融合趋势信息的时间序列符号聚合近似方法.

To solve the problem of losing trend information when representing time series with symbolic aggregate approximation method (SAX), this paper proposed a new time series symbolic aggregate approximation method integrating morphological trend information.Based on the maximum and minimum values in the subsequence and their relative positions, this method defined a new trend index to describe the trend information of the subsequence segments, and used the symbol vector integrating trend index to approximately represent the time series.For the proposed representation method, this paper gave a new distance metric and used it to conduct classification experiments on the UCR datasets and motor torque dataset.The experimental results show that the proposed method obtains higher classification accuracy than the SAX method on most datasets, and can effectively make up for the deficiency of losing local trend when the SAX method represents time series. [ABSTRACT FROM AUTHOR]