User portrait generation system based on patient proinflammatory dietary behavior analysis

Abstract

The invention belongs to the technical field of artificial intelligence and medical health crossing, and particularly relates to a user portrait generation system based on patient pro-inflammatory dietary behavior analysis. According to the system, a dynamic user portrait containing a pro-inflammatory risk level, a behavior vulnerability dimension and an intervention suggestion is generated by collecting dietary and physiological data, calculating a time serialization pro-inflammatory load index, identifying a dietary behavior pattern and fusing multi-dimensional features such as clinical features, gene features, microorganism features and lifestyle features by utilizing an interpretable machine learning model; precise nutrition intervention and continuous tracking are supported; the system outputs a dietary behavior vulnerability dimension through an interpretable machine learning model, so that a doctor or a nutritionist can quickly understand a core problem of a patient and formulate a targeted strategy, and doctor-patient communication efficiency and intervention compliance are greatly improved.

Description

Technical Field

[0001] This invention belongs to the field of interdisciplinary technology of artificial intelligence and medical health, specifically involving a user profile generation system based on the analysis of patients' pro-inflammatory dietary behaviors. Background Technology

[0002] In the field of digital health and precision nutrition management, building user profiles based on individual behavioral data has become an important technological approach for achieving personalized interventions. This field integrates nutritional epidemiology, behavioral science, and artificial intelligence technologies, aiming to identify potential health risks and provide customized recommendations by analyzing users' dietary habits, lifestyles, and health indicators. Among these, pro-inflammatory dietary behaviors, as key triggers for chronic inflammation and related metabolic diseases, require quantitative assessment and dynamic tracking for disease prevention.

[0003] Among them, the user profile generation system based on the analysis of patients' pro-inflammatory dietary behaviors focuses on extracting pro-inflammatory features from multi-source dietary records (such as food frequency questionnaires, mobile application logs, or image recognition data) and constructing a structured user profile reflecting an individual's level of inflammatory risk. The core of this technical direction lies in transforming complex dietary patterns into calculable indicators of inflammatory potential and modeling their correlation with other dimensions of the user's health to support clinical decision-making or health management services.

[0004] Existing technologies face multiple challenges in creating such user profiles: dietary data collection relies on user input, leading to recording bias and incompleteness; pro-inflammatory scoring models are mostly based on static food composition databases, making it difficult to dynamically adapt to individual metabolic differences and the impact of food processing methods; existing profiling systems lack the ability to integrate temporal and contextual factors (such as mealtime and emotional state), resulting in coarse-grained profiles and limited predictive validity; furthermore, strong semantic heterogeneity between different data sources and the lack of a unified framework for representing pro-inflammatory behaviors restrict the generalized application of profiling in cross-platform health services. Therefore, a systematic solution is urgently needed that can integrate multi-dimensional dynamic dietary behavior data, accurately quantify individual pro-inflammatory risks, and generate highly interpretable user profiles. Summary of the Invention

[0005] The purpose of this invention is to provide a user profile generation system based on the analysis of patients' pro-inflammatory dietary behaviors, in order to solve the problem that the current management of chronic inflammation-related diseases lacks technical means to dynamically, quantitatively and structurally model the individual dietary pro-inflammatory potential, which makes it difficult to formulate personalized nutritional intervention strategies and predict and track the intervention effects.

[0006] The technical solution of this invention is a user profile generation system based on patient pro-inflammatory dietary behavior analysis, comprising a dietary data acquisition module, a pro-inflammatory index calculation module, a behavior pattern recognition module, a multi-dimensional feature fusion module, and a dynamic profile generation module. The dietary data acquisition module continuously acquires the user's dietary intake records via a mobile terminal or wearable device. These records include food types, intake amounts, intake times, intake frequencies, and accompanying physiological state data. The pro-inflammatory index calculation module, based on the dietary intake records, calls a pre-built dietary pro-inflammatory potential database to perform pro-inflammatory score mapping on each food item and performs weighted calculation based on intake amounts to generate a time-series-based individual pro-inflammatory load index. The behavior pattern recognition module is used to analyze the time series data... The individual pro-inflammatory load index is used for time-series pattern mining to identify statistically significant dietary behavior patterns, including periodic bursts of high pro-inflammatory load, stable patterns of persistently low pro-inflammatory load, and correlation patterns of pro-inflammatory load fluctuating with diurnal rhythms or emotional states. The multi-dimensional feature fusion module is used to perform cross-domain alignment and feature embedding of the dietary behavior patterns with the user's clinical indicator data, gene polymorphism information, gut microbiome characteristics, and lifestyle parameters to construct a high-dimensional heterogeneous feature vector. The dynamic profile generation module is used to generate a structured user profile based on the high-dimensional heterogeneous feature vector through an interpretable machine learning model. The user profile includes pro-inflammatory risk level, dietary behavior vulnerability dimension, sensitive window period for nutritional intervention, and personalized dietary adjustment suggestions.

[0007] Furthermore, the dietary pro-inflammatory potential database is constructed based on large-scale epidemiological studies and in vitro cytokine release experimental data. Each food item is associated with a standardized pro-inflammatory score, which comprehensively reflects the potential impact of the food on serum C-reactive protein, interleukin-6, and tumor necrosis factor-α levels after intake. The score ranges from -10 to +10, with negative values ​​indicating anti-inflammatory effects and positive values ​​indicating pro-inflammatory effects.

[0008] Furthermore, when performing weighted calculations, the pro-inflammatory index calculation module uses the logarithmic transformation value of the intake as a weighting factor and introduces a food processing method correction coefficient. The food processing method correction coefficient quantifies and adjusts the amplification or inhibition effect of pro-inflammatory potential according to different processing methods such as frying, grilling, pickling, or raw consumption. The correction coefficient ranges from 0.5 to 2.0.

[0009] Furthermore, the behavior pattern recognition module uses a sliding window combined with a hidden Markov model to classify the individual pro-inflammatory load index sequence into states, identifying three types of pro-inflammatory states: low, medium, and high. It then uses the Viterbi algorithm to decode the most likely state transition path and extracts the state dwell time, state transition frequency, and state triggering conditions as the core parameters of the behavior pattern.

[0010] Furthermore, the multi-dimensional feature fusion module adopts a graph neural network architecture to achieve cross-domain feature alignment, where nodes represent feature entities from different data sources, edges represent biological associations or statistical correlations between features, and neighbor information is aggregated through a multi-layer message passing mechanism to generate a unified embedding vector for each user, wherein the embedding vector has a dimension of 128.

