Evaluation of Rough set and Fuzzy rough set models with application to multi-attribute decision making
Archana K. Ratnaparkhi, Dattatraya Shankar Bormane, Rajesh Ghongade · 2019
Uncertainty and randomness is inherent to real valued datasets. With the advent of fuzzy concepts, noise tolerant models are being studied by researchers nearly for a decade. Fuzzy sets and extended fuzzy sets are now widely used to represent the vague and ambiguous nature of datasets. The proposed work investigates the relationship between the output class and the optimal attribute subset on the basis of rules induced using rough sets, vaguely quantified rough sets, ordered weighted average fuzzy rough sets and fuzzy rough k-nearest neighbor model. Fuzzy rough sets and its extended models have been applied on bench mark databases. The novelty of the work lies in the application of these models on optimally generated subsets. The optimal generation of attributes is carried out using discernibility matrix based approach. This approach is also compared with the standard correlation and mutual information based approaches. The results indicate that fuzzy rough k-nearest neighbour approach combined with discernibility matrix based attribute generation outperforms other methods in terms of classification accuracy. Sensitivity analysis indicates the efficacy of the proposed model over traditional approaches.