Abstract: Deep mining and underground space have been more comprehensively utilized in recent years,but it has led to increasingly frequent rock burst disasters. In order to predict the occurrence of rockburst more accurately,a principal component analysis method and an improved Bayes discriminant method are proposed. The prediction of rockburst is accomplished by selecting indexes such as rock uniaxial compressive strength,brittleness coefficient,rock elastic energy,etc. and establishing a comprehensive rockburst prediction model. When the principal component analysis method is used,the original data needs to be preprocessed to eliminate the correlation effect,and the main factors affecting the rockburst can be determined only after the Bayes discriminant condition is satisfied.The improved Bayes discriminant method is used to classify the rockburst intensity. This method adds a legal correction to the traditional Bayes model,which can solve the problems including that the area to which the indicators belong is uneven,the prior probability gap is large,and the posterior probability is located near the classification boundary,resulting in misjudgment. The research shows that the accuracy of the traditional Bayes discriminant prediction sample reaches 93.18%,and the accuracy and generalization ability of the model have been greatly improved after the improvement,which contributes to a new method and idea for rockburst research.