[0011] Furthermore, the interpretable machine learning model used in the dynamic profile generation module is a gradient boosting tree ensemble model based on the attention mechanism. While predicting the pro-inflammatory risk level, the model outputs the contribution weight of each input feature to the prediction result. The contribution weight is used to generate interpretive labels for the dietary behavior vulnerability dimension, including "high sugar driven", "red meat dependent", "processed food sensitive" or "dietary fiber deficient".

[0012] Furthermore, the dynamic profile generation module is also configured with a profile update triggering mechanism. When newly collected dietary data causes the 7-day moving average of the individual's pro-inflammatory load index to change by more than 15%, or when the behavior pattern recognition module detects a new significant behavior pattern, the incremental update of the user profile is automatically triggered, and a profile version change log is recorded.

[0013] Furthermore, the system also includes an intervention feedback closed-loop module, which receives user feedback data on personalized dietary adjustment suggestions. The feedback data includes suggestion adoption rate, actual intake deviation, and subjective feeling score. The system then feeds the feedback data back into the multi-dimensional feature fusion module to dynamically calibrate the intervention sensitive window period parameters in the user profile.

[0014] Furthermore, the clinical indicator data include fasting blood glucose, glycated hemoglobin, lipid profile, liver function indicators, and systemic inflammatory markers; the gene polymorphism information includes single nucleotide polymorphism sites related to inflammatory pathways, covering TNF, IL6, CRP, and PPARG genes; the gut microbiome characteristics include the relative abundance ratio of pro-inflammatory bacteria to anti-inflammatory bacteria; and the lifestyle parameters include sleep duration, physical activity intensity, stress level, and smoking and drinking status.

[0015] Furthermore, the system operates in a cloud-edge collaborative computing architecture. The dietary data acquisition module is deployed on the user terminal device, the pro-inflammatory index calculation module and the behavior pattern recognition module are deployed on the edge computing node to ensure data privacy and real-time performance, and the multi-dimensional feature fusion module and the dynamic profile generation module are deployed on the cloud server to utilize high-performance computing resources for complex model inference.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: This invention, for the first time, constructs a complete technical chain from raw dietary behavior to structured user profiles. By quantifying the pro-inflammatory load of an individual's diet and identifying its dynamic behavioral patterns, it overcomes the limitations of traditional nutritional assessments that rely solely on static dietary frequency questionnaires or single nutrient analysis. The system introduces multi-source heterogeneous health data for cross-domain fusion, enabling the generated user profiles to not only reflect dietary behavior itself but also embed individual biological background and lifestyle context, significantly improving the clinical relevance and intervention guidance value of the profiles. The adopted dynamic update mechanism and intervention feedback closed-loop design give the user profiles continuous evolution capabilities, enabling them to truly reflect the dynamic process of user behavior changes and physiological responses, providing a quantifiable, traceable, and interpretable decision support tool for precision nutritional intervention in chronic inflammation-related diseases. In addition, the system outputs a dietary behavior vulnerability dimension through an interpretable machine learning model, enabling doctors or nutritionists to quickly understand the core issues of patients and develop targeted strategies, greatly improving the efficiency of doctor-patient communication and intervention adherence. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall technical solution architecture of the present invention; Figure 2 This is a schematic diagram of the core principle framework of multi-dimensional feature fusion and dynamic portrait generation in this invention. Detailed Implementation

[0018] Example 1 To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be described in detail below with reference to specific embodiments. This embodiment provides a user profile generation system based on the analysis of patients' pro-inflammatory dietary behaviors. The core objective is to construct a dynamic, quantitative, and structured technical means to assess the pro-inflammatory potential of an individual's diet, thereby solving the technical challenges of developing personalized nutritional intervention strategies and predicting and tracking intervention effects in the current management of chronic inflammation-related diseases. This system achieves a comprehensive transformation from raw dietary behavior data to structured user profiles with clinical guidance through refined data collection, multi-level data processing, intelligent pattern recognition, and interpretable machine learning models.

[0019] The system's workflow begins with the continuous acquisition of dietary data, followed by a quantitative assessment of the pro-inflammatory potential of this data, and further identification of unique dietary behavior patterns among users. Next, the system deeply integrates these behavioral patterns with multi-source, heterogeneous biomedical and lifestyle data, ultimately generating structured user profiles through interpretable machine learning models. The entire system architecture is designed as a cloud-edge collaborative computing model, ensuring a balance between data privacy, real-time processing, and high-performance, complex model inference.

[0020] Dietary Data Acquisition Module: This module forms the front end of the system, continuously acquiring users' dietary intake records via mobile terminals or wearable devices. These records include detailed information on food types, intake amounts, timing, frequency, and accompanying physiological data. The module's design prioritizes user-friendliness, data continuity, accuracy, and privacy.

[0021] Data Acquisition Mechanism: This module utilizes smartphones, smartwatches, and other mobile devices or wearable devices that users carry daily as data input interfaces. A specially designed application is deployed on the mobile device, allowing users to conveniently record their dietary intake through various methods such as text input, voice recognition, image recognition, and barcode scanning. For food types, the application provides a preset food database for intelligent matching, and also allows users to customize new food entries and upload images for auxiliary recognition. Intake is quantified using standardized units (e.g., grams, milliliters, servings) and can be obtained through user input, image estimation, or smart scale integration. Intake time and frequency are automatically recorded by the application.

[0022] Accompanying physiological data acquisition: In addition to dietary information, this module also integrates and collects users' physiological data through wearable devices (such as smartwatches and smart bracelets). This data includes, but is not limited to, heart rate, steps, sleep duration, body temperature, and fasting and postprandial blood glucose data obtained from a blood glucose meter connected via an external interface. All physiological data is synchronized to the mobile terminal application in real time or near real time and timestamped with the corresponding dietary intake records to form a complete diet-physiological correlation dataset.

[0023] Data Preprocessing and Storage: Before data is transmitted to the backend processing, the dietary data acquisition module performs preliminary data verification and cleaning on the terminal device. This includes checking the data format compliance, removing duplicate records, and handling null and outlier values ​​(e.g., through interpolation, mean filling, or warning the user to correct). After local encryption, the data is uploaded to edge computing nodes or cloud storage via a secure transmission protocol (e.g., Transport Layer Security). To ensure data traceability and integrity, each data record includes a unique identifier, timestamp, geographic location information (optional), and data source device identifier. The data is stored in a distributed time-series database in the cloud for efficient querying and analysis by subsequent modules.

[0024] Pro-inflammatory index calculation module: This module is responsible for converting the raw dietary intake records obtained by the dietary data acquisition module into a time-series-based individual pro-inflammatory load index. This conversion process is the core component of this system for quantifying the pro-inflammatory potential of diet.

