APPLICATION OF WEIGHTED DISTANCE DISCRIMINANT ANALYSIS METHOD BASED ON ROUGH SET TO ROCKBURST PREDICTION

<正>Rockburst was one of the main engineering geological hazards at deep high terrestrial stress zone and was influenced by many factors.Mahalanobis distance method was a new rockburst prediction method,in which the importance of each factor was thought to be equal.But the importance of each factor was different in fact. Therefore it was necessary to weight each factor.It was presented the rockburst prediction method of weighted Mahalanobis distance method could be developed on the basis of rough set.In this paper,a method of determining weighting coefficient was proposed based on rough set theory.Determining weighting coefficient was translated into estimating significance of attributes among rough set.According to the analysis of main causes of rockburst,the compressive strength,tensile strength,elastic energy index of rock and the maximum tangential stress of the tunnel wall were chosen as the criterion indices.By analyzing the significance of conditional attribute set for decision attribute,the weighing coefficient of each factor was determined by rough set.Finally,the model was applied to predict rockbrust in practical engineering.The results shows that the method is reliable and it can improve the evaluating accuracy of Mahalanobis distance method.