Predictive Modeling of Compressive Strength in Coconut Shell-Embedded Concrete Using Ensemble Regression Models

This study is based on the exploration of ensemble regression models using Adaboost Regressor, Random Forest Regressor, Gradient Boosting Regressor, and XGBoost Regressor for predicting the compressive strength of coconut shell concrete. In the study normalization technique, and 5-fold cross-validation for the training and testing of the models over a dataset of 483 samples are employed, constituting various concrete mix features. It produced the result showing competitive performance with the top performance being Adaboost Regressor that resulted in a score of 0.81 R2 and a minimal MSE of 1.16. With the help of SHAP, it has been analyzed and observed that influential pivotal effects have been outlined by influential features including Age, Coarse Aggregate, and Water content in the predictions. Age of concrete has emerged highest important factor that puts it in a leading role for determining compressive strength. The research itself gives insight into the optimal formulation of concrete, harmonizing between structural integrity and sustainability. Ensemble models, coupled with SHAP, give inspiring directions for advancing concrete technology in eco-friendly construction practices.