[0025] Data Input and Food Entry Matching: This module receives structured dietary intake records from the dietary data acquisition module. For each food category in a record, the module first performs a precise match against a pre-built dietary pro-inflammatory potential database. The matching process employs an algorithm based on text similarity (e.g., cosine similarity, Jaccard similarity) and incorporates hierarchical relationships from food ontology for multi-level matching to ensure accuracy. If a new food entry is encountered that does not exist in the database, the system marks it as an item awaiting review, which may trigger manual review or supplementary learning through an external nutrition knowledge graph.

[0026] Dietary Pro-inflammatory Potential Database: This database forms the basis for the pro-inflammatory score mapping in this module. It is constructed based on large-scale epidemiological study data and in vitro cytokine release experimental data. Epidemiological studies provide evidence of the long-term correlation between food intake and population inflammatory markers (e.g., high-sensitivity C-reactive protein, interleukin-6, tumor necrosis factor alpha), while in vitro experiments provide direct effects of food extracts or metabolites on the secretion of inflammatory factors by immune cells. Each food entry in the database is associated with a standardized pro-inflammatory score, ranging from -10 to +10. Negative values ​​indicate that the food has an anti-inflammatory effect, inhibiting the inflammatory response or promoting inflammation resolution; positive values ​​indicate a pro-inflammatory effect, inducing or exacerbating the inflammatory response. For example, deep-sea fish rich in omega-3 fatty acids may have a negative score, while processed foods high in saturated fat and sugar may have a positive score. The database is maintained using a version control mechanism and is regularly updated and validated based on new scientific research findings.

[0027] Weighted Calculation of Pro-inflammatory Score: After successfully matching a food item and obtaining its baseline pro-inflammatory score, this module performs a weighted calculation based on intake. The weighting factor here uses the logarithmic transformation of intake to reflect the biological characteristic that the pro-inflammatory load is not linearly related to intake, but rather its marginal effect gradually decreases with increasing intake. Specifically, in the calculation, if the intake is... Then the weighting factor is or The choice of the cardinality is based on the characteristics of the data distribution.

[0028] Food Processing Method Correction: To more accurately assess the pro-inflammatory potential of food, this module further introduces a food processing method correction coefficient. This coefficient quantifies the amplification or inhibition of the pro-inflammatory potential of food based on different processing methods such as frying, grilling, pickling, raw consumption, and boiling. The correction coefficient ranges from 0.5 to 2.0. For example, for a food with an original pro-inflammatory score of 2, if it is fried, its correction coefficient may be 1.5, amplifying the pro-inflammatory score to 2 × 1.5 = 3; if it is eaten raw or simply boiled, the correction coefficient may be 0.8, reducing the pro-inflammatory score to 2 × 0.8 = 1.6. These correction coefficients are constructed based on research data on the effects of food science, nutrition, and cooking on nutrient composition, and can be refined according to different food categories. Finally, the individual pro-inflammatory load index for each meal or day is the sum or average of the corrected pro-inflammatory scores of all ingested foods and weighting factors, forming continuous time series data.

[0029] Behavioral Pattern Recognition Module: This module aims to perform in-depth analysis of the time-series individual pro-inflammatory load indices generated by the pro-inflammatory index calculation module to identify statistically significant dietary behavioral patterns. These patterns can reveal the inherent regularities and potential risks of users' dietary behaviors.

[0030] The foundation of time-series pattern mining: This module receives continuous time series of individual pro-inflammatory load indices. To effectively identify patterns, the system first applies a sliding window technique to segment the time series. The width of the sliding window can be dynamically adjusted according to the time granularity of the analysis (e.g., 24 hours, 7 days) and the fluctuation characteristics of the data, while the step size of the window determines the density of pattern recognition. Through the sliding window, the system transforms the continuous pro-inflammatory load index sequence into a series of local subsequences with temporal context.

[0031] Hidden Markov Model State Classification: Within each sliding window, or over the entire time series, the module uses a Hidden Markov Model to classify the individual pro-inflammatory load index sequence into states. A Hidden Markov Model is a statistical model used to describe a stochastic process containing latent variables. In this system, the observable variable is the pro-inflammatory load index, while the latent variable is the user's dietary pro-inflammatory state. The system pre-defines or learns three pro-inflammatory states—low, medium, and high—through clustering algorithms.

[0032] Low pro-inflammatory state: The corresponding pro-inflammatory load index is consistently below a certain preset threshold, indicating that the diet is mainly composed of anti-inflammatory foods and the intake of processed foods is extremely low.

[0033] Moderate pro-inflammatory state: The corresponding pro-inflammatory load index fluctuates in the middle range, indicating that the diet is relatively balanced, or that pro-inflammatory foods are occasionally consumed, but the frequency and amount are within a controllable range.

[0034] High pro-inflammatory state: The corresponding pro-inflammatory load index is consistently higher than a certain high threshold, indicating that the dietary structure contains a large proportion of pro-inflammatory foods or a large intake of processed foods.

[0035] The training data for the Hidden Markov Model comes from a large number of historical pro-inflammatory load index sequences of users. The model parameters, including the initial state probability, the state transition probability matrix, and the observation probability distribution, are estimated using the expectation-maximization algorithm.

[0036] Viterbi Algorithm Decoding and Pattern Parameter Extraction: After the Hidden Markov Model (HMM) is trained, this module uses the Viterbi algorithm to decode the individual pro-inflammatory load index sequence to find the most likely state transition path. The Viterbi algorithm can efficiently calculate the most likely hidden state sequence for a given observation sequence. By analyzing this optimal path, the system can extract core behavioral pattern parameters. State dwell time: The length of time a user remains in a pro-inflammatory state (e.g., a hyper-pro-inflammatory state). A longer hyper-pro-inflammatory state dwell time indicates a sustained risk of dietary pro-inflammatory effects.

[0037] State transition frequency: How frequently a user transitions between different pro-inflammatory states. Frequent transitions from low to high pro-inflammatory states, followed by rapid recovery, may indicate periodic "indulgence" behavior.

[0038] State triggers: The module identifies external events or internal physiological changes that lead to state transitions through association analysis and machine learning classifiers (e.g., decision trees or logistic regression). These triggers may include specific times (e.g., weekends), specific emotional states (e.g., stress, anxiety), or associations with specific physiological indicators (e.g., blood sugar fluctuations). For example, the system might identify a "periodic burst pattern of high pro-inflammatory load" that typically occurs at weekend dinners, or that "pro-inflammatory load fluctuates with emotional state" and significantly increases when the user feels stressed.

