Artificial intelligence-driven food quality prediction: Applying machine learning ensemble models for dynamic...

This study presents a food chemistry-driven approach to predict post-slaughter pork quality dynamics, focusing on the biochemical mechanisms governing pH evolution and meat color development over 48 h. The interconversion of myoglobin redox states and the cascade of glycolytic reactions fundamentally determine meat quality through complex chemical pathways. Using multi-source data from 1284 pigs across 24 breeds, we developed and optimized machine learning ensemble models, including LightGBM, XGBoost, and Random Forest, to capture these chemical transformations. Our ensemble learning methods achieved superior prediction accuracy, with R² values exceeding 0.7 for both pH and color attributes. Feature importance analysis revealed early pH values and muscle structural characteristics as critical predictors, aligning with key biochemical processes such as anaerobic glycolysis, protein denaturation kinetics, and myoglobin redox reactions. These insights underscore the intricate interplay between chemical mechanisms and sensory traits, providing a precise alternative to traditional evaluation methods. The chemical basis of pork quality prediction established in this study offers a scalable framework for integrating genetic, environmental, and processing factors, ultimately enhancing quality control precision and sustainability in meat production through the application of food chemistry principles.