt or low-lactose milk.

[0050] 2. Dietary recommendations are generated.

[0051] The large model generates dietary recommendations based on nutrition and health analysis reports, individual data, and the "Chinese Dietary Reference Intakes (2023 Edition)" in the knowledge base. Output example (excerpt from a complete user report) User: Ms. Zhang, 32 years old, mid-pregnancy, Hb 105g / L, calcium intake 420mg / day, lactose intolerance. ① Nutritional deficit 1. Calcium (-580 mg) 2. Iron (-10 mg) 3. Dietary fiber (-9 g) ② Food mapping Calcium: 250 ml (300 mg) low-lactose yogurt + 100 g (138 mg) firm tofu + 10 g (117 mg) sesame paste → Total 555 mg Iron: 50 g duck blood (12.5 mg) + 100 g stir-fried beef with green peppers (4 mg) + 200 g strawberries (1.2 mg) → Total 17.7 mg Dietary fiber: 50g dry oatmeal (5g) + 200g broccoli (6g) + 1 apple (4g) → 15g total ③ 7-day dietary recommendations (Day 1 example) Meal combinations, cooking methods, and key nutritional contributions of dining out Breakfast: Oatmeal and milk porridge (40g oatmeal + 200ml low-lactose milk) + 100g strawberries (microwave for 5 minutes) + 240mg calcium + 4g dietary fiber Lunch out: Stir-fried beef with green peppers set meal (80g beef + 150g rice) Add 100g of broccoli (listed under Meituan's "Light Meal" label) which contains 3.2mg of iron and 3g of dietary fiber. 100g low-sugar yogurt (convenience store refrigerated display case) 120mg calcium Dinner: Duck blood and tofu soup (50g duck blood + 100g firm tofu) + 50g dry oatmeal + 100g stir-fried broccoli (cooked for 15 minutes). Calcium 258mg + Iron 12.5mg + Dietary fiber 5g Total for the day: Calcium 618 mg (target 62%) Iron 15.7 mg (meeting the target) 12 g of dietary fiber (replenish the remaining 13 g the next day). In summary, this invention provides a method for nutritional health analysis and dietary recommendation generation based on machine learning and large-scale models. It addresses issues such as high testing costs and long cycles, making high-frequency monitoring difficult; nutrient deficiency risk assessment relying on manual interpretation, leading to strong subjectivity; and dietary recommendations often using static templates, lacking individualization and dynamic adjustment capabilities. By constructing and optimizing a nutrient risk assessment model, rapid and non-invasive prediction of nutrient status such as vitamin D, albumin, and ferritin can be achieved; subsequently, a large-scale model is used as the engine to generate personalized dietary recommendations. This method can be widely applied in hospital nutrition departments, physical examination centers, maternal and infant institutions, and family health management scenarios.

[0052] Device Example 1 According to embodiments of the present invention, a device for nutritional health analysis and dietary recommendation generation based on a large model is provided. Figure 4 This is a schematic diagram of a large-model-based nutritional health analysis and dietary recommendation generation device according to an embodiment of the present invention, such as... Figure 4 As shown, the large-model-based nutritional health analysis and dietary recommendation generation device according to an embodiment of the present invention specifically includes: Nutrient risk assessment module 40 is used to automatically build and optimize a nutrient risk assessment model on a visual machine learning platform using Raman spectroscopy data and historical datasets of nutrient indicators. Based on newly monitored Raman spectroscopy data, this model predicts the user's nutrient status information. Specifically, it is used for: Raman spectral data covering specific wavelength band spectral intensity information are acquired using a portable or desktop Raman spectrometer; key nutritional status indicators are obtained through clinical biochemical testing, wherein the key nutritional status indicators specifically include: vitamin D, albumin, and / or ferritin; The algorithm integrates feature selection algorithms and / or classification / regression algorithms from the algorithm pool, and automatically optimizes the hyperparameters of the feature selection algorithms and / or classification / regression algorithms. It automatically builds and optimizes multiple nutrient risk assessment models for specific nutrients on a visual machine learning platform, and performs five-fold hierarchical cross-validation on the nutrient risk assessment models to select the optimal model as the final nutrient risk assessment model. The optimization methods specifically include Bayesian optimization and Hyperband early cessation strategy.

[0053] The newly monitored Raman spectral data undergoes wavelength calibration, baseline correction, smoothing, denoising, and normalization. The feature selection module corresponding to the final nutrient risk assessment model is then invoked to extract key spectral variables, which are input into the final nutrient risk assessment model. The model outputs predicted nutrient index values, maps these predicted values ​​to corresponding nutrient index levels, and outputs a structured report. The structured report specifically includes: predicted nutrient index values, nutrient index levels, confidence scores, and visualization charts.

[0054] The nutrition and health analysis and dietary recommendation generation module 42 is used to generate a nutrition and health analysis report by dynamically searching an authoritative knowledge base based on user basic information, historical dietary data, and health monitoring data using a large language model as an engine, and employing retrieval enhancement generation technology. Based on the nutrition and health analysis report, it generates personalized dietary recommendations. The health monitoring data includes nutrient status information. Specifically, the nutrient status information includes nutrient index levels, which specifically include: normal, insufficient, deficient, and severely deficient.

[0055] The embodiments of the present invention are device embodiments corresponding to the above method embodiments. The specific operation of each module can be understood with reference to the description of the method embodiments, and will not be repeated here.

