In order to accurately predict the pollution of transmission line insulators in complex environment and to achieve warning of pollution flashover, a method for identification of pollution characteristics of transmission line insulator and pollution prediction based on data driven is proposed.Combined with improved rough set and sample weighting method, a comprehensive quantitative model of insulator pollution characteristics based on PSO-BP neural network, and important characteristics affecting pollution can be quantified.On the basis of identification, a weighted support vector machine based on improved rough set is constructed to predict the pollution of the insulator and identify the risk of pollution flashover.The results show that the method is completely based on data driving and the charac-teristics of insulators with different operating environments and complex data types can be accurately identified.Compared with other methods, the proposed pollution prediction and risk identification method is more accurate and has smaller error due to the importance of the characteristics.The method proposed has good application prospects.