基于PrefixSpan的序列模式挖掘改进算法  

Improved Algorithm for Sequential Pattern Mining Based on PrefixSpan

作　　者：汪林林[1,2] 范军[1] 

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065 [2]重庆工学院,重庆400050

出　　处：《计算机工程》2009年第23期56-58,61,共4页Computer Engineering

摘　　要：针对序列模式挖掘算法PrefixSpan在挖掘过程中需要构造大量投影数据库的不足,提出IPMSP算法,在递归挖掘过程中,通过检查序列数据库关于前缀的前缀,避免对同一频繁前缀模式构造重复投影数据库,同时舍弃对非频繁项的存储并在投影序列数小于最小支持度时停止扫描投影数据库,从而提高PrefixSpan算法的时空性能。实验结果证明,IPMSP算法在时间和空间性能上优于PrefixSpan算法。Aiming at the PrefixSpan algorithm produce huge amount of project databases in mining sequence patterns, this paper proposes an Improved PrefixSpan algorithm for Mining Sequential Patterns（IPMSP） alaorithm. By avoid produce duplicated project databases with the same prefix pattern through checking the prefix with regard to prefix of the sequence database and abnegating the non-frequent items and project databases which sequential number is lower than minimum support in the recursive mining process, the performance of Pref＇ixSpan is well improved. Experiment results shows that the time and space performance of IPMSP algorithm are better than that of PrefixSpan.

关 键 词：序列模式 PREFIXSPAN算法 投影数据库 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]