【作者】 谢坤武； 陈世强；

【Author】 Xie Kunwu and Chen Shiqiang (School of Information Engineering,Hubei Institute for Nationalities,Enshi 445000)

【机构】 湖北民族学院信息工程学院；

【摘要】 现有的针对分类数据的算法需要多次扫描数据库,对于数据开采经常处理的大容量数据,多遍I/O操作是一项沉重的系统开销.CACD(clustering algorithm for categorical data)是针对分类属性数据的聚类算法,该算法采用压缩技术缩小需要处理的数据量以提高效率,同时算法提出了一种新的基于压缩数据结构的标准用于衡量分类数据的相似度.CACD只需扫描数据库一遍,算法理论分析和实验分析都表明该算法比同类针对分类数据的聚类算法效率要高,并且压缩技术对聚类结果的质量影响不大.更多还原

【Abstract】 Existing algorithms for clustering categorical data require several passes over the databases,and obviously the role of I/O overhead is significant for very large databases.CACD(clustering algorithm for categorical data) is an algorithm for clustering categorical data.It improves the efficiency by using compression technology to reduce the size of original data sets,and it also presents a new similarity criterion for compressed categorical data.CACD only needs to scan the database one time.The experimental results show that CACD performs better than other algorithms for clustering categorical data.The compressing technology does not affect too much the quality of clustering result.更多还原