Traffic forecasting plays a significant role in Network Management as well as in Congestion Control and Network Security. Accurate traffic prediction based burst and unstable point can significantly improve network performance substantially while satisfying Quality of Service (QoS) requirements. In this paper, a new traffic forecasting method of Grey theory assembled with Chaos and SVR was presented (GCSVR).In this method, we employed the chaos theory to analysis the time series, adopted the Grey theory to smooth the series, make the series has a high regularity. In the experiment section, two models for short term forecast are examined: the original SVR and the GCSVR.Through the demonstration, the precision of forecasting by the GCSVR has a better performance than the original SVR.