一种基于加权规则的显著模式挖掘算法  

SIGNIFICANT PATTERN MINING ALGORITHM BASED ON WEIGHTED RULES

作　　者：熊政[1,2] 王金明[2] 郑海雁[1,2] 李昆明[1] 徐立臻[2] 崇志宏[2] 

机构地区：[1]江苏方天电力技术有限公司智能电网产品中心,江苏南京211189 [2]东南大学计算机科学与工程学院,江苏南京211189

出　　处：《计算机应用与软件》2015年第11期53-56,112,共5页Computer Applications and Software

基　　金：国家自然科学基金项目(60973023)

摘　　要：在频繁项集挖掘过程中会发现事务或关系数据集中项目具有不同的重要性,而一些经典的频繁模式挖掘算法仅考虑项目频数这一属性来进行挖掘操作。针对该问题为不同的项目添加不同权重,提出一个新的加权规则模型,定义一种特殊的模式即显著模式。构造一棵类似于FPTree树的、具有高度压缩存储特性的数据结构树——SPTree(Significant Pattern Tree),之后基于SPTree树提出一个新颖的挖掘显著模式的算法DMSP(Data Mining Significant Pattern)。实验结果验证DMSP算法能够高效地挖掘显著模式。该算法可以有效解决由于项目重要性各不相同而导致的问题,有利于发现更多有研究价值的信息。In the process of mining frequent itemsets, it will find that the itemsets in transaction or relational data sets have different importance. However, some classic frequent pattern mining algorithms only consider the property of items frequency number to complete the mining operations. In light of this problem, we gave different weights to different item, and proposed a new weighted rule model, defined a special pattern which is named as the significant pattern. We constructed a new data structure tree named SPTree （significant pattern tree）, which is similar to FPTree but with a high degree of compressing storage characteristic, then based on SPTree we proposed a novel mining significant patterns algorithm DMSP （data mining significant pattern）. Experimental result verified that DMSP algorithm could efficiently mine the significant pattern. The conclusion was that the DMSP algorithm could effectively solve the problems caused by different importance of data items, and was conducive to find more information with research value.

关 键 词：SPTree树 DMSP算法 加权规则 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]