STOCHASTIC OPTIMIZATION OF EXPANDED POLYSTYRENE SIZE AND PROPORTION IN CONCRETE USING MACHINE LEARNING

Abstract

This paper presents a study on the prediction of the impact of incorporating EPS into concrete using ML algorithms. A dataset of 125 samples from reputable international journals was analyzed using Python tools. The primary objective of this study is to develop a reliable algorithm to optimize the size and proportions of EPS beads in concrete mixes without compromising 28-day compressive strength. Gradient Boosting Regression, Random Forest Regression, and XGBoost models were applied and evaluated using MAE, MSE, RMSE, and R² score to select the best predictive model. Based on this newly developed algorithm, an experimental test was conducted, and the results closely matched the predicted values with only slight differences. The study recommends further experimental investigations to understand the interaction between EPS beads and other concrete ingredients. It also emphasizes the importance of multidisciplinary approaches and mitigation strategies for better integration of EPS in concrete design.