Representation and clustering of time series by means of segmentation based on PIPs detection

SAX is the representative time series representation method. SAX used the PAA technique to reduce the dimension of time series. But PAA technique has the demerit that cannot represent various movement shapes of time series exactly in lower dimensional space, since its smoothing effect distorts the dynamic characteristic of time series. Therefore, this paper suggests new representation method of time series using PIPs detection technique. The proposed method can represent various movement shapes of times series exactly than SAX. Because the PIP is the most important factor that determines the movement shapes of time series. The experimental result shows that the proposed time series representation is superior to SAX.