Ensemble Clustering for Boundary Detection in High-Dimensional Data

The emergence of novel data collection methods has led to the accumulation of vast amounts of unlabelled data. Discovering well separated groups of data samples through clustering is a critical but challenging task. In recent years various techniques to detect isolated and boundary points have been developed. In this work, we propose a clustering methodology that enables us to discover boundary data effectively, discriminating them from outliers. The proposed methodology utilizes a well established density based clustering method designed for high dimensional data, to develop a new ensemble scheme. The experimental results demonstrate very good performance, indicating that the approach has the potential to be used in diverse domains.