[0039] Multidimensional Feature Fusion Module: The core task of the multidimensional feature fusion module is to perform cross-domain alignment and feature embedding of the dietary behavior patterns output by the behavior pattern recognition module with the user's clinical indicator data, gene polymorphism information, gut microbiome characteristics, and lifestyle parameters to construct a high-dimensional heterogeneous feature vector. This process aims to integrate the user's multifaceted health information to form a comprehensive and unified digital representation.

[0040] Data Sources and Feature Extraction: This module receives input from multiple independent data sources. Dietary behavior pattern data: Behavioral pattern parameters generated by the behavior pattern recognition module, such as the residence time and switching frequency of high, medium and low pro-inflammatory states, as well as the corresponding triggering conditions.

[0041] Clinical indicators include fasting blood glucose, glycated hemoglobin, lipid profile (e.g., total cholesterol, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, triglycerides), liver function indicators (e.g., alanine aminotransferase, aspartate aminotransferase, total bilirubin), and systemic inflammatory markers (e.g., high-sensitivity C-reactive protein, erythrocyte sedimentation rate). These data are standardized and cleaned using the medical institution's electronic health record system or user-uploaded test reports.

[0042] Genetic polymorphism information: This includes single nucleotide polymorphisms (SNPs) associated with inflammatory pathways, such as specific sites in the tumor necrosis factor (TNF), interleukin-6 (IL6), C-reactive protein (CRP), and peroxisome proliferator-activated receptor gamma (PPARG) gene. This data was obtained from gene sequencing services and genotyped (e.g., 0, 1, 2 represent different allele combinations).

[0043] Gut microbiome characteristics: mainly including the relative abundance ratio of pro-inflammatory to anti-inflammatory bacteria. These data were obtained through 16S ribosomal RNA sequencing or metagenomic sequencing analysis, and after bioinformatics processing, the abundance data of key microbial taxonomic units were extracted.

[0044] Lifestyle parameters include sleep duration, physical activity intensity (e.g., daily steps, minutes of exercise), stress level (estimated through questionnaires or heart rate variability from wearable devices), and smoking and drinking status. These data are supplemented partly by physiological data from the dietary data collection module and partly synchronized through user questionnaires or external smart devices.

[0045] Graph Neural Network Architecture: To achieve effective alignment and fusion of cross-domain features, this module adopts a graph neural network architecture. In this architecture: Node representation: Feature entities from different data sources are abstracted as nodes in the graph. For example, a user's pro-inflammatory dietary pattern can be a node, their fasting blood glucose level can be another node, a specific genotype can be a node, and the abundance of a certain gut microbiota can also be a node. To capture heterogeneity, nodes can be assigned different types (e.g., "dietary pattern node", "clinical indicator node", "gene node").

[0046] Edge representation: Edges between nodes represent biological associations or statistical correlations between features. For example, there are known biological pathway associations between genes and disease biomarkers, causal relationships between dietary patterns and gut microbiota, and statistical correlations between different clinical indicators. This association information can be obtained from existing biomedical knowledge graphs, literature, or through correlation analysis of large-scale anonymized data. The weight of the edge can represent the strength or confidence level of the association.

[0047] Message Passing Mechanism: Graph neural networks aggregate neighbor information through a multi-layered message passing mechanism. At each layer, each node collects information from its neighbors and merges its own features with those of its neighbors using an aggregation function (e.g., summation, averaging, max pooling, or more complex attention mechanisms) to update its own representation. This iterative message passing enables information to propagate and integrate efficiently throughout the heterogeneous graph, capturing complex high-order relationships between multimodal data.

[0048] Feature Embedding: After multiple layers of message passing, the graph neural network generates a unified embedding vector for each user. This embedding vector is a low-dimensional, dense representation of all fused features, with its dimension set to 128. This 128-dimensional vector captures the user's comprehensive features across multiple dimensions, including dietary behavior, clinical aspects, genetics, microbiome, and lifestyle, and provides highly refined input for the subsequent dynamic profile generation module.

[0049] Dynamic User Profile Generation Module: This module is the core of the system's decision-making process. Its function is to generate structured user profiles based on the high-dimensional heterogeneous feature vectors output by the multi-dimensional feature fusion module, using an interpretable machine learning model. This profile not only provides pro-inflammatory risk assessment but also delves into the underlying dietary behavioral vulnerabilities and offers specific intervention recommendations.

[0050] Interpretable Machine Learning Model: This module employs an attention-based gradient boosting tree ensemble model. Gradient boosting trees (such as XGBoost or LightGBM) are powerful ensemble learning models, adept at handling heterogeneous data and exhibiting good predictive performance. Introducing attention mechanisms, such as through SHAP (Shapley Additive Explanations) or LIME (Local Interpretable Model-agnostic Explanations), allows the model to output the contribution weights of each input feature (i.e., each dimension in the 128-dimensional embedding vector, or back to the original features) to the prediction result while simultaneously predicting the pro-inflammatory risk level. These contribution weights are crucial for the model to "explain" its decision-making process.

[0051] User profile content composition: The generated structured user profile contains several key components: Pro-inflammatory risk level: Based on the prediction results, the system classifies the user's pro-inflammatory risk into three levels: low, medium, and high. This provides clinicians and users with an intuitive risk assessment.

[0052] Dietary behavior vulnerability dimension: Based on the feature contribution weights output by the model, the module can identify the dietary-related features that have the greatest impact on users' pro-inflammatory risk. By semantically mapping these high-contribution features, the system generates explanatory labels, such as "high-sugar driven" (indicating that high sugar intake is the main driver of pro-inflammatory risk), "red meat dependent" (indicating that high red meat intake is a key factor), "processed food sensitive" (indicating a particularly strong reaction to processed foods), or "fiber deficient" (indicating that insufficient dietary fiber intake leads to increased pro-inflammatory risk). These dimensions directly point out the weakest links in users' dietary behavior that most need improvement.

[0053] Sensitive window for nutritional intervention: By combining the state transition trigger conditions of the behavioral pattern recognition module with the fluctuation cycle of clinical indicators, the system predicts the time period when a user is most likely to accept or most needs nutritional intervention. For example, if the system identifies that a user is prone to high pro-inflammatory dietary behaviors when under stress, the system will indicate that this is a sensitive intervention window when the user's stress level rises.

[0054] Personalized dietary adjustment recommendations: Based on pro-inflammatory risk levels, dietary behavior vulnerability dimensions, and intervention sensitivity windows, the system generates specific and actionable dietary adjustment recommendations. These recommendations can be macro-level (e.g., "increase vegetable and fruit intake"), micro-level (e.g., "replace the staple food at dinner with brown rice and reduce red meat intake"), or even precise down to the recommendation and avoidance of specific foods. The recommendations are filtered and optimized based on the user's historical preferences, allergy history, and clinical contraindications.

