Incremental approaches to updating reducts under dynamic covering granularity

摘要

In real-world situations, knowledge acquisition of dynamic covering decision information systems(DCDISs) under variations of object sets, covering sets and covering granularity is an important research topic of covering-based rough set theory. In this paper, firstly, we introduce the concepts of the refining and coarsening coverings when revising attribute value sets and investigate the updating mechanisms of related families in DCDISs with dynamic covering granularity. Meanwhile, we discuss the relationship between reducts of the original covering decision information systems(OCDISs) and those of DCDISs and provide the incremental algorithms for updating reducts by making full use of the existing results from OCDISs. Finally, we perform the experiment on eight data sets downloaded from UCI Machine Learning Repository, which verifies that the proposed algorithms achieve better performance in terms of stability and computational time.