Ensemble learning models to predict the compressive strength of geopolymer concrete: a comparative study for geopolymer composition design

Currently, single machine learning models are mostly used for predicting the compressive strength of geopolymer concrete, but the use of single models has limitations. This study proposes the use of an integrated model to predict the compressive strength of geopolymer concrete. However, there are few applications of ensemble learning model and lack of model optimization. In this study, an improved beetle antennae search (IBAS) algorithm was proposed to tune the hyperparameters of decision tree (DT). random forest (RF), and K-nearest neighbor (KNN) models to predict the compressive strength of geo-polymer concrete. The focus of this paper is to compare the reliability and efficiency of IBAS algorithm applied to three integrated learning models for the prediction of geopolymer concrete compressive strength. The test results show that the corresponding R values are 0.9043, 0.6866, 0.9024, respectively. Therefore, it can be judged that the DT-IBAS integrated model has the worst prediction effect in these three models. In addition, the minimum RMSE values obtained by RF-IBAS and KNN-IBAS models in the ten-fold verification were 5.9 and 7.1, respectively. Therefore, RF-IBAS has the best predictive performance in comparison. On the other hand, the molar concentration of NaOH is the most important factor affecting the compressive strength of geopolymer concrete. Through the importance score test, the importance score of NaOH molar concentration (4.2981) far exceeds that of other input variables. Therefore, it is necessary to focus on the molar concentration of NaOH when making geopolymer concrete.