Enhancing rough clustering with outlier detection based on evidential clustering

Abstract

Soft clustering plays an important role in many real world applications. Fuzzy clustering, rough clustering, evidential clustering and many other approaches are used effectively to overcome the rigidness of crisp clustering. Each approach has its own unique features that set it apart from others. In this paper, we propose an enhanced rough clustering approach by combining the strengths of rough clustering and evidential clustering. The rough K-means algorithm is augmented with an ability to determine outliers in datasets using the concepts from the Evidential c-means algorithm. Different experiments are carried on various datasets and it is found that the modified rough K-means can effectively detect outliers with relatively smaller computational complexity. © 2013 Springer-Verlag.