Incremental attribute reduction in incomplete decision systems

Abstract

According to whether the underlying information decision system varies with time, methods for attribute reduction can be categoried as static and dynamic two groups. While most existing work is done for the former, seveval approaches have been developed recently for the latter if the information system is complete, i.e. contains no missing values on any attribute. As to dynamic attribute reduction in incomplete decision systems, there is no work known to our knowledge. In this paper, with the introduction of lower approximation attribute reduction into incomplete decision systems, we present an incremental attribute reduction updating scheme based on discernibility matrices when object set is added to an incomplete decision system.