[0055] User profile update trigger mechanism: To ensure the real-time performance and accuracy of user profiles, the dynamic profile generation module is configured with an automatic update trigger mechanism. This mechanism will automatically trigger an incremental update of the user profile when one of the following two situations occurs: The newly collected dietary data caused a change of more than 15% in the 7-day moving average of the individual pro-inflammatory load index. This means that the user's overall dietary habits or pro-inflammatory levels have changed significantly and need to be reassessed.

[0056] The behavior pattern recognition module detected new and significant behavioral patterns. For example, the user shifted from a persistent low pro-inflammatory load pattern to a periodic burst pattern of high pro-inflammatory load.

[0057] When the triggering conditions are met, the system will automatically start a new profile generation cycle and record the profile version change log in order to track the evolution history of user profiles and the effects of interventions.

[0058] Intervention Feedback Closed-Loop Module: The intervention feedback closed-loop module is key to the system's continuous optimization and personalized learning. It receives user feedback data on personalized dietary adjustment recommendations and feeds this data back into the multi-dimensional feature fusion module for dynamically calibrating the intervention sensitivity window parameters in the user profile.

[0059] Feedback Data Collection: This module actively or passively collects user performance feedback data through mobile applications or associated smart health devices. This data includes: Suggestion adoption rate: The percentage of system suggestions that users actually adopt. For example, out of 5 system-recommended adjustments, 3 are implemented.

[0060] Actual intake deviation: The difference between the user's actual dietary intake and the system's recommendations. For example, the system recommends an intake of 100 grams of protein, but the user actually consumes 80 grams.

[0061] Subjective feeling rating: The subjective rating of the user's physical feeling after adopting the suggestions (e.g., mental state, gastrointestinal comfort, improvement of inflammatory symptoms), usually in the form of Likert scales.

[0062] Feedback Data Injection and Parameter Calibration: The received execution feedback data, after cleaning and standardization, is injected back into the multidimensional feature fusion module. In this module, this feedback data is treated as a special type of "contextual feature," participating in the feature fusion process along with the original clinical, genetic, and microbiome data. In this way, the system can learn the user's response patterns to different intervention recommendations and the actual effects of different intervention strategies.

[0063] Dynamic calibration of intervention sensitivity window parameters: Feedback data is particularly important for calibrating intervention sensitivity window parameters in user profiles. For example, if the system provides advice during a "high-pressure" period, but user feedback shows low adoption rates and negative subjective experiences, the system will learn that the user has low intervention sensitivity in this specific context. Conversely, if a suggestion has a high adoption rate and good results within a specific time period, the system will increase the weight given to that time period as a sensitivity window. This dynamic calibration process can be achieved through reinforcement learning algorithms or adaptive Bayesian updates, enabling the system to more accurately predict when intervention will yield the best results based on the user's actual behavior and physiological responses, thereby improving the effectiveness and compliance of interventions.

[0064] Cloud-edge collaborative computing architecture: This system adopts a cloud-edge collaborative computing architecture, which aims to balance the needs of data privacy, real-time processing capabilities and high-performance computing resources.

[0065] The dietary data collection module is deployed on user terminal devices: the client of the dietary data collection module is directly deployed on mobile terminals and wearable devices, responsible for the generation, preliminary verification, local encryption, and secure uploading of raw data. This maximizes the privacy of users' sensitive personal data, because the raw data is encrypted before leaving the user's device, and users have greater control over their data.

[0066] The pro-inflammatory index calculation module and the behavior pattern recognition module are deployed on edge computing nodes: edge computing nodes can be home smart gateways, community health station servers, or small regional data centers. Deploying these two modules on edge nodes offers the following advantages: Data privacy protection: The calculation of pro-inflammatory index and behavioral patterns involves detailed dietary information of users. Processing this information at the edge can reduce the transmission of raw sensitive data over wide area networks and reduce the risk of leakage.

[0067] Real-time performance: Performing computation at the edge can significantly reduce the latency of data transmission to the cloud, supporting near real-time pro-inflammatory load assessment and behavioral pattern analysis, thereby enabling more timely health feedback and early warning.

[0068] Network bandwidth optimization: After a large amount of raw dietary data is processed at the edge, only high-level pro-inflammatory indices and behavioral pattern summaries need to be transmitted to the cloud, effectively saving network bandwidth resources.

[0069] The multi-dimensional feature fusion module and the dynamic portrait generation module are deployed on a cloud server: the cloud server has powerful computing resources, storage capacity, and elastic scalability, and can support: Complex model inference: The graph neural network in the multi-dimensional feature fusion module and the attention mechanism gradient boosting tree model in the dynamic portrait generation module require a lot of computing resources for model training, inference and continuous optimization, which can be completed efficiently on high-performance computing clusters in the cloud.

[0070] Cross-domain data integration: Heterogeneous data such as clinical, genetic, and microbiome data are usually stored in different centralized databases. The cloud can easily integrate and manage these data in a unified manner.

[0071] Model updates and management: Centralized cloud-based model version management, global optimization, and secure deployment ensure that all users can enjoy the latest and most accurate profile services.

[0072] Communication between edge nodes and cloud servers utilizes encrypted application programming interfaces (APIs) and message queuing services to ensure the security and reliability of data transmission. This layered architecture effectively balances data security, computational efficiency, and system scalability.

[0073] This embodiment constructs a closed-loop intelligent user profile generation system through the collaborative work of the aforementioned modules. It no longer relies on traditional static questionnaires or single-nutrient analysis, but instead quantifies the pro-inflammatory load of an individual's diet and identifies its dynamic behavioral patterns, overcoming the limitations of traditional assessments. The system innovatively introduces multi-source heterogeneous health data for cross-domain fusion, enabling the generated user profile to not only reflect dietary behavior itself but also embed individual biological background and lifestyle context, significantly enhancing the clinical relevance and intervention guidance value of the profile. The adopted dynamic update mechanism and closed-loop intervention feedback design give the user profile continuous evolution capabilities, accurately reflecting the dynamic process of user behavior changes and physiological responses, providing a quantifiable, traceable, and interpretable decision support tool for precision nutritional intervention in chronic inflammation-related diseases. Furthermore, the system outputs a dietary behavior vulnerability dimension through an interpretable machine learning model, enabling doctors or nutritionists to quickly understand the patient's core issues and develop targeted strategies, significantly improving doctor-patient communication efficiency and intervention adherence.