[0056] Device Example 2 This invention provides an electronic device, such as... Figure 5 As shown, it includes: a memory 50, a processor 52, and a computer program stored in the memory 50 and executable on the processor 52, wherein the computer program, when executed by the processor 52, performs the steps as described in the method embodiment.

[0057] Device Example 3 This invention provides a computer-readable storage medium storing an information transmission implementation program, which, when executed by a processor 52, performs the steps described in the method embodiment.

[0058] The computer-readable storage media described in this embodiment include, but are not limited to, ROM, RAM, disk, or optical disk.

[0059] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for nutritional health analysis and dietary recommendation generation based on a large model, characterized in that, include: By using Raman spectroscopy data and historical datasets of nutritional indicators, a nutrient risk assessment model is automatically built and optimized on a visualization machine learning platform. Based on newly monitored Raman spectroscopy data, the nutrient risk assessment model predicts the user's nutrient status information. Using a large language model as the engine, based on user basic information, historical dietary data and health monitoring data, the system dynamically retrieves authoritative knowledge bases using retrieval enhancement generation technology, generates a nutrition and health analysis report, and generates personalized dietary recommendations based on the nutrition and health analysis report. The health monitoring data includes nutrient status information.

2. The method according to claim 1, characterized in that, Using Raman spectroscopy data and historical datasets of nutritional indicators, a nutrient risk assessment model is automatically built and optimized on a visualization machine learning platform. This includes: Raman spectral data covering specific wavelength band spectral intensity information are acquired using a portable or desktop Raman spectrometer; key nutritional status indicators are obtained through clinical biochemical testing, wherein the key nutritional status indicators specifically include: vitamin D, albumin, and / or ferritin; The algorithm integrates feature selection and classification / regression algorithms from the algorithm pool and automatically optimizes the hyperparameters of the feature selection and / or classification / regression algorithms. It automatically builds and optimizes multiple nutrient risk assessment models for specific nutrients on a visual machine learning platform. The nutrient risk assessment models are then subjected to five-fold hierarchical cross-validation, and the optimal model is selected as the final nutrient risk assessment model. The optimization methods specifically include Bayesian optimization and Hyperband early cessation strategy.

3. The method according to claim 1, characterized in that, The nutrient status information specifically includes: nutrient index levels, which specifically include: normal, insufficient, deficient, and severely deficient.

4. The method according to claim 1, characterized in that, Based on newly monitored Raman spectral data, the nutrient risk assessment model predicts the user's nutrient status information, specifically including: The newly monitored Raman spectral data undergoes wavelength calibration, baseline correction, smoothing, denoising, and normalization. The feature selection module corresponding to the final nutrient risk assessment model is then invoked to extract key spectral variables, which are input into the final nutrient risk assessment model. The model outputs predicted nutrient index values, maps these predicted values ​​to corresponding nutrient index levels, and outputs a structured report. The structured report specifically includes: predicted nutrient index values, nutrient index levels, confidence scores, and visualization charts.

5. A device for nutritional health analysis and dietary recommendation generation based on a large model, characterized in that, include: A prediction module is constructed to automatically build and optimize a nutrient risk assessment model on a visualization machine learning platform using Raman spectroscopy data and historical datasets of nutrient indicators. Based on newly monitored Raman spectroscopy data, the nutrient risk assessment model predicts the user's nutrient status information. The generation module is used to dynamically retrieve authoritative knowledge bases based on user basic information, historical dietary data, and health monitoring data using a large language model as the engine, and to generate a nutrition and health analysis report. Based on the nutrition and health analysis report, personalized dietary recommendations are generated. The health monitoring data includes nutrient status information.

6. The apparatus according to claim 5, characterized in that, The prediction construction module is specifically used for: Raman spectral data covering specific wavelength band spectral intensity information are acquired using a portable or desktop Raman spectrometer; key nutritional status indicators are obtained through clinical biochemical testing, wherein the key nutritional status indicators specifically include: vitamin D, albumin, and / or ferritin; The algorithm integrates feature selection algorithms and / or classification / regression algorithms from the algorithm pool, and automatically optimizes the hyperparameters of the feature selection algorithms and / or classification / regression algorithms. It automatically builds and optimizes multiple nutrient risk assessment models for specific nutrients on a visual machine learning platform, and performs five-fold hierarchical cross-validation on the nutrient risk assessment models to select the optimal model as the final nutrient risk assessment model. The optimization methods specifically include Bayesian optimization and Hyperband early cessation strategy.

7. The apparatus according to claim 5, characterized in that, The nutrient status information specifically includes: nutrient index levels, which specifically include: normal, insufficient, deficient, and severely deficient.

8. The apparatus according to claim 5, characterized in that, The prediction construction module is specifically used for: The newly monitored Raman spectral data undergoes wavelength calibration, baseline correction, smoothing, denoising, and normalization. The feature selection module corresponding to the final nutrient risk assessment model is then invoked to extract key spectral variables, which are input into the final nutrient risk assessment model. The model outputs predicted nutrient index values, maps these predicted values ​​to corresponding nutrient index levels, and outputs a structured report. The structured report specifically includes: predicted nutrient index values, nutrient index levels, confidence scores, and visualization charts.

9. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the steps of the method for generating nutritional health analysis and dietary recommendations based on a large model as described in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores an implementation program for information transmission, which, when executed by a processor, implements the steps of the method for generating nutritional health analysis and dietary recommendations based on a large model as described in any one of claims 1 to 4.

Images