Deep Learning Neural Networks for Food Risk Traceability Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current food traceability systems primarily focus on information collection and query services within the circulation link, leading to issues like information islands and link fractures, with limited research on basic information classification.
Innovation Solution
A computer-implemented method using deep learning neural networks for food risk traceability information classification, employing self-learning AI models to classify traceability information by converting factors into vectors and adjusting parameters through feedback mechanisms for autonomous learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If label carrier methods are used for food traceability information collection, then information can be recorded and queried, but information islands and link fractures occur
Solution Approach 1:
The patent merges multiple information sources and circulation link data into a unified analysis framework, combining production information, circulation information, and risk information into a single comprehensive evaluation system to eliminate information islands and achieve integrated traceability analysis
Solution Approach 2:
The patent creates a universal information classification framework that can handle multiple types of traceability information (production, circulation, risk, etc.) through a single standardized system, enabling the system to process diverse information types uniformly and reduce integration complexity
2Ease of operation
If circulation link oriented traceability is implemented, then information query services are provided, but basic information classification is insufficient
Solution Approach 1:
The patent segments traceability information into distinct classification categories (production information, circulation information, risk information, etc.) with specific classification dimensions and attributes, enabling precise classification while maintaining ease of query operations through structured organization
Solution Approach 2:
The patent applies different classification standards and dimensions to different types of traceability information, tailoring the classification approach to the specific characteristics of each information type (e.g., production data vs. circulation data) to achieve precise classification while preserving operational ease
3Measurement precision
If deep learning neural networks are used for information classification, then classification accuracy improves, but computational complexity increases
Solution Approach 1:
The patent performs preliminary data preprocessing, feature extraction, and information standardization before inputting data into the deep learning model, preparing the data in advance to reduce the computational burden during classification and improve model efficiency
Solution Approach 2:
The patent extracts key features and critical information from raw traceability data before classification, removing unnecessary information and focusing the deep learning model on the most relevant features, thereby reducing computational complexity while maintaining high classification accuracy
Data Source
AI summary
The present disclosure provides a method, a device and a computer readable storage medium for food risk traceability information classification. The method includes: building a deep learning neural networks model by a self-learning ability of an artificial intelligence model, initializing weights and a bias of the built model, and obtaining an original deep learning neural networks model; obtaining samples of food risk traceability information, dividing the samples, and obtaining factors of the food risk traceability information, converting the factors into vectors of the food risk traceability information; inputting the vectors into the original deep learning neural networks model, and obtaining original classification vectors of current food risk traceability information; and inputting the original classification vectors into a loss function, obtaining a loss rate of the original classification vectors, and determining the original classification vectors as a target classification result in response to the loss rate being within a preset range.