[0074] Compared to existing technologies, the core technological advancement of this invention lies in transforming an individual's dietary pro-inflammatory potential from an abstract concept into a dynamic, quantifiable indicator, and revealing its behavioral patterns through deep time-series analysis. Secondly, it achieves seamless fusion of multi-source heterogeneous health data through graph neural networks, solving the challenge of integrating biological, clinical, and behavioral data using traditional methods. Thirdly, the introduction of an interpretable machine learning model enables user profiles to not only possess predictive capabilities but also provide semantic explanations of behavioral vulnerabilities, greatly enhancing their clinical application value. Finally, through dynamic updates and a feedback loop mechanism, the entire system can continuously learn and self-optimize, adapting to dynamic changes in individual health conditions, forming a continuously evolving precision health management platform.

[0075] Example 2 Based on the aforementioned embodiments, this embodiment further elaborates on the construction process, maintenance mechanism, and interaction details between the dietary pro-inflammatory potential database and the pro-inflammatory index calculation module, and provides a specific explanation of the method for determining the food processing method correction coefficient.

[0076] The construction of a dietary pro-inflammatory potential database is fundamental to the accurate quantification of individual pro-inflammatory loads in this system. This database construction was a complex, multi-stage, interdisciplinary process designed to ensure that the pro-inflammatory scores associated with each food item have a high degree of scientific basis and credibility.

[0077] Data source integration and preprocessing: Database construction begins with large-scale data collection and integration. Major data sources include: Large-scale epidemiological study data: This involves collecting cohort, case-control, and cross-sectional data from global studies on the correlation between dietary patterns, specific food intake, and levels of chronic inflammatory markers (such as serum C-reactive protein, interleukin-6, and tumor necrosis factor alpha). These data are typically presented in the form of articles or reports and require systematic literature review, data extraction, and standardization.

[0078] In vitro cytokine release experimental data: This study integrates laboratory results from cell culture models to test the effects of different food extracts, nutrients, or their metabolites on the release of inflammatory factors (e.g., interleukin-1 beta, tumor necrosis factor alpha) from immune cells (e.g., macrophages, lymphocytes). These data provide biological evidence that food directly influences inflammatory pathways.

[0079] Clinical intervention trial data: This involves collecting data from randomized controlled trials on the effects of specific dietary interventions (e.g., the Mediterranean diet, high-fat diet) on the levels of inflammatory markers in the human body. These data directly reflect the real effects of food on the human body.

[0080] All collected data underwent rigorous quality control and preprocessing before integration, including removing duplicate data, handling missing values, standardizing units of measurement, unifying disease diagnostic criteria, and calculating effect sizes and performing pooled analyses on different research findings (e.g., through meta-analysis).

[0081] Pro-inflammatory scoring quantification model: Based on data integration, a multi-factor comprehensive assessment model is used to quantify the pro-inflammatory score of each food. This model weights and integrates epidemiological association strength, in vitro experimental evidence, and clinical trial results.

[0082] Epidemiological weighting: This assesses the strength and statistical significance of the correlation between food intake and inflammatory marker levels using statistical methods (e.g., regression analysis, hazard ratio calculation). Foods with stronger and more consistent correlations are assigned higher weights.

[0083] Strength of evidence from in vitro experiments: The direct pro-inflammatory or anti-inflammatory effects of food are scored based on the magnitude of changes in the release of inflammatory factors and their dose-response relationship observed in in vitro experiments.

[0084] Clinical evidence level: Assign appropriate weights based on the level of evidence of the clinical intervention trial (e.g., evidence from randomized controlled trials is superior to that from observational studies) and the effect size.

[0085] Ultimately, the standardized pro-inflammatory score for each food was determined. Calculated using the following formula:

[0086] in, The weighting coefficients representing each piece of evidence, summing to 1; The standardized effect size represents the various pieces of evidence. The effect sizes are uniformly mapped to a rating range of -10 to +10. -10 represents the strongest anti-inflammatory effect, and +10 represents the strongest pro-inflammatory effect.

[0087] Database Structure and Maintenance: The Dietary Pro-inflammatory Potential Database adopts a relational database structure, with core tables including: Food body table: Includes unique food identifier, food name (Chinese and English), classification, common cooking methods, main nutritional components, etc.

[0088] Pro-inflammatory scoring scale: unique identifier for stored food, standardized pro-inflammatory score, literature citations on which the scoring is based, update date, etc.

[0089] Correction coefficient table: Stored food categories, processing methods, corresponding correction coefficient values, sources, etc.

[0090] Maintaining the database is an ongoing process. New scientific research findings and data are regularly imported into the database and validated through an expert review mechanism. Whenever new research alters the pro-inflammatory score or correction factor for a specific food, the database is incrementally updated, and a version change log is maintained to ensure all calculations are based on the latest scientific evidence.

[0091] Determination of correction coefficients for food processing methods Determining the correction factor for food processing methods is a crucial component of the accuracy of the pro-inflammatory index calculation module. Its purpose is to reflect the impact of different processing methods on the intrinsic pro-inflammatory or anti-inflammatory properties of food. The correction factor ranges from 0.5 to 2.0.

[0092] Experimental data acquisition: The correction coefficients were mainly determined in the following ways: Nutritional experiments: This involves analyzing changes in the nutritional components (e.g., lipid oxidation products, final glycation products, trans fatty acids, and antioxidants) of the same batch of food after different processing methods (e.g., raw, boiled, steamed, baked, fried, pickled). These changes are closely related to the intensity of the inflammatory response.

[0093] Bioactivity assay: Food extracts processed by different methods were subjected to in vitro cytokine release experiments to directly measure their effects on inflammatory responses.

[0094] Expert assessment and consensus: For processing methods lacking direct experimental data, a committee composed of experts in nutrition, food science, and inflammation and immunology conducts an assessment and determines preliminary correction coefficient values ​​based on existing knowledge and experience.

[0095] Iterative optimization and validation: The initially set correction coefficients will be validated through backtracking analysis of historical user data. For example, observe whether the actual changes in users' inflammatory marker levels after ingestion using a specific processing method are consistent with the predicted pro-inflammatory score. If a significant deviation exists, the correction coefficients will be iteratively adjusted and optimized.

[0096] For example, frying and grilling typically generate more advanced glycation end products (AGEs) and oxidized fats, which have strong pro-inflammatory effects, so their correction factor is usually greater than 1. Boiling and steaming, on the other hand, generally retain the nutrients in food and reduce the formation of harmful substances, so their correction factor may be close to 1 or slightly less than 1. Pickled foods may have pro-inflammatory potential due to their high salt content and certain additives, so their correction factor may be slightly higher than 1.

