Abstract: In order to predict rockburst risk quickly and accurately, a model was built for prediction of rock burst risk based on NRS-ACPSO-SVM. Before building the model, firstly, on the basis of a comprehensive analysis of influence factors of rockburst risk was made on the basis of the data on Chongqing Yanshitai Mine taken as an example,10 main indexes influencing rock burst such as coal thickness,coal seam dip angle,buried depth and so on were selected as the characteristic indicators of rock burst risk. Secondly, the NRS theory was used to reduce the dimensionality of characteristic indicators, the reduction set consisted of the key attributes affecting the rock burst risk were extracted. Finally, in order to avoid the random selection of SVM model parameters, ACPSO algorithm was used to optimize the SVM's parameters, the reduction set as the input into ACPSO-SVM model was trained, and trained ACPSO-SVM model was used to predict rockburst risk of testing samples, and a comparison was made between the model and other models in the prediction results. The results show that NRS can effectively reduce attributes, and simplify model's structure, the accuracy and efficiency of prediction model are improved; using ACPSO to optimize SVM model can make the results to avoid getting into the local extremum, and improve the convergence speed and prediction accuracy, the model can be used to effectively predict risk level of rock burst.

Key words: rockburst, risk prediction, neighborhood rough set(NRS), support vector machine(SVM), adaptive chaotic particle swarm optimization(ACPSO)