The efficient transportation of real-time variable bit rate (VBR) video traffic in high-speed networks is currently an active area of research. The capability to predict VBR video traffic can significantly improve the effectiveness of numerous management tasks, including dynamic bandwidth allocation and congestion control. This paper proposes an adaptive traffic prediction method for VBR MPEG videos, a major multimedia application. Rapid traffic variations due to scene changes are analyzed, then a prediction scheme using the identification of scene changes related to I and P frames is presented. For predicting multiplexed MPEG traffic, a prediction interval is derived that represents a highly correlated traffic sequence. In addition, to reduce the prediction error, a less fluctuating signal instead of the original multiplexed traffic is used as the input for the predictor. Simulation results show that the proposed method is able to predict the original traffic more accurately than the conventional LMS method.