A novel piecewise linear segmentation for time series

Time series representation is one of the fundamental tasks in Time Series Data Mining (TSDM). Due to the advantage of easy understanding and implementation, Piecewise Linear Representation (PLR) has been widely used in compression, indexing, and similarity measurement of time series data. In this paper, we introduce a novel online PLR segmentation method. It is based on determining the cumulative radian error for each data point. The proposed method is demonstrated by applying to real stock market indices data and shows its effectiveness and superiority.