小波滤波在时间序列频繁模式挖掘中的应用

与布尔型数据的频繁模式挖掘相比,时间序列的频繁模式挖掘是一个相对复杂的问题,目前对此类问题还缺少深入的研究.通过对小波滤波的研究,提出了一种时间序列的频繁模式挖掘算法,Frequent-Wavelet算法.该算法的特点是采用多孔平滑滤波器组对时间序列做低通平滑处理,用得到的多个尺度序列表示原序列,较好地解决了时间序列的平凡相似问题和时间轴伸缩问题.实验表明,Frequent-Wavelet算法对于时间序列的频繁模式挖掘具有较好的效果.

Application of wavelet filter in frequent time series patterns mining

Compared with frequent patterns mining of Boolean data,frequent time series patterns mining is more complex,and deep studies on this issue are rare. Based on the study of wavelet filter,an algorithm for frequent patterns mining of time series,named Frequent-wavelet,was proposed.The characteristic of this algorithm is to pass time series through the wavelet filter,a trous-smooth-filter,and represent the original time series using the obtained scale sequences.The algorithm can solve the problem of trivial similarity and time axis stretching of time series.The experiment shows that our algorithm is effective for frequent time series patterns mining.