This study investigates the effectiveness of an XGB regression model combined with SHapley Additive exPlanations (SHAP) analysis for understanding surface roughness (Ra) and tool wear (VB) in high-speed milling of AA6061 aluminum alloy. Bayesian Optimization, a technique that efficiently explores hyperparameter space, is employed to fine-tune the Extreme Gradient Boosting (XGBoost) model for optimal performance. The model achieves high accuracy, with low Root Mean Squared Error (RMSE) and a high R2 value on the testing dataset. This demonstrates the model's capability to understand the intricate connections between cutting parameters (such as depth of cut, feed rate, cutting speed, and cutting time) and the resulting Ra and VB. SHAP analysis provides further insights, revealing the overall importance of each parameter and how they influence Ra and VB in individual milling experiments. This granular understanding empowers manufacturers to optimize high-speed milling processes for AA6061 by adapting cutting parameters based on their specific goals.