Predicting rockburst damage scale in seismically active mines using a classifier ensemble approach

Abstract

Rockburst still remains one of the most important sources of hazards in underground mine industry. This phenomenon results in many accidents and casualties in many underground mine projects around the world. Despite the scientific achievements in rock mechanics and engineering, the accurate prediction of rockburst damage potential is still challenging due to the complexity of seismic event occurrence. Hence, this paper aims to develop a reliable classifier ensemble to rockburst intensity in underground mine excavations subjected to seismicity. High quality rockburst database consisting of 254 case histories was used for the study. The classifier ensemble was developed through aggregation of several commonly used machine learning classifiers using the weighted voting. The performance of the classifier ensemble was evaluated using several indicators, namely: accuracy, recall, precision, and F1-score. The overall results indicate that the proposed classifier ensemble achieved good performance metrics and outperformed some existing empirical methods. It is concluded that the classifier ensemble could assist engineers to properly assess rockburst damage and contribute to selecting adequate ground control measures.