Generalized Relation Based Knowledge Discovery in Interval-valued Information Systems

A novel generalized relation is proposed in this paper to discover the knowledge of an interval-valued information system more objective based on the concept of interaction rates, which are the length of the intersection between two objects divided by the length of the first object according to an attribute. So, the two interaction rates between two objects are usually different. The generalized relation may be asymmetric and need not to change the two measures between two objects into identical value. This means the generalized relation is better for presenting the knowledge in an interval-valued information system. In order to substantiate the conceptual arguments numerical examples are given.