Arandom forest (RF) modelfor rock burst identification was established on the basis of the RF theory to forecast rock burst risk rapidly and accurately. Ten indices, ie, coal seam, dip angle, buried depth, structure situation, change of dip angle, change of coal thickness, gas concentration, roof management, pressure relief and shooting were used as the criterion indices for rock burst prediction in the proposed model on the basis of analysis of rock burst impact. Twenty-four typical rock burst instances of a coal mine were used to createa RF classifier. RF is a combination of tree predictors, and variable importance is measured by Gini importance (GI) when the forest grows. The GI shows that structure situation was the most important indicator, followed by shooting and dip angle. Another 12 groups of rock burst instances were tested as forecast samples, and the predicted results were in accordance with actual situation.

rock burst  /  random forest  /  variable importance