Outlier Detection Using Divide-and-Conquer Strategy in Density Based Clustering

This paper proposes an algorithm to output clusters and outliers in a divide-and-conquer manner. The concepts of density in context of the objects and their clusters are defined as computable mathematical notions. A density-based clustering approach is followed to identify core objects and outliers in each cluster. The utility of proposed clustering method for outlier detection is demonstrated through experiments performed over some popular and real life datasets. The results support the claim about the efficacy of proposed algorithm for purpose of outlier detection. It does not suffer from the drawback of over-rejection.