An Experiment with Distance Measures for Clustering

Distance measure plays an important role in clustering data points. Choosing the right distance measure for a given dataset is a non-trivial problem. In this paper, we study various distance measures and their effect on different clustering techniques. In addition to the standard Euclidean distance, we use Bit-Vector based, Comparative Clustering based, Huffman code based and Dominance based distance measures. We cluster both synthetic datasets and one real life dataset using the above distance measures by employing k-means, matrix partitioning and dominance based clustering algorithms. We analyse the results of our study using a real life dataset of cricket and compare the accuracy of various techniques using synthetic datasets.

• Type: Conference
• Year: 2008
• Where: COMAD
• Authors: Ankita Vimal, Satyanarayana R. Valluri, Kamalakar Karlapalem
• Keywords: COMAD 2008, Distance Measures, Knowledge Management, Real Life Dataset, Right Distance Measure