An improved accuracy measure for rough sets

In rough set theory, the accuracy measure is an important numerical characterization that quantifies the imprecision of a rough set caused by its boundary region. However, it does not take into consideration the granularity of the partition induced by an equivalence relation. Therefore, the imprecision of a rough set is not well characterized by the traditional accuracy measure. This paper first analyzes the limitations of the accuracy measure, and then improves it by information theory. Finally, four theorems show that the improved measure has some good properties and hence is more suitable to measure the imprecision of rough sets.