一种新的快速挖掘频繁子树算法  

A new fast mining frequent subtree algorithm

作　　者：唐德权[1] 刘绪崇[1] TANG Dequan;LIU Xuchong(Department of Information Technology,Hunan Police Academy,Changsha 410138,China)

机构地区：[1]湖南警察学院信息技术(网监)系,湖南长沙410138

出　　处：《湘潭大学学报（自然科学版）》2022年第2期96-106,共11页Journal of Xiangtan University(Natural Science Edition)

基　　金：国家自然科学基金(61471169);湖南省科技重大专项(2017SK1040);湖南省教育厅重点项目(20A172)。

摘　　要：挖掘隐藏在大型标签数据集中丰富的语义信息是数据挖掘的重要任务之一.基于成千上万标签的半结构化数据集,提出了从给定包含一棵或多棵标签树的数据集中,找出所有满足用户最小支持度阈值频繁子树方法.首先采用树和森林的规范表示,使用扩展操作生成候选子树集,进一步提出有根有序标签树的挖掘算法.通过确定自由树中心,将自由树转换成有根有序标签树.该方法不仅解决了一般自由树规范化问题,而且能直接应用到半结构化数据集中.实验结果表明,该方法能够快速有效地从大型标签数据集中挖掘所有频繁子树.Mining the rich semantic information hidden in large label datasets is one of the important tasks in data mining.A method is proposed to find all frequent subtrees that meet the user’s minimum support threshold from a given dataset containing one or more label trees.The standard representation of row tree and forest is first used,the candidate subtree set is generated by extension operation,and the mining algorithm of rooted ordered label tree is further proposed.By determining the center of the free tree,the free tree is transformed into a rooted ordered label tree.This method not only solves the normalization problem of general free tree,but also the mining algorithm of rooted ordered label tree can be directly applied to semi-structured data set.The experimental results show that our proposed method can quickly and effectively mine all frequent subtreesin from large label datasets.

关 键 词：数据挖掘 标签数据集 频繁子树 有根有序树 自由树 

分 类 号：TP311.2[自动化与计算机技术—计算机软件与理论]