Abstract
In this paper, we propose a rough set analysis of information tables whose attribute values may take set-values. For such information tables, the traditional rough set analysis based on the indiscernibility relation is not sufficient since we can characterize the objects not only by the set-valued attribute values theirselves but also by their elements. In order to treat the multiplicity of the characteristics of objects, we extend the rough sets to the case when a cover is given. An elementary set of the cover is a collection of objets whose attribute values include common elements. Corresponding to the lower approximations of extended rough sets, we consider decision rules. We extract all decision rules based on a decision matrix.
