Abstract: In order to accurately predict rock burst disasters in coal mines, a prediction method based on particle swarm optimization algorithm (PSO) optimized least squares support vector machine (LSSVM), namely PSO-LSSVM method for rock burst classification prediction, was proposed. In this method, 10 indicators including mining depth, geological structure, coal hardness coefficient, maximum principal stress, angle variation of coal seam, variation of coal seam thickness, roof rock thickness, mining technology, rock strength of roof and floor were taken into consideration comprehensively to build the prediction index system of rock burst. The nuclear parameter σ and penalty factor f of the LSSVM model were optimized by using PSO search method, and then the optimized parameters were input into the LSSVM model. The classification prediction method of rock burst disasters based on PSO-LSSVM method was established, and the working face was forecasted by it. The results show that compared with other prediction methods, the PSO-LSSVM method has the characteristics of high computational efficiency, high accuracy and simple operation, and the field application effect is good.