Supervised data-driven approach to predict split tensile and flexural strength of concrete with marble waste powder

The utilization of marble waste powder (MWP) as a supplementary cementitious material in concrete, serving as a replacement for cement, holds the potential to enhance split tensile strength (STS) and flexural strength (FS), alongside offering environmental advantages. However, it is crucial to determine the optimal dosage of MWP, ensuring meticulous mix design and testing procedures to maximize the concrete's strength and overall performance. This research endeavor seeks to investigate a supervised data-driven approach for predicting STS and FS in concrete composites incorporating MWP, along with other cementitious materials such as silica fume (SF), granite powder (GP), and fly ash (FA), and their influence on the STS and FS of MWP-incorporated concrete. Ten distinct machine learning (ML) algorithms, including multivariate linear regression (MVLR), support vector regression (SVR), artificial neural networks (ANN), decision tree regressor (DT), random forest regressor (RF), adaptive boosting regressor (AdB), light gradient boosting machine (LGBM), gradient boosting regressor (GBR), extreme gradient boosting (XGB), and cat boost, are employed to assess the predictive capabilities of these models for FS and STS datasets. Statistical metrics like correlation coefficient (R2), root mean square error (RMSE), mean absolute error (MAE), and mean absolute percentage error (MAPE) are used to evaluate the performance of each ML algorithm. To enhance model efficiency, hyperparameter tuning and a 5-fold cross-validation technique are implemented. Among the ML algorithms tested, the cat boost algorithm demonstrates superior performance in predicting STS, while the ANN algorithm excels in predicting FS. Additionally, SHAP dependency plots are utilized to ascertain the feature importance in the best-performing models. The analysis reveals that features such as curing age, water, and cement play a more significant role in predicting STS, whereas attributes like cement, concrete type, and sand hold greater importance in predicting FS.