Integrating multi-dimensional quality profiling with explainable machine learning for discriminant modeling of Yimeng Black Goat meat

This study evaluated the meat quality of M. longissimus dorsi and M. semimembranosus from Yimeng Black Goats (BG) and Huanghuai White Goats (WG) using multi-dimensional parameters (pH, water-holding capacity, color, texture, electronic nose). Classification models (PLS-DA, SVM, BP-ANN) were developed and interpreted via SHAP analysis. Results showed BG had higher pH and water-holding capacity (P < 0.05), while WG exhibited higher lightness (L*), hardness, and stronger volatile flavor responses. SVM achieved the best performance (97.22% accuracy, 96.88% precision, 98.08% recall, 97.33% F1 score). SHAP interpretation identified post-mortem hardness at 45 min (T1-1) as the most discriminative feature, with breed- and muscle-specific contribution patterns. This study provided an intelligent framework for quality-based classification and targeted processing of goat meat.