基于频繁矩阵的Apriori算法改进  

Improved apriori algorithm based on frequent matrix

作　　者：刘敏娴[1] 马强[2] 宁以风[1] 

机构地区：[1]江苏师范大学现代教育技术中心,江苏徐州221116 [2]徐州市政府经济信息中心,江苏徐州221006

出　　处：《计算机工程与设计》2012年第11期4235-4239,共5页Computer Engineering and Design

摘　　要：针对Apriori算法效率不高的问题,提出一种基于频繁模式矩阵的方法来挖掘最大频繁项目集。算法的基本思想是:只需扫描原始事务数据集一次,将事务数据转换成压缩矩阵,矩阵中保留了项目间的关联信息,同时只存放逻辑型数据,数据挖掘只采用逻辑运算,在挖掘过程中根据条件不断的对事务数据集和候选集进行剪枝,减少了不必要的开销。当数据量较大时,在效率上有一定的优势。实验结果表明改进后的算法具有良好的性能,提高了挖掘的速度。Aiming at the low efficiency problem in Apriori algorithm, a method on the basis of frequent pattern compressed matrix is presented to dig the biggest frequent items set. The basic idea of this algorithm is that original transaction data sets are scanned only once, and transaction data are transformed into a matrix, relevant information between the items is kept in matrix, meanwhile only logical data are stored. Data mining only adopts logical operation. In the process of mining , the transaction data set and candidate set are continuously pruned according to the conditions, unnecessary spending is reduced. When data quantity is bigger, there are a certain advantages in efficiency. Experiments prove that improved algorithm has good functions and increases the speed in mining.

关 键 词：频繁项集 矩阵 数据挖掘 关联规则 支持度 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]