小波变换的离群时序数据挖掘分析

Analysis of Time Series Outlier Mining Based on Wavelet Transform

摘要: 针对时序数据进行离群数据挖掘方法的研究。通过对时序数据进行离散小波变换,将其从时域空间变换到频域空间,使时序数据映射为多维空间的点。该方法具有多尺度、时移不变性等特点,经离群时间序列进行离散小波变换后,不仅具有良好的保距性又达到降低维数目的。然后提出一种基于距离的离群时序数据挖掘算法。仿真试验表明了该方法的有效性。

Abstract: In this paper, the outlier mining method for time series data is investigated. DWT is used to transform the time series data from time domain to frequency domain. The time series data can be mapped into the multidimensional points in multidimensional space. We proposed a distanced-based algorithm to mine the outliers. The simulation results show the effectiveness of the method.