挖掘关联规则的蚁群算法  

Ant Colony Algorithm for Mining Association Rules

作　　者：屠莉[1] 陈崚 

机构地区：[1]南京航空航天大学信息科学与技术学院 [2]扬州大学计算机系

出　　处：《南京邮电大学学报（自然科学版）》2006年第5期36-40,共5页Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition

基　　金：国家自然科学基金(60473012);国家科技攻关项目(2003BA614A-14);江苏省自然科学基金(BK20005047);南京大学软件新技术国家重点实验室开放基金资助项目

摘　　要：基于蚂蚁的进化算法已经被广泛地应用于各种组合优化问题。首次结合蚁群算法提出了关联规则的蚁群挖掘算法(RA3)。利用数据集中属性和属性值对应超顶点和子顶点而构成无向图。蚂蚁对在无向图中不断地分别搜索挖掘频繁项集合,从而构成双规则前件、后件,并根据规则质量对双规则加以取舍。提出的算法可以在标准测试集中有效地实现关联规则的提取。与经典的Apriori和FP-growth算法比较,模拟实验证明,本文算法可以较快的挖掘出关联规则,具有较好的规则质量。Ant-based evolutional algorithms has been widely applied to kinds of combinatorial optimization problems. In this paper, we first present an ant colony algorithm for association rule discovery（RA^3）. A digraph is constructed where attributes and attribute values correspondingly represent the super-vertices and sub-vertices. Ant couples separately search and excavate the frequent items from the digraph which could be constructed as the double rules＇ antecedents or rules＇ consequents. Then, the double rules could be selected or removed by their qualities. This algorithm could extract association rules efficiently on several standard data bases. Compared with classical Apriori and FP-growth algorithms, the results indicate that our algorithm is able to discover association rules quickly with better accuracy than other methods.

关 键 词：关联规则 蚁群算法 无向图 蚂蚁对 

分 类 号：TP311.11[自动化与计算机技术—计算机软件与理论]