混合型频繁无序树挖掘算法  

Hybrid mining algorithm for frequent unordered trees

作　　者：李海波[1] 王元珍[1] 左琼[1] 

机构地区：[1]华中科技大学计算机科学与技术学院,湖北武汉430074

出　　处：《华中科技大学学报（自然科学版）》2009年第5期61-63,71,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

摘　　要：吸收深度优先和广度优先算法的优点，充分利用Apriori原则，提出了一种连接-扩展混合型的频繁无序树挖掘算法．首先在兄弟模式子树间进行连接，然后只在模式子树的最右叶顶点上进行扩展．分别用人工数据集和现实数据集进行测试，结果表明，本文算法比uFreqt算法性能提高0．5～1．0倍，比HybridTreeMiner算法性能提高5～10倍．After the advantages of depth-first algorithm and breadth-first one were analyzed, an algorithm for mining frequent subtrees in the database of unordered trees was proposed, in which joiningextending combined enumeration method was used to give full pay to Apriori rule＇s speciality. Sibling pattern trees was joined at first, and only extends on the right-most leaf node of a pattern tree. Proved by performance test on artificial and practical data set, the algorithm is 0.5 - 1.0 times faster than uFreqt algorithm, and 5-10 times faster than HybridTreeMiner algorithm.

关 键 词：数据挖掘 频繁无序树 算法 图 同构 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]