Comparison of Density-Based and Distance-Based Outlier Identification Methods in Fuzzy Clustering

Abstract

Outlier identification is a process of identification of any kind of abnormality in the data regarding to context or behavior of data objects. In literature, various outlier identification methods such as statistical methods, clustering-based methods, and proximity-based methods have been proposed. However, in fuzzy clustering, mainly proximity-based methods (density-based and distance-based) are used for outlier identification. In this paper, we have compared a density-based outlier identification method and a distance-based outlier identification method in context of fuzzy clustering. Experimental results of these outlier identification methods show higher stability and accuracy of distance-based method over density-based method.