Abstract: Goat meat is favored by consumers for its high protein and low-fat content. The Yimeng Black Goat and Huanghuai White Goat, as local goat breeds in Shandong Province, have received less systematic research in terms of meat quality. This study focused on the longissimus dorsi muscle and semimembranosus muscle of these two breeds, measuring pH, water retention capacity, color, texture parameters, electronic nose data, and volatile flavor substances. Principal component analysis (PCA) was used to analyze the correlation of multiple indicators, and a backpropagation artificial neural network (BP-ANN) classification model was constructed. The Shapley Additive Explanations (SHAP) method was employed to identify key features, and partial least squares discriminant analysis was used to screen key flavor substances. The Mantel correlation analysis was combined to preliminarily reveal the correlation between meat quality and key flavor substances. The results demonstrated that, in comparison with Huanghuai white goats, Yimeng black goats exhibited lower cooking loss, higher a^* values, and greater concentrations of volatile flavor compounds. The BP-ANN model developed using 27 quality indicators achieved a classification accuracy of 97.5% for breed and meat cut in the test dataset, indicating strong discriminative performance. SHAP analysis identified seven key predictors influencing model output, including h value and S14 sensor response. Mantel correlation analysis revealed significant associations between these critical quality indicators and specific flavor compounds, namely n-butyl aldehyde, n-octanol, and ethyl acetate, suggesting that lipid oxidation derivatives may serve as pivotal mediators linking mutton's eating quality and flavor profile. These findings indicate that Yimeng black goats possess superior eating quality characteristics, and the established BP-ANN model offers a reliable methodological framework for future rapid, objective, and intelligent grading and discrimination of mutton.