Abstract: During the establishment of the support vector regression（SVR) prediction model of infiltration route data in tailing dam，the prediction accuracy and the computational time are two difficult problems to control，which severely constrain the extensive application of the SVR model.To solve these two problems，the rough set（RS) algorithm is adopted to reduce the attribute for training samples，then the prediction model of infiltration route is established with the SVR algorithmThe cases proved that RS-SVR model effectively lowered the difficulty of SVR model iteration，and made the prediction accuracy improved.It can be seen that the RS-SVR is feasible not only in theory but in application.

Key words: Infiltration route, Rough set, Reduction, Support vector regression