Data-driven model for ternary-blend concrete compressive strength prediction using machine learning approach

Abstract

Ternary-blend concrete is a complex composite material, and the nonlinearity in its compressive strength behavior is unquestionable. Entirely many models have been developed to accurately predict the ternary-blend concrete compressive strength, such as ANN, SVM, random forest, decision tree, to mention but a few. This study underscores the better predictive performance and successful application of the least square support vector machine (LSSVM), a machine learning model for predicting the compressive strength of ternary-blend concrete. Coupled simulated annealing (CSA) was applied to the LSSVM model as an optimization algorithm. In addition, the genetic programming (GP) model was used as a benchmark model to compare the performance of the LSSVM-CSA model. The predictive performance of the LSSVM-CSA was compared with that of some of the proposed models in well-known studies where the same datasets were used. The model proposed in this study outperformed other studies, yielding an R2 value of 0.954.