An Algorithm for Similar Sub-patterns Discovery From Time Series
Jun Zhang · Journal of Nanjing Normal University · 2005
General method of similar sequence mining based on time series is to transform time series into discrete character series and cluster them into different sets, then compute the Euclidean distance between querying series and these sets to measure their similarity. These methods ignore the position and holistic characteristic of time series and work with high time complexity, according to which this paper proposes an algorithm of searching the key points which divides the time series into line segments. After checking the fitness of each line segments, we can quickly mine the similar sub-sequence with pattern distance measurement and quick pruning method.