Supervised machine learning–based prediction and sensitivity analysis of compressive strength in red-mud-based sustainable concrete

This study develops supervised machine learning models based on least squares support vector machine (LSSVM) and its hybrid variants integrated with particle swarm optimization (PSO), ant colony optimization (ACO) and grey wolf optimization (GWO) to predict CS of red-mud-based concrete. A comprehensive database of 198 experimental datasets including eight input variables and one output variable employed for model development and validation of models. Model performance was evaluated using statistical and error indices. The results indicate that hybrid optimization significantly enhances prediction accuracy, with the PSO-LSSVM model achieving the best performance (R2 = 0.944, RMSE = 0.047, MAE = 0.034), outperforming the standalone LSSVM model (R2 = 0.909, RMSE = 0.058, MAE = 0.045). Sobol sensitivity analysis reveals that fly ash content is the most influential parameter (0.583), followed by water (0.352), superplasticizer (0.277) and red mud (0.259). The findings demonstrate that hybrid LSSVM models provide a reliable and efficient approach for accurate CS prediction and mix optimization in sustainable red-mud-based concrete. The developed models were further integrated into a user-friendly web-based application to facilitate rapid and practical compressive strength prediction for engineers and researchers.