A Novel Robust Fuzzy Rough Set Model for Feature Selection

The existing fuzzy rough set (FRS) models all believe that the decision attribute divides sample into several “clear” classes, and this data processing method makes model sensitive to noise information when conducting feature selection. To solve problem, paper proposes a robust (RS-FRS) based on representative samples. Firstly, membership degree of samples is defined reflect its fuzziness uncertainty, RS-FRS constructed reduce influence does not need parameters for in advance can effectively complexity human intervention. On basis, related properties are studied, pair selection algorithm (SPS) used In paper, tested analysed open 12 datasets. experimental results show proposed select most relevant features has certain robustness information. good applicability improve performance