An explainable boosting-based ensemble machine learning model for predicting the properties of date palm fiber reinforced concrete

Natural fibers from date palm trees have been utilized in concrete, mortar and blocks to boost thermal insulation, ductility, energy absorption, and acoustical properties. but exhibits poor bonding with the cement paste and cause reduction in strength and durability of the concrete. For suitable application and acceptability of the date palm fibers (DPF) in concrete, methods to counteract its detrimental impacts on the concrete's performance needs to be developed. Hence, in this research, powdered activated carbon (PAC) due to its high reactivity was utilized as an additive to concrete containing DPF to mitigate its negative effects. The composite was made by adding different proportions of DPF between 1% and 3% of cement weight, and 1–3% PAC by cement. Machine learning (ML) technique was employed for predicting the DPF reinforced concrete properties for both time and cost savings. Five distinct ML algorithms were adopted to predict the hardened properties of sustainable concrete: Gradient Boosting, Adaptive boosting, Light Gradient Boosting Machine, Extreme Gradient Boosting, and Categorical Boosting. The XGBoost and CatBoost models outperformed the other algorithm models in terms of the prediction accuracy of hardened properties when compared using a testing dataset. It was observed that the XGBoost model outperformed others in terms of compressive and split tensile strength predictions with correlation coefficients of 0.992 and 0.985, respectively. Similarly, the CatBoost model showed better efficiency in flexural strength and Water Absorption prediction, with high correlation coefficients of 0.989 and 0.985, respectively.