基于Markov链的Web访问序列 挖掘算法研究及性能分析

引入正向、逆向Markov一步状态转移概率矩阵构造序列数据库，并将逐层投影的PrefixSpan序列挖掘算法改为伪投影和隔层投影算法结合，以改进经典序列算法中存在的时间或空间开销太大的缺陷。性能分析表明，与经典算法相比，这种基于Markov链的Web访问序列模式挖掘新算法能够通过较少的计算量和空间复杂度获得较优的Web访问序列模式。

Research and Performance Analysis on Web Accessing Sequence Data Mining Algorithm Based on Markov's Chain

The transition probability matrix to building sequence database of forward and backward Markov one-step state is introduced. Using pseudo projection combined with bi-level projection algorithm instead of level-by-level projection, it can improve classical sequence mining algorithms that usually have much time and space complexity properties. Performance shows that the proposed one can decrease time and space overhead while searching for Web accessing sequence pattern when compared with classical algorithm.