Extra Trees Ensemble: A Machine Learning Model for Predicting Blast-Induced Ground Vibration Based on the Bagging and Sibling of Random Forest Algorithm

Abstract

In this paper, the extra trees ensemble (ETE) technique was introduced to predict blast-induced ground vibration in open pit mines. It was developed based on the extension of random forest (RF) algorithm to bagging and sibling the predictors. Accordingly, the ETE used a simple algorithm to construct the decision trees (DTs) models as the predictors. Next, it combines the constructed predictors to achieve as-good performance in predicting blast-induced ground vibration. Herein, more than 300 blasting events were implemented and their parameters, as well as the intensity of blast-induced ground vibration, were measured and collected for this study. The ETE model was then developed based on the collected dataset for predicting blast-induced ground vibration. In addition, the RF model was also applied to compare with the ETE model. The results showed that the ETE model is superior to the RF model in predicting blast-induced ground vibration. Its performance and accuracy are outstanding and should be used in practical engineering to control the adverse effects of blast-induced ground vibration in open pit mines.