Quantifying compressive strength in limestone powder incorporated concrete with incorporating various machine learning algorithms with SHAP analysis

The use of waste and recycled materials in concrete is one potential solution to lessen the impact of environmental problems from the concrete industry. The purpose of this work is to use machine learning algorithms to forecast and create an empirical formula for the compressive strength (CS) of limestone powder (LP) incorporated concrete. Eight distinct machine learning models—XGBoost, Gradient Boosting, Support Vector Regression, Linear Regression, Decision Tree, K-Nearest Neighbors, Bagging, and Adaptive Boosting—were trained and tested using a dataset that included 339 experimental data of varying mix proportions. The most significant factors were used as input parameters in the creation of LP-based concrete models, and these included cement, aggregate, water, super plasticizer, cement, and additional cementitious material. Several statistical measures, such as mean absolute error (MAE), coefficient of determination (R²), mean square error (MSE), root man square error (RMSE) and mean absolute percentage error (MAPE), were used to evaluate the models. XGBoost model outperforms the other models with R² values of 0.99 (training) and 0.89 (testing), with RMSE values between 0.065 and 4.557. To ascertain how the input parameters affected the outcome, SHAP analysis was done. It was demonstrated that superplasticizer, cement, and SCM significantly affected the CS of limestone powder concrete (LPC) with high SHAP values. By eliminating experimental procedures, reducing the demand for labor and resources, increasing time efficiency and offering insightful information for enhancing LPC design, this research advances the development of sustainable building materials using machine learning.