Density Based Outlier Detection Technique

Abstract

Outlier Detection has become an emerging branch of research in the field of data mining. Detecting outliers from a pattern is a popular problem. Detection of Outliers could be very beneficial, knowledgeable, interesting and useful and can be very destructive if remain unexplored. We have proposed a novel density based approach which uses a statistical measure i.e. standard deviation to identify that a data point is an outlier or not. In the current days there are large variety of different solutions has been efficiently researched. The selection of these solutions is sometimes hard as there is no one particular solution that is better than the others, but each solution is suitable under some specific type of datasets. Therefore, when choosing an outlier detection method to adapt to a new problem it is important to look on the particularities of the specific dataset that the method will be applied. To test the validity of the proposed approach, it has been applied to Wisconsin Breast Cancer dataset and Iris dataset.