Comparison between Random forest and M5P to predict the compressive strength of concrete modified with solid wastes.

Abstract

The efficacy of two machine learning algorithms to predict the concrete compressive strength is investigated in this research. For this objective, a vast amount of experimental data from numerous academic research articles was statically evaluated and modelled. In all, 265 observations were considered in this investigation 70% of the data was used for training, while the remaining 30% was used for testing. The data used in this paper was divided randomly. Cement, ground granulated blast furnace slag, limestone powder, fly ash, rice husk ash, fine aggregate, coarse aggregate, silica fume, superplasticizer, water, viscosity modifying admixture, and coarse aggregate were among the 11 input parameters employed in the study. The concrete compressive strength was taken as the output parameter of the model. To assess the models' prediction abilities, correlation coefficient (CC), the mean absolute error (MAE) and root mean square error (RMSE) values were employed. From the results it was concluded that RF-based model outperforms the M5P Model. The CC, RMSE and MAE values for RF are 0.9867, 3.7887 and 2.9244 for training and 0.8668, 9.1723 and 5.8552 for testing stage respectively.