Soft computing in estimating the compressive strength for high-performance concrete via concrete composition appraisal

This study investigates the predictive performance of Concrete Compressive Strength (CCS) for high-performance, based on concrete mixture constituents and proportioning. A new ensemble computational technique - Boosting Smooth Transition regression trees (BooST), is adopted and compared with other contemporary methods for higher predictive accuracy and analyses. With variations in CCS performances due to complexities in concrete compositions, ten unique models are created and divided into three sets from which several analytic techniques are employed to predict CCS at 28 days for high-performance. The results showed that BooST dominance in prediction accuracy over the other methods with minimal errors and better fit to experimental laboratory results. (C) 2020 Elsevier Ltd. All rights reserved.