Distance based fast outlier detection method

In this paper, we propose a new method to capture outliers in an efficient way. The proposed approach has two steps, clustering-pruning step and outlier factor step. In clustering-pruning step, the entire input data set is clustered into disjoint clusters using a clustering algorithm and based on the outlier factor of the centroids of the disjoint clusters, we prune away some clusters. In outlier factor step, we calculate outlier score for each point in the unpruned clusters. Based on the outlier score we declare the top-n points with the highest score as outliers. The experimental results using real data set demonstrate that even though the number of computations are less, the proposed method performs better than the existing method.