Identifying Outliers Using Voronoi Circles

A new outliers detection technique has been proposed in this paper in context of Computational geometry. Empty circles, Voronoi diagram, and convex hull are used in this proposed new technique for anomaly detection. In the context of empty circles, it is the fact that no data points are there inside the boundary of the circle. This fact played a vital role to find the outliers. It is obvious that most of the outliers reside within the circumference of the large empty circle (LEC). We revealed that the circumference points of LEC and boundary points of the convex hull are likely to be the outliers of data sets. It is well known that the K-nearest neighbor (KNN) outlier finding method is very much popular in the context of outliers finding. However, till now computational geometry is not used in state of the art for outlier detection models. In this work, for the first time, we used computational geometry to detect outliers. Our algorithm is successfully implemented with PYTHON using various benchmark data sets. The experimental results establish the novelty and validity of the proposed method over the standard KNN technique.