Study of Visual Technqiues for Determing Clustering Tendnecy Over the Unlabelled Datasets

Abstract

Finding similarity features between a set of data objects is a primary step in assessment of clusters. Currently, visual techniques such as visual access tendency (VAT), spectral VAT (SpecVAT) and other variants of VAT are widely used for determining the number of clusters. Determining the number of clusters for given data is known as cluster tendency. Popular clustering techniques, such as k-means and other graph-based techniques produces the clusters without knowing the knowledge of cluster tendency. Thus, this paper surveys the visual approaches for addressing the problem of cluster tendency that can be useful for improving the quality of clusters in k-means and graph-based clustering approaches. VAT and SpecVAT are major visual approached and that can be tested on synthetic datasets and presented cluster assessment results in observation study.