一种改进的最大频繁项集挖掘算法  

AN IMPROVED ALGORITHM FOR MINING MAXIMUM FREQUENT ITEM SETS

作　　者：胡德敏[1] 赵瑞可[1] 

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《计算机应用与软件》2012年第12期186-188,共3页Computer Applications and Software

摘　　要：研究基于条件模式基排序的最大频繁项集挖掘算法。通常在基于FP-tree(frequent pattern tree)的最大频繁项集挖掘算法中,影响执行效率的主要是递归和超集检测。因此提出了改进的最大频繁项集挖掘算法S-FP-MFI(sorted frequent pattern tree for maximal frequent item set),根据条件模式基含有的项目数对条件模式基进行动态排序,以减少递归次数;另外基于MFI-tree(maximalfrequent item tree)的投影策略减少了超集检测时间。实验表明S-FP-MFI算法在支持度较小的情况下,具有优越性。An algorithm of mining maximal frequent item sets based on conditional pattern base sorting is studied. In general, the main factors affecting the execution efficiency of maximal frequent item sets mining algorithms based on FP-tree （ frequent pattern tree） are the recursion and superset checking. Therefore this paper proposes an improved maximal frequent item sets mining algorithm S-FP-MFI （ sorted frequent pattern tree for maximal frequent item set）. According to the number of items of conditional pattern base, this algorithm sorts conditional pattern base in order to reduce recursion times, and the projection strategy adopting MFI-tree （maximal frequent item tree） also reduces the superset checking time as well. Experimental results testify the predominance of the proposed algorithm in condition of low support threshold.

关 键 词：递归 最大频繁项集 频繁模式树 条件模式基 超集检测 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]