基于邻接序列模式挖掘的网络流量分析

Network Trend Analysis based on Contiguous Sequential Mining Patterns

在网络流量模式挖掘中,发现邻接序列模式(CSP)是一个重要问题,为网络流量分析提出了一种新的树型数据结构。为了有效存储包含指定项的所有序列,该树组合了前缀树和后缀树,这种特殊的树结构确保了CSP检测的有效性。实验表明与已有方法相比,使用该结构不仅改进了CSP挖掘的时间性能,而且改进了空间性能。

Finding Contiguous Sequential Patterns （CSP） is an important problem in network trend mining. In this paper we propose a new data structure, UpDown Tree, for CSP mining. An UpDown Tree combines suffix tree and prefix tree for efficient storage of all the sequences that contain a given item. The special structure of UpDown Tree ensures efficient detection of CSPs. Experiments show that UpDown Tree improves CSP mining of both time and memory usage comparing to existing methods.