Machine learning study demonstrates accurate prediction of drilled hole surface roughness in hybrid sisal-cotton composites, indicating enhanced manufacturing quality control.
Key Points
To develop a super-learner ensemble machine learning framework capable of accurately predicting the surface roughness of drilled holes in hybrid sisal-cotton reinforced polyester composites.
Trained decision trees, random forests, gradient boosting, extreme gradient boosting, and a combined super-learner ensemble model.
Optimized model parameters using k-fold cross-validation with grid search, and evaluated feature importance using Shapley Additive Explanations (SHAP).
The super-learner model outperformed all individual algorithms, achieving a coefficient of determination (R²) of 99.4% on the test dataset.
The ensemble model minimized prediction error, recording a mean absolute error of 4.75%, a mean absolute percentage error of 3.30%, and a root mean square error of 5.32%.