A Comparative Evaluation of Supervised and Unsupervised Methods for Detecting Outliers

Outlier refers to a sample which is deviated with the rest of the samples in the dataset. Such outliers can be categorized as global outliers, collective outliers, and contextual outliers based on the behaviour and the degree of its deviation from the normal samples. Methods like supervised, unsupervised and semi-supervised are employed to find these kinds of outliers. This paper sheds light on the layout and performance analysis of supervised and unsupervised outlier detection methods in determining the aforementioned outliers. To understand the performance of these methods in predicting outliers from few benchmarks and realtime data sets, the data mining tools like Rapid Miner and R is used.