Evaluation of Subspace Clustering Quality
Subspace clustering methods seek to find clusters in different subspaces within a data set instead of searching them in full feature space. In such a case there is a problem how to evaluate the quality of the clustering results. In this paper we present our method of the subspace clustering quality estimation which is based on adaptation of Davies-Bouldin Index to subspace clustering. The assumptions which were made to build the metrics are presented first. Then the proposed metrics is formally described. Next it is verified in an experimental way with the use of our clustering method IBUSCA. The experiments have shown that its value reflects a quality of subspace clustering thus it can be an alternative in the case where there is no expert’s evaluation.
Unable to display preview. Download preview PDF.
- 3.Ester, M., Kriegel, H.-P., Sander, J., Xu, X.: A density-based algorithm for discovering clusters in large spatial databases with noise. In: KDD, pp. 226–231 (1996)Google Scholar
- 4.Glomba, M., Markowska-Kaczmar, U.: IBUSCA: A Grid-based Bottom-up Subspace Clustering Algorithm. In: Proceedings of the Sixth International Conference on Intelligent Systems Design and Applications (ISDA 2006). IEEE Computer Society, Los Alamitos (2006)Google Scholar
- 5.Han, J., Kember, M.: Data Mining: Concept and Techniques. In: Cluster Analysis, pp. 335–393. Morgan Kaufman Publishers/ Academic Press (2001)Google Scholar
- 6.Newman, S., Hettich, D., Blake, C., Merz, C.: Uci repository of machine learning databases, http://www.ics.uci.edu/mlearn/MLRepository.html