基于自相关获取周期的时间序列模式挖掘算法

Time Sequence Mining Algorithm Based on Autocorrelated Period Obtaining

文章针对局域网通话流量监测数据,提出了一种新的基于时间序列自相关的关联规则挖掘方法。该方法通过对数据进行平滑处理、求自相关函数,从而确定信号的周期;通过动态阈值对时间序列进行压缩处理提取流量趋势序列,提出DS-MMFI算法完成序列模式的挖掘。结果表明此算法能够有效去除时间序列的波动,且序列模式挖掘结果符合实际通信特征。

The article is based on the actual data of a local area network call traffic, it proposes a new association rule mining algorithm based on time-series by autocorrelation. The method determines the signal cycle by smoothing and finding the autocorrelation function. To get traffic trends sequence, it uses the dynamic threshold for the compression of time-series. It completes sequential pattern mining by using DS-MMFI method. The results show that the method can effectively remove the volatile of time-series, and the sequence mining results are in accordance with realistic communication features.