Authors: Yassine dahbi, Hamza Naciri, Hamza Zaouri, Ouahib Alaoui
Abstract: This study examines the use of GradientBoostingRegressor, StackingRegressor, and Gradient Boosting Regression with HistGradientBoosting in developing models that predict the compressive strength (fcu) and splitting tensile strength (fsp) of steel fiber-reinforced recycled aggregate concrete (SFR-RAC). The information comprises 465 compressive strength and 339 splitting tensile strength data of concrete mixes with varied ratios. Training and model testing were performed using 80/20 split with PSO for the hyperparameter optimization. The performance of the model was measured with four statistical metrics: coefficient of determination (R²), mean absolute error (MAE), root mean squared error (RMSE), and mean absolute percentage error (MAPE). Out of the models, Gradient Boosting Regression with HistGradientBoosting performed better in terms of prediction, with StackingRegressor taking the second rank. SHapley Additive exPlanations (SHAP) and feature importance were employed to determine the influence of input parameters on model predictions. From the results obtained, it was evident that the water content, cement content, and fiber ratio influence considerably the strength of SFR-RAC. The models give good insights regarding SFR-RAC mixture behavior, which is helpful in the production of environmentally friendly concrete with greater enhanced strength. Future research can enhance the data and use other predictor variables to further support these models.
DOI: https://doi.org/10.5281/zenodo.16437728
