一般信息系统的PoClustering与概念格

传统聚类方法生成的子集,一般来说都是不相交的.而严格的不相交分类结构,不能充分表现象本体这样的事物间丰富的类关系.在基因本体中,类与子类既不是简单的树也不是格结构,而是一个有向非循环图,其任何子女都可能有多个父结点.PoClustering是相异数据的一种无损聚类方法,概念格则反映了数据的对象和属性的对应关系.采用了PoClustering方法,在保持尽量多的信息的前提下建立一般数据集(信息系统)的属性确定下的概念化分类,讨论了它的算法,从概念格的角度研究了这种类的结构特征.

PoClustering for a General Information System and Concept Lattice

Classical clustering algorithms generate disjoint subsets,such as graph-theoretical clustering,density- based clustering and k-means type clustering,etc.Since strict taxonomies form disjoint partitions,these structures are insufficient for capturing categorizations with richer relationships,such as ontologies.The class-subclass relationships in Gene Ontology are neither a simple tree,nor a lattice structure.Instead,it is a directed acyclic graph,where any child can have multiple parents.PoClustering is a lo...