Outlier Detection using Unsupervised Learning Techniques

Abstract

Outlier Detection is a technique to detect anomalous events or outliers during analysis of the data in various domains such as computer network intrusion detection, fraud detection, medical and public health, sensor network, text analysis etc. Analysis of outliers leads to get some interesting information. Many outlier detection techniques have been proposed over decades. Those are broadly classified into three classes such as supervised technique, semi-supervised technique and unsupervised technique. Supervised and semi-supervised techniques have a dependency on the labeled dataset to train the model. Hence, unsupervised techniques are popular due to its independency upon labeled training dataset. Many parametric and non-parametric outlier detection approaches have been proposed over the last couple of decades. The existing neighborhood-based non-parametric unsupervised approaches like LOF, symmetric neighborhood, LDOF are proven to be effective when outliers are in a region of variable density. However, these techniques wrongly treat an outlier point as inlier in certain scenarios (outlier located between a dense cluster and close to a sparse cluster).