Adaptive time series segmentation algorithm based on trend turning points and state changes

Time series segmentation is a key problem in time series data mining, which directly affects the accuracy and efficiency of data mining. However, some existing time series segmentation methods not only need to set some important parameters manually without prior knowledge but also need to traverse all the data in the time series, resulting in high computational complexity. To solve these problems, an adaptive time series segmentation algorithm based on trend turning points and state changes (ATS-TSC) is proposed, which combines important trend turning points and state change index to optimize the segmentation results from locality to globality, improving segmentation accuracy. Moreover, it can adaptively realize segmentation without setting any parameters and has strong interpretability. The experimental results on UCR datasets show that the proposed segmentation methods are superior to the existing methods.