[0097] The pro-inflammatory index calculation module generates highly refined and personalized time series of individual pro-inflammatory load indices by accurately matching dietary data, accessing the latest version of the pro-inflammatory potential database, and combining intake weighting and processing method correction. This series not only reflects the pro-inflammatory effects of users' dietary behaviors but also captures their dynamic changes and potential risks, providing a solid data foundation for subsequent behavioral pattern recognition and user profile generation.

[0098] Example 3 Based on the aforementioned embodiments, this embodiment further elaborates on the working principle of the explainable machine learning model, the generation logic of the dietary behavior vulnerability dimension, and the specific implementation details of the profile update triggering mechanism in the dynamic profile generation module, and provides a detailed explanation of the generation and optimization process of personalized dietary adjustment suggestions.

[0099] Explaining the working principle of machine learning models: The dynamic portrait generation module adopts an attention-based gradient boosting tree ensemble model, whose core advantage lies in providing high-precision predictions while clearly revealing the decision-making basis.

[0100] Gradient Boosting Tree Ensemble: This model consists of multiple weak learners (usually decision trees) combined sequentially. Each tree learns from the residuals of the previous tree, progressively optimizing prediction performance. This model can handle high-dimensional heterogeneous feature data and automatically captures non-linear relationships and interactions between features.

[0101] Attention Mechanism and Feature Contribution Weights: In this model, the attention mechanism is not the self-attention layer in traditional deep learning, but rather refers to evaluating the contribution of each input feature to the final prediction result of the model (e.g., pro-inflammatory risk level) through methods such as SHAP (Shapley Additive Explanations) or LIME (Local Interpretable Model-agnostic Explanations).

[0102] SHAP Value Calculation: Based on cooperative game theory, the SHAP value fairly distributes the contribution of each feature to each prediction. For each user, the SHAP value quantifies how each input feature (e.g., "frequency of high sugar intake," "specific genotype of gene locus XX," "relative abundance of pro-inflammatory bacteria") affects the model's output, and sums these impact values, equaling the difference between the predicted value and the baseline value (the predicted value when all features are averaged). A positive SHAP value indicates that the feature pushes up the predicted value, while a negative value indicates that it pushes it down. By calculating the SHAP values ​​of all features, the precise contribution weight of each feature to pro-inflammatory risk in a user profile can be obtained.

[0103] The generation logic of the dietary behavior vulnerability dimension: Based on the feature contribution weights output by the model, the system can further generate an interpretable dietary behavior vulnerability dimension.

[0104] Feature contribution ranking: For each user, the system will sort the features in descending order based on the absolute value of their SHAP values ​​and find the top N features (e.g., N is set to 5 or 10) that have the greatest impact on the user's pro-inflammatory risk.

[0105] Feature semantic mapping: This process maps these high-contribution features to predefined dietary behavior vulnerability dimensions. This mapping process is based on nutritional knowledge and expert experience. For example: If the SHAP value of the "high sugar intake frequency" trait is the highest and positive, the user may be labeled as "high sugar driven".

[0106] If the SHAP value of the "red meat intake" characteristic is significantly positive, it may be labeled as "red meat dependent".

[0107] If the SHAP value of the "processed food intake ratio" characteristic is high and positive, it may be labeled as "processed food sensitive".

[0108] If the SHAP value of the "Dietary Fiber Intake" characteristic is significantly negative (indicating that a lack of dietary fiber increases the risk), it may be labeled as "Dietary Fiber Deficiency".

[0109] If the SHAP value of the "Frequency of Vegetable and Fruit Intake" characteristic is significantly negative, it may be labeled as "Insufficient Vegetable and Fruit Intake".

[0110] These dimensions provide highly semantic interpretations, helping doctors, nutritionists, and users quickly understand the root causes of pro-inflammatory risks. A user can possess multiple vulnerability dimensions simultaneously, reflecting the complexity of their dietary behaviors.

[0111] Implementation details of the portrait update triggering mechanism The profile update trigger mechanism ensures that user profiles can reflect users' latest health status and behavioral changes in a timely manner.

[0112] Individual Pro-inflammatory Burden Index 7-day Moving Average Change Detection: The system continuously calculates the 7-day moving average of each user's individual pro-inflammatory burden index. Each day, the system compares the current 7-day moving average with the previous day's 7-day moving average. If the change (absolute value) exceeds 15%, i.e. This triggers a profile update. This moving average-based detection method smooths out short-term fluctuations and focuses on long-term trends in dietary pro-inflammatory potential.

[0113] The behavior pattern recognition module detects new significant behavioral patterns: Profile updates are also triggered when the module detects, via a Hidden Markov Model, that a user is switching from one primary pro-inflammatory state (e.g., persistently low pro-inflammatory load) to another (e.g., periodic bursts of high pro-inflammatory load), or when it detects a new, previously unobserved steady-state pattern. This is typically done by comparing the newly identified pattern with the set of patterns recorded in the user's historical profile; if a significant difference exists and passes a statistical significance test, an update is triggered.

[0114] Log Recording and Notifications: Each time a user profile is updated, the system automatically records a detailed profile version change log, including the update time, triggering reason, and key differences between the old and new profile versions. Simultaneously, the system sends notifications to relevant healthcare professionals and users, reminding them that their user profile has been updated and suggesting they review the latest health report and recommendations.

[0115] Generation and optimization of personalized dietary adjustment recommendations: The generation of personalized dietary adjustment recommendations is an intelligent recommendation process based on multi-dimensional information of user profiles, and is continuously optimized through an intervention feedback loop.

[0116] Suggestion Library and Rule Engine: The system maintains a vast library of dietary adjustment suggestions, including strategies such as food substitutions, changes in cooking methods, and adjustments to meal timing. Each suggestion is associated with a specific pro-inflammatory risk factor and a vulnerability dimension related to dietary behavior. A rule-based expert system then filters an initial set of suggestions from the library based on the user's pro-inflammatory risk level, key vulnerability dimensions, and intervention sensitivity window.

[0117] Constraints and Preference Filtering: Based on the initial suggestion set, the system will further consider the user's personalized information for filtering and optimization. Clinical contraindications: For example, high-protein diet recommendations will be filtered out for patients with renal insufficiency.

[0118] Allergy history: Filter out foods that the user is known to be allergic to.

[0119] Dietary preferences: By analyzing users' historical dietary data or questionnaires, we can understand their taste preferences and dietary habits, and prioritize recommending foods and cooking methods that users are more likely to accept.

[0120] Feasibility assessment: Considering the user's geographical location, seasonal factors, and purchasing power, recommend available and affordable food options.

