Abstract: Real-time prediction of video source traffic is an important step in network resource management and end-to-end quality-of-service (QoS) strategies. However, together with the long-range-dependence (LRD) and the traffic non-stationarity, it suggests that conventional prediction tools, which only use short-range dependence (SRD), are not appropriate for VBR video traffic prediction. In this paper, by analyzing the correlation structure of multifractal coefficients, the original LRD trace can be converted to a series of SRD sequence in multifractal domain. Because the LRD feature of trace is used, the multi-step performance of proposed multifractal model is much better than traditional methods.