【作者】 吴丹； 鲍玉斌； 刘运涛； 于戈； 孙焕良； 王大玲；

【机构】 东北大学信息科学与工程学院；

【摘要】 <正>1引言在数据挖掘、文档分类和多媒体索引等新兴领域所处理的数据集规模巨大且复杂,表现为数据集的元组数目大、属性个数多,以及属性间关系复杂。这使得数据处理变得相当困难。人们除了需要发现数据集中具有代表性的数据和属性,还希望进一步发现属性之间的相关性。目前,已提出了许多更多还原

【Abstract】 Finding correlations between the attributes and grouping those attributes are challenge work when dealing with multimedia information.Former algorithm FD-ASE employs the fractal dimension as approximation of significance and takes two cycles to generate the results,however,it produces O(n~2) complexity of time.This paper presents a fast algorithm FACE for finding the correlations between attributes base on new definitions and efficient fractal dimension calculation.The algorithm defines two kinds of correlation-attribute groups,and employs the definition to accelerate indicating the most significant attributes and generating the groups of correlation attributes.Our contribution is reducing the complexity of finding groups.Experimental results on synthetic data sets demonstrate the advantages of our algorithm.更多还原