Personalized Nutritional Recommendation System Using Dynamic User Portraits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users face challenges in accurately determining their health state and selecting appropriate nutritional products due to reliance on subjective feelings and lack of nutritional knowledge, leading to ineffective symptom alleviation.
Innovation Solution
A computer-implemented method that acquires physiological and behavioral data to generate user portraits, using a knowledge graph and recall models to recommend personalized nutritional products by integrating static and dynamic labels, and sorting models for precise product suggestions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If users determine their physical condition based on vague subjective feelings, then they can avoid hospital visits and reduce time loss, but they cannot accurately determine their nutritional requirements and select appropriate products
Solution Approach 1:
The patent introduces a recommendation system as an intermediary between users and nutritional products. This system uses multiple data sources (physiological data, behavioral data, genetic test data) and processing modules (label generation, user portrait construction, recall models, sorting models) to accurately determine nutritional requirements and provide personalized recommendations, resolving the contradiction between avoiding hospital visits and accurately determining nutritional needs
2Loss of time
If users take nutritional products according to their own experience, then they can avoid professional medical advice and reduce time loss, but the type, quantity and frequency of products taken are random and ineffective
Solution Approach 1:
The patent enables users to serve themselves by providing personalized nutritional recommendations based on their individual data. The system automatically analyzes physiological data, behavioral data, and genetic test data to generate customized product recommendations, eliminating the need for professional medical advice while ensuring reliability and effectiveness
3Ease of operation
If users lack nutritional knowledge, then they can avoid the complexity of understanding nutritional information, but they cannot accurately determine which products to take based on their state data
Solution Approach 1:
The recommendation system acts as an intermediary that translates complex nutritional information into personalized recommendations. It processes user data through multiple modules (label generation, user portrait construction, recall models) to automatically determine appropriate products, eliminating the need for users to understand complex nutritional knowledge while maintaining accurate product selection
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
The present application discloses computer-implemented method, apparatus and medium for providing a recommendation of personalized nutritional products. The method includes: acquiring physiological data information in a plurality of dimensions of a user; generating a plurality of static labels that indicate a health state of the user according to the physiological data information in the plurality of dimensions; acquiring a plurality of behavior data of the user; generating a plurality of dynamic labels that indicate user preferences according to the plurality of behavior data; generating a user portrait based on the static labels and the dynamic labels; generating a plurality of recall lists that indicate correspondence between the user and different nutritional products through a trained recall model based on the plurality of dynamic labels and based on a knowledge graph of nutritional products, and constructing a merged recall list that integrates the plurality of recall lists; constructing a feature vector used for a sorting model by using the merged recall list and the user portrait, and inputting a calculated feature vector into the sorting model; and sorting the merged recall list according to a score calculated by the sorting model for an inputted feature vector, so as to determine the recommendation of personalized nutritional products suitable for the user.