Another direction of the researches is the segmenting of the time-series data,
Reference [8] suggested dividing timeseries sequence into meaningful sub-sequence and in [9] a fixed length window was used to segment time series into subsequences and a time series was then represented by the primitive shape patterns that were formed. Because the window length became one of the most important factors affecting the segmenting performance here, Reference [10] proposed a conception of flexible window length. And in [12] a genetic algorithm was involved in segmenting the stock time-series and in [11] a perceptual important point
locating method was introduced in detecting patterns from the segmenting results