The task of complex time series predicting is hard to be accomplished with only one single predicting model.In this paper

a novel multi-scale incremental predictor is proposed.This predictor decomposes the complex time series into a series of intrinsic mode functions(IMF) and a residual signal with empirical mode decompo sition firstly

and then an Incremental Independent Vector Combination Predicting algorithm in Kernel Space(ⅡVCPKS) is constructed for predicting every IMF or residual signal.The propo sed predictor is competent for predicting the complex time series in real time.Experimental results showed that the propo sed method performed very well in the task of predicting complex time series.