Indirect Estimation of the Tensile Strength of Plastic-Infused Concrete using Gradient Boosting

Abstract

This paper applies gradient boosting (GB), a machine learning (ML) methodology for modeling the tensile strength (TS) of concrete made with waste plastic. Firstly, for development of GB models, the database including 235 data records was obtained from the existing studies. Following that, several GB models were developed by using the combination of different hyperparameters and their performance was validated through several statistical metrics. The optimum model achieved R² values of 0.9 and 0.89 for the training and testing datasets, respectively. The root mean square error (RMSE) was noted as 0.29 MPa for training and only marginally higher at 0.32 MPa in testing meanwhile mean absolute error (MAE) was found 0.25 MPa in training and 0.27 MPa in testing. These results demonstrate the capability of GB modeling in predicting TS of concrete.