Finding outliers using mutual nearness based ranks detection algorithm

Outlier detection is concerned with finding exceptional behaviours of objects in the datasets. We propose a novel approach for finding outliers using outlierness score - average of mutual nearness based ranks of each object that emphasize on crucial issue whether an object is significant for its nearest neighbors or not. A centrally located object in the cluster has relatively low outlierness score since it is among the nearest neighbors of its own nearest neighbours. On the other hand, an object at the periphery of a cluster has high outlierness score because its nearest neighbors are closer to the points. Use of outlierness score method eliminates the problem of density calculation in the neighborhood of the point and this improves performance. Outlier detection is becoming a growingly useful tool in many applications such as fraud detection in credit cards, exposing criminal behaviours in e-commerce, computer intrusions identification, detecting health problems, analysing satellite images.