A Machine Learning Framework for Meat Safety Risk Assessment with Cross-Domain Data Fusion

This is the authors' abstract. We don't add key points for this paper.

The escalating complexity of global meat supply chains and rising consumption levels heighten public health risks, particularly in high-volume regions. To address this, we developed a novel risk assessment framework integrating multisource heterogeneous data from seven key origins. The framework implements LLM-enhanced data governance, a Delphi-AHP dual-dimensional indicator system, and an optimized Random Forest model, achieving predictive accuracy between 88.25% and 99.24%. Analysis identified pork as the highest-risk meat category, fresh meat as the riskiest processing type, and a shift since 2022 toward biological hazards, especially pathogens, as the dominant risk over chemical contaminants. This data-driven tool enables proactive, targeted risk management and supports regulatory decision-making for enhanced meat safety.