A hybrid important points identification for time series: Financial case

Important points identification is the key of the piecewise linear segmentation for time series. However, nearly all existing approaches are always perceptually important points (PIPs) focused while neglecting the domain related important points (DIPs) which might be of great interests to the domain experts. In order to preserve more important information relating to the particular domain after segmentation, a hybrid method to identify important points from both perceptual and domain perspectives is presented. We show the validity and effectiveness of the proposed method via a financial case.