Study On The Approaches For Dynamically Updating Approximations Under The Dominance Relation Based Rough Sets

Rough set theory has been successfully applied in many areas as a kind of mathematical tools to deal with indiscernible data. However, Rough set theory under the indiscernibility relation does not consider attributes with preference-ordered domains. Therefore, Dominance-based Rough Sets Approach (DRSA) was proposed to address a large number of multiple criteria decision problems with a preference order.At present, research on DRSA mainly focused on static data. However, the data in decision systems is continuously accumulated and constantly updated. Aiming to the dynamic characteristics of the data, to study incremental learning approaches based on DRSA has important practical significance.In this thesis, we discuss approaches for incremental updating approximations while attributes'adding or deleting and attribute values'coarsening and refining based on the DRSA. Firstly, we propose some relevant theorems and algorithms about dynamically updating approximations, respectively when attributes'adding and deleting and attribute values'coarsening and refining. A comparative analysis of time complexities is outlined between the dynamic algorithm and original algorithm for updating approximations. Experimental evaluation validates the proposed approached for dynamically updating approximations. It may contribute to improve the efficiency of knowledge discovery based on DRSA.