Machine learning-based prediction of concrete strength properties with coconut shell as partial aggregate replacement: A sustainable approach in construction engineering

Abstract

This study investigates the application of machine learning (ML) models to predict the compressive, flexural, and split tensile strength of concrete incorporating coconut shell as a partial replacement for coarse aggregate. This study utilizes a comprehensive dataset compiled from reputable literature, encompassing various experimental samples and input variables. Statistical analyses, including Pearson correlation and frequency distribution, lay the groundwork for preprocessing, involving standard scaling of features. Five prominent ML models, namely, support vector regression, random forest regression, gradient boosting regression, extreme gradient boosting regression, and multi-layer perceptron regression, are trained on the preprocessed dataset. The models' performances are rigorously evaluated using R², RMSE, MAE, MAPE, and Comprehensive Performance Index (CPI) metrics. The feature importance analysis unveils the critical role of variables such as the age of concrete, coarse aggregate, and water in shaping concrete strength. gradient boosting regression consistently emerges as the top-performing model. This study concludes with insights into the implications for sustainable construction practices and suggests future research directions, emphasizing the continual refinement of predictive models and on-site validation for real-world applications in construction engineering.