Computational intelligence enhances compressive strength estimation of sustainable concrete using industrial waste

Abstract The promising results exhibited by industrial waste, notably fly ash (FA) and blast furnace slag (BFS), position them as potential supplementary materials for reducing cement usage in concrete to reduce environmental impact. This study presents a novel stacking‐based approach to enhance the accuracy of compressive strength (CS). The stacking model integrates ensemble learning methods, including Random Forest (RF), Gradient Boosting (GBR), Extreme Gradient Boosting (XGB), and a Bi‐directional Long Short‐Term Memory (Bi‐LSTM) as the base model, and a Catboost regressor as the meta‐estimator. The dataset is subjected to the grid search method and 10‐fold cross‐validation to assess the model for all base models, resulting in an R 2 of over 0.9 and R 2 of 0.9676 in the stacking model. The study utilizes Shapley Additive Explanations (SHAP) analysis to enhance model explainability, revealing how features like cement, BFS, and FA interact to influence CS. Further, SHAP interaction plots confirm that BFS in the 200–350 kg/m 3 range, FA in the 180–200 kg/m 3 , and SP of 20–30 kg/m 3 can be ideal for developing sustainable concrete. Additionally, the research highlights that concrete age, up to 200 days, correlates with increased CS. A composition‐based relationship between the input features, mainly industrial waste, and the target features is explained using the reverse design method, which relies on SHAP results. These findings suggest that the stacking model outperformed all employed base models, providing a comprehensive and robust methodology for adopting sustainable construction practices.