挖掘最大频繁项集的改进蚁群算法

摘要

Mining Maximal Frequent Itemsets(MFI) is to find a maximal subset that appears frequently in datasets. There are many algorithms to effectively solve MFI.Ant Colony Optimization(ACO) is a new method to solve MFI.However,there are two bottlenecks: The ACO algorithm takes too much time and solves imprecisely for MFI.A novel ACO algorithm with max-min ant system and association graph is proposed. The tour graph is constructed. Ant colony constructs local maximal frequent itemsets under the instruction of dynamic pheromone and heuristic factor. It discovers global maximal frequent itemsets by new local and global update mechanism. Compared experiments show that this algorithm is fast and effective.%最大频繁项集挖掘用于发现频繁地出现在数据集中的最大子集,目前已经有许多有效的算法.应用蚁群算法挖掘最大频繁项集是一种新的方法,但是该算法往往迭代次数多,提取率低.结合频繁项集关联图和最大最小蚂蚁系统,提出一种新的蚁群算法.算法构造蚁群路径图,蚁群在动态的信息素和启发式因子指导下构造局部最大频繁项集,通过新的局部更新和全局更新机制发现全局最大频繁项集.时比实验表明,算法挖掘速度快,提取率高.