[0121] Multi-objective optimization and reinforcement learning: To generate optimal personalized recommendations, the system employs a multi-objective optimization algorithm to balance objectives such as reducing pro-inflammatory load, improving user compliance, and ensuring nutritional balance. User adoption rate and actual effect data provided by the intervention feedback loop module are used to train the reinforcement learning model. Through continuous trial and error and learning, the reinforcement learning model optimizes the recommendation generation strategy, enabling the system to dynamically adjust its recommendation algorithm based on users' personalized responses, thereby maximizing the long-term effectiveness of the intervention. For example, if a user has high compliance with the recommendation to "reduce sugar intake" and the effect is significant, the system will continue to reinforce this type of recommendation in the future; conversely, if a recommendation is theoretically effective but user compliance remains low, the system will learn to provide the user with alternative, more readily adoptable solutions.

[0122] This embodiment demonstrates, through a detailed explanation of the dynamic profile generation module, how the system, within a complex feature space, utilizes interpretable machine learning techniques to create a unique and insightful health profile for each user. These profiles not only provide intuitive risk assessments but also reveal deep-seated behavioral drivers and can be continuously optimized based on user feedback. This provides unprecedented intelligent support for precision nutritional intervention in chronic inflammation-related diseases, achieving a seamless transformation from data to executable and optimizable personalized health management plans.

Claims

1. A user profile generation system based on patient pro-inflammatory dietary behavior analysis, characterized in that, It includes a dietary data acquisition module, a pro-inflammatory index calculation module, a behavior pattern recognition module, a multi-dimensional feature fusion module, and a dynamic profile generation module; The dietary data acquisition module is used to continuously acquire the user's dietary intake records through a mobile terminal or wearable device. The dietary intake records include food types, intake amounts, intake times, intake frequencies, and accompanying physiological state data. The pro-inflammatory index calculation module is used to call a pre-built dietary pro-inflammatory potential database based on the dietary intake record, map the pro-inflammatory score for each food item, and perform weighted calculation in combination with the intake amount to generate a time-series individual pro-inflammatory load index. The behavior pattern recognition module is used to perform time-series pattern mining on the time-series individual pro-inflammatory load index to identify statistically significant dietary behavior patterns. The multidimensional feature fusion module is used to perform cross-domain alignment and feature embedding of the dietary behavior pattern with the user's clinical indicator data, gene polymorphism information, gut microbiome characteristics and lifestyle parameters to construct a high-dimensional heterogeneous feature vector; The dynamic profile generation module is used to generate a structured user profile based on the high-dimensional heterogeneous feature vector through an interpretable machine learning model. The user profile includes pro-inflammatory risk level, dietary behavior vulnerability dimension, sensitive window period for nutritional intervention, and personalized dietary adjustment suggestions. The behavior pattern recognition module uses a sliding window combined with a hidden Markov model to classify the individual pro-inflammatory load index sequence into three types of pro-inflammatory states: low, medium, and high. It then uses the Viterbi algorithm to decode the most likely state transition path and extracts the state dwell time, state transition frequency, and state triggering conditions as the core parameters of the behavior pattern.

2. The user profile generation system based on patient pro-inflammatory dietary behavior analysis according to claim 1, characterized in that, The dietary pro-inflammatory potential database is constructed based on large-scale epidemiological studies and in vitro cytokine release experimental data. Each food item is associated with a standardized pro-inflammatory score, which comprehensively reflects the potential impact of the food on serum C-reactive protein, interleukin-6 and tumor necrosis factor-α levels after intake. The score range is -10 to 10, with negative values ​​indicating anti-inflammatory effects and positive values ​​indicating pro-inflammatory effects.

3. The user profile generation system based on patient pro-inflammatory dietary behavior analysis according to claim 1, characterized in that, When performing weighted calculations, the pro-inflammatory index calculation module uses the logarithmic transformation value of intake as a weighting factor and introduces a food processing method correction coefficient. The food processing method correction coefficient quantifies and adjusts the amplification or inhibition effect of pro-inflammatory potential based on different processing methods such as frying, grilling, pickling, or raw consumption. The correction coefficient ranges from 0.5 to 2.

0.

4. The user profile generation system based on patient pro-inflammatory dietary behavior analysis according to claim 1, characterized in that, The multi-dimensional feature fusion module adopts a graph neural network architecture to achieve cross-domain feature alignment, where nodes represent feature entities from different data sources, edges represent biological associations or statistical correlations between features, and neighbor information is aggregated through a multi-layer message passing mechanism to generate a unified embedding vector for each user, wherein the embedding vector has a dimension of 128.

5. The user profile generation system based on patient pro-inflammatory dietary behavior analysis according to claim 1, characterized in that, The interpretable machine learning model used in the dynamic profile generation module is a gradient boosting tree ensemble model based on an attention mechanism. While predicting the pro-inflammatory risk level, this model outputs the contribution weights of each input feature to the prediction result. These contribution weights are used to generate interpretive labels for the dietary behavior vulnerability dimension.

6. The user profile generation system based on patient pro-inflammatory dietary behavior analysis according to claim 1, characterized in that, The dynamic profile generation module is also equipped with a profile update triggering mechanism. When newly collected dietary data causes the 7-day moving average of the individual's pro-inflammatory load index to change by more than 15%, or when the behavior pattern recognition module detects a new significant behavior pattern, the incremental update of the user profile is automatically triggered, and a profile version change log is recorded.

7. The user profile generation system based on patient pro-inflammatory dietary behavior analysis according to claim 1, characterized in that, It also includes an intervention feedback closed-loop module, which receives user feedback data on the implementation of personalized dietary adjustment suggestions. The implementation feedback data includes suggestion adoption rate, actual intake deviation and subjective feeling score. The implementation feedback data is then injected back into the multi-dimensional feature fusion module to dynamically calibrate the intervention sensitive window period parameters in the user profile.

8. The user profile generation system based on patient pro-inflammatory dietary behavior analysis according to claim 1, characterized in that, The clinical indicators include fasting blood glucose, glycated hemoglobin, lipid profile, liver function indicators, and systemic inflammatory markers; the gene polymorphism information includes single nucleotide polymorphism sites related to inflammatory pathways, covering TNF, IL6, CRP, and PPARG genes; the gut microbiome characteristics include the relative abundance ratio of pro-inflammatory bacteria to anti-inflammatory bacteria; and the lifestyle parameters include sleep duration, physical activity intensity, stress level, and smoking and drinking status.

9. The user profile generation system based on patient pro-inflammatory dietary behavior analysis according to claim 1, characterized in that, The system runs on a cloud-edge collaborative computing architecture. The dietary data acquisition module is deployed on user terminal devices, the pro-inflammatory index calculation module and the behavior pattern recognition module are deployed on edge computing nodes, and the multi-dimensional feature fusion module and the dynamic profile generation module are deployed on cloud servers.

Images