<p>The use of statistical and empirical models for the prediction of surface roughness always lacks generalizability and is not reliable when applied to unseen datasets. To increase generalizability and accuracy in predicting responses, researchers developed artificial intelligence and machine learning. The objective of this research is to develop advanced machine learning to predict surface roughness. For instance, the decision tree (DT), adaptive boosting (ADB), gradient boosting (GB), and extreme gradient boosting (XGB)s are used to predict surface roughness. The extreme gradient boosting (XGB) outperforms the other models by scoring the highest regression coefficient (<i>R</i>² = 98.7%) and least mean absolute error (MAE = 7.4%), least mean absolute percentage error (MAPE = 3.58%), and least root mean square error (RMSE = 8.0%). Further, based on the SHAP, the coolants have higher effects on surface roughness followed by cutting speed and feed rates respectively, and the tool overhangs have the least effects. Increasing cutting speed and coolant, decreases surface roughness while increasing feed rate and tool overhangs decreases surface roughness. The extreme gradient boosting (XGB) has highest generalizability and can be applied to every section of industrial and manufacturing sector for prediction of surface roughness; thus, it can save time and resource.</p>