Abstract: Accurate and effective grading prediction of rockburst is of great significance to reduce rockburst disasters.Maximum tangential stress of surrounding rock σθ，uniaxial compressive strength of rock σc，uniaxial tensile strength of rock σt and elastic deformation energy index of rock Wet are selected as the evaluation indexes of rockburst.Multidimensional normal cloud model is used as the prediction method of rockburst grade.Rough set is used to determine the weight of evaluation indexes.15 sets of rockburst examples at home and abroad are used as model construction samples，and the grading prediction model of rockburst based on rough set-multidimensional normal cloud is established.Firstly，fuzzy clustering analysis is carried on the evaluation samples，the evaluation indexes are filtered through the rough set，and the weight of evaluation indexes is determined.The determination degree of evaluation indexes is calculated according to the multidimensional normal forward cloud generator.Then，the comprehensive determination degree of evaluation samples is calculated by using the weight of evaluation indexes and the determination degree of evaluation indexes.According to the principle of maximum membership degree，the rockburst grade of evaluation samples is determined.The rockburst samples of Xinhua Mining Company are used to validate the grading prediction model of rockburst based on rough set-multidimensional normal cloud.The calculated results are consistent with the actual situation，which further indicates that the grading prediction model of rockburst based on rough set-multidimensional normal cloud has great practicality.

Key words: Rockburst, Grading prediction, Rough set, Multidimensional normal cloud, Comprehensive determination degree