摘要
为了实现对数据流的序列模式挖掘,提出了基于数据流的序列模式挖掘算法MFSDS-1和MFSDS-2,它们均通过调整入选度的大小来调整保存信息的粒度.算法MFSDS-2利用分层存储结构,不仅能更好地保存序列信息,而且可以通过与全局序列模式的对比得到当前活动的一些异常序列模式.实验结果表明,基于分层存储的算法MFSDS-2的效率比算法MSFDS-1高.
Although sequence pattern mining has been deeply studied, it is a challenge to extend to data streams. In the paper, algorithm MFSDS-1 and MFSDS-2 are presented for mining sequential patterns from data stream. Both of the algorithms use the storage of elected sequences which are accommodated by the elected rate. In algorithm MFSDS-2 a new structure based on levels is proposed, which not only can store the sequence well, but also can be used to find the abnormal sequence. The experiments results show that MFSDS- 2 is more efficient than MFSDS-1.
出处
《江南大学学报（自然科学版）》
CAS
2007年第6期763-768,共6页
Joural of Jiangnan University (Natural Science Edition)
基金
江苏省自然科学基金项目(BK2005135)
关键词
序列模式挖掘
频繁序列
数据流
sequence pattern mining
frequent sequenee
data stream
