关键词: 重要点分割，平滑处理，特征值，KNN算法

Abstract: Because the time series has a large amount of data, detecting it directly will has a high complexity. And this paper proposed an outlier sub-sectuences detection algorithm based on importance points segmentation of time series to relieve the problem. Outlier detection of sub-sectuences can offset the limitations of the outlier detection of points. This algorithm firstly obtains a series of smoothed important points, and then divides the sequences according to them, mcanwhile extracts the four characteristic values of each sub-sequence: length, height, mean and standard deviation, and applies these four characteristic values to the curopean euclidcan distance. Finally, it detects the outlier sulrsequences with the KNN algorithm Experimental results show that the algorithm is effective and reasonable.

Key words: Importance points segmentation, Smooth process, Characteristic values, KNN algorithm