Prediction of compressive strength of rice husk ash concrete based on stacking ensemble learning model

Abstract

By replacing cement in concrete production with rice husk ash (RHA), the amount of cement used and its environmental impact can be reduced. The objective of this study is to accurately determine the compressive strength of rice husk ash (RHA) concrete using a machine learning model. Stacking is an excellent fusion strategy. It uses meta-learner to better learn the prediction results of multiple base learners and improve the performance of the mode. In this research, a stacking ensemble learning-based compressive strength prediction model for rice husk ash (RHA) concrete is developed. The ensemble learning model is the first layer of the stacking model; the linear regression model is the second layer. The optimal configuration of base learners was experimentally determined, and the stacking model was contrasted with other mainstream methods. Using the base learner XGBoost model, the importance of the input feature variables was assessed. The findings reveal that the created stacking ensemble learning model can successfully fuse the prediction outputs of base learners and increase the predictive accuracy of the model. The performance evaluation indices of the established stacking model are as follows: RMSE = 2.344, MAE = 1.764, and R2 = 0. 987. The developed models were compared with previous studies and the model accuracy was better than previous studies. The developed model was applied to the new dataset and the model showed good performance. The cement and age are the two most important parameters impacting the compressive strength of rice husk ash (RHA) concrete.