独立于应用的数据库聚类技术

Application-independent databases clustering technology

文章提出了一种独立于应用的数据库聚类技术,是多数据库挖掘的重要步骤,处于数据准备阶段,也是分组规则合成的前提,该技术主要包括多数据库最优划分方法,该方法将数据库的属性集当作其特征。数据库最优划分方法采用非对称二元变量相似度计算方法得到数据库间相似度,利用分裂层次聚类法对数据库进行完全划分,然后借鉴k中心点方法提出最大树方法选出对应簇中心,最后利用自适应模糊C-均值聚类方法的评价函数获得最优划分。

A kind of application-independent databases clustering technology is presented, which is an important step of the multi-database mining process in the data preparation phase, and is the require- ment of the group-rule synthesis. The technology mainly includes a best partitioning approach of databases which takes the attributes of a database as its characteristic. Through the best partitioning approach of databases, the asymmetric binary variable similarity computing method is applied to obtaining the similarity between two databases, and the divisive hierarchical clustering method is utilized to completely divide all given databases. Then the maximum-tree method derived from k-medoids is used to select the corresponding centers, and the score function of adaptive fuzzy C-means clustering （AFCMC） is used to find the best partition.