Multigranulation Rough Set Methods and Applications Based on Neighborhood Dominance Relation in Intuitionistic Fuzzy Datasets

With the redundancy and complexity of information and data, how to acquire the samples that meet the requirements is an inevitable task in data analysis. There is a general consensus that the neighborhood rough set (NRS) has become the mainstream method for data mining and knowledge classification. Whereas, the limitations still exist in the neighborhood relation for it cannot more accurately reflect the dominance relations that commonly exist in actual data, nor can it select the required data according to different conditions. Enlightened by this idea, this paper focuses on the intuitionistic fuzzy neighborhood dominance relation, which both refines the relationship between samples in the neighborhood and mines the needed samples in data analysis. On this basis, we define the neighborhood dominance rough set (NDRS) model in intuitionistic fuzzy ordered information system (IFOIS). Moreover, we establish the multigranulation neighborhood dominance rough set (MNDRS) from multiple perspectives, and discuss related properties between NDRS and MNDRS. Meanwhile, we compare the NDRS with other rough set models from the roughness and the dependence degree viewpoints. Finally, we adopt nine UCI data sets and implement a series of experiments to illustrate the feasibility and effectiveness of the proposed models.