An Efficient Visual Analysis Method for Cluster Tendency Evaluation, Data Partitioning and Internal Cluster Validation

Year, pages: 2013, 1013 - 1037

Visual methods have been extensively studied and performed in cluster data analysis. Given a pairwise dissimilarity matrix D of a set of n objects, visual methods such as Enhanced-Visual Assessment Tendency (E-VAT) algorithm generally represent D as an n times n image I( overlineD) where the objects are reordered to expose the hidden cluster structure as dark blocks along the diagonal of the image. A major constraint of such methods is their lack of ability to highlight cluster structure when D contains composite shaped datasets. This paper addresses this limitation by proposing an enhanced visual analysis method for cluster tendency assessment, where D is mapped to D' by graph based analysis and then reordered to overlineD' using E-VAT resulting graph based Enhanced Visual Assessment Tendency (GE-VAT). An Enhanced Dark Block Extraction (E-DBE) for automatic determination of the number of clusters in I( overlineD') is then proposed as well as a visual data partitioning method for cluster formation from I( overlineD') based on the disparity between diagonal and off-diagonal blocks using permuted indices of GE-VAT. Cluster validation measures are also performed to evaluate the cluster formation. Extensive experimental results on several complex synthetic, UCI and large real-world data sets are analyzed to validate our algorithm.

Prabhu, P., Duraiswamy, K. 2013. An Efficient Visual Analysis Method for Cluster Tendency Evaluation, Data Partitioning and Internal Cluster Validation. In Computing and Informatics, vol. 32, no.5, pp. 1013-1037. 1335-9150.

Prabhu, P., Duraiswamy, K. (2013). An Efficient Visual Analysis Method for Cluster Tendency Evaluation, Data Partitioning and Internal Cluster Validation. Computing and Informatics, 32(5), 1013-1037. 1335-9150.