Support Vector Clustering with Outlier Detection

Abstract

This paper presents a cluster validity measure with outlier detection for support vector clustering (SVC) algorithm. We proposed an outlier detection approach for dealing with noise data and a cluster validity process for identifying an optimal cluster configuration with suitable parameters without a priori knowledge regarding the given data sets. Since SVC is a kernel-based clustering approach, the parameter of kernel functions and the soft-margin constants in Lagrangian functions play a crucial role in the clustering results. A validity measure with outlier detection has been developed to automatically determine suitable parameters. Using these parameters, the SVC algorithm can identify an optimal cluster number and increase its robustness to outliers and noise. The simulations have been conducted to demonstrate the effectiveness of the proposed cluster validity measure and outlier detection for benchmark datasets.