Compressive strength evaluation of ultra-great workability concrete by creating hybridized regression analysis

Abstract

In the present article, the applicability of proposing hybridized regression evaluation on the ultra-great workability concrete ( UGWC ) was assessed with the goal of declining the assigned time and experimental works. UGWC can be included in several by-products or physical elements. For proposing suggested models, several hybridized regressions were produced, in which the best values of principal attributes of support vector regression (SVR) were clarified by metaheuristic methods such as sine cosine algorithm (SCA), Imperialist competitive algorithm (ICA), called ICA-SVR, and SCA-SVR. The criteria supply that both hybridized ICA-SVR and SCA-SVR models conclude justifiable ability in estimation outline. Taking into account integrated models, the ICA-SVR results in the weakest accuracy compared to SCA-SVR, with R 2 and VAF accounting for 0.9098, and 90.66%, respectively. The highest and most incredible performance owned by the SVR model hybridized with the SCA algorithm, where it was able to receive the largest value of R 2 and VAF in both training and testing datasets, and the smallest value of RMSE , and MAE in both training and testing datasets. MAE value in the training and testing phase for the ICA-SVR model is roughly fourfold and six times than SCA-ANFIS, which depicts the wonderful performance of SCA-SVR in the estimation system. So, the hybridized SCA-SVR can receive the highest precision in comparison with ICA-SVR as well as the published essay ( R 2 =0.801).