摘要： 用浮动搜索算法对时间序列进行特征选择得到低维特征参数，采用WSTB 方法实现对高维时序的相似性搜索。首先用浮动搜索算法对高维时间序列降维处理，得到特征参数后进行样本线性分段，建立时序曲线箱和相应索引。其次对样本序列和相似距离进行快速计算，不用逐个检查子序列箱的内容就进行快速索引。最后还验证了该方法的通用性和有效性。

关键词: 特征选择；线性分段；相似性搜索；时间序列

Abstract: In this paper, a WSTB-based algorithm for high-dimensional similarity search is proposed based on the float searching algorithm which is used to select features. The float searching algorithm to reduce the dimension of time series is used to get the piecewise linear features of the sample. When the subsequence bin for time series and the index of the bin is built, the sample series is calculated with the similarity distance. Quick index can be realized without checking the content of the bin, because the calculation gotten from comparison one by one is avoided. At last,the currency and efficiency of the algorithm are proved.

Key words: Feature selection; Piecewise linear representation; Similarity search; Time series