Research on Pattern Recognition of Agricultural Product Safety Incidents and Intelligent Tracing of Traceability Paths Based on Artificial Intelligence

Ensuring the safety of agricultural products has become an urgent global concern, especially in the context of increasingly complex supply chains and frequent contamination incidents. However, most existing traceability approaches remain fragmented and static, focusing on isolated functions such as event logging or basic product tracking, and lacking an integrated framework for dynamic path reconstruction and causal reasoning. To address these limitations, this study proposes a unified AI-powered framework that integrates pattern recognition, intelligent path tracing, and knowledge graph-based reasoning to enhance both detection accuracy and traceability depth for agricultural safety incidents. The proposed framework employs machine learning models-including Long Short-Term Memory (LSTM) and XGBoost-for risk pattern recognition, while incorporating intelligent path tracing algorithms such as <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{A}^{*}$</tex> search and reinforcement learning (RL) to dynamically reconstruct contamination pathways. Furthermore, a domain-specific knowledge graph is developed to support causal chain analysis and interpretable decision-making. Experimental results on a multi-source dataset demonstrate the system's effectiveness: the LSTM model achieved an F1-score of 0.788 in identifying high-risk batches, and the RL-based tracing agent reached an F1score of 0.87 in reconstructing contamination paths. These findings validate the framework's potential as a scalable and interpretable solution for intelligent agricultural product traceability and risk management.