分布式环境下全局模糊频繁项目集的快速挖掘

为了解决模糊关联规则挖掘算法需要用户事先给定模糊集和相应隶属度函数的问题,提出基于分布式聚类自动生成模糊集及隶属度函数的算法GFAM.该算法利用分布式K-Means聚类算法对每个数值型属性进行聚类,求得聚类中心,由此构造全局模糊集,定义全局隶属度函数.DFAR算法根据构造的全局模糊集及隶属度函数进行分布式模糊频繁项目集的快速挖掘,采用全局-局部站点模式,其中包括局部模糊频繁项目集产生算法FLF和全局模糊频繁项目集产生算法FGF.实验结果表明,该算法能准确地生成全局模糊频繁项目集,在求解全局模糊频繁项目集过程中,传送局部模糊候选项目集支持数的通信量为O(n),提高了算法的挖掘效率.

Fast Algorithms for Mining Global Fuzzy Frequent Itemsets in Distributed Systems

Most of the algorithms proposed for mining fuzzy association rules assume that fuzzy sets and their membership functions are given.The paper proposes a method GFAM for automatically generating fuzzy sets and their corresponding membership functions based on distributed clustering.The paper also introduces an algorithm DFAR for mining f global fuzzy frequent itemsets in distributed systems.For this purpose,firstly,an efficient distributed clustering algorithm K-Means is used to group the data of each attribute and compute medoid of each cluster.Secondly,global fuzzy sets of quantitative attributes and their corresponding membership functions are generated by the medoids.Finally,the algorithm DFAR,which includes algorithms FLF and FGF,is proposed to mine fuzzy frequent itemsets in distributed databases.The experiments show the method can mine global fuzzy frequent itemsets precisely and requires only O(n) messages for computing support count of candidate fuzzy itemsets.