Classifier-Independent Visualization of Supervised Data Structures Using a Graph
Supervised data structures in high dimensional feature spaces are displayed as graphs. The structure is analyzed by normal mixture distributions. The nodes of the graph correspond the mean vectors of the mixture distributions, and the location is carried out by Sammon’s nonlinear mapping. The thickness of the edges expresses the separability between the component distributions, which is determined by Kullback-Leibler divergence. From experimental results, it was confirmed that the proposed method can illustrate in which regions and to what extent it is difficult to classify samples correctly. Such visual information can be utilized for the improvement of the feature sets.
KeywordsFeature Selection Method High Dimensional Feature Space Confusion Matrice Component Distribution Projection Pursuit
- 6.Mori, Y., Kudo, M., Toyama, J., Shimbo, M.: Comparison of Low-Dimensional Mapping Techniques Based on Discriminary Information. In: Proceedings of the Second International ICSC Symposium on Advances in Intelligent Data Analysis (2001) (2001) ,CD-ROM Paper #1724-166Google Scholar
- 7.Tenmoto, H., Mori, Y., Kudo, M.: Visualization of Class Structures using Piecewise Linear Classifiers. Proceedings of Logic Applied to Technology (LAPTEC), 104–111 (2002)Google Scholar
- 8.Taylor, P.C., Hand, D.J.: Finding ’Superclassifications’ with an Acceptable Misclassification Rate. Journal of Applied Statistics 26, 579–590 Google Scholar
- 13.Ueda, N.: EM Algorithm with Split and Merge Operations for Mixture Models (Invited). Transactions of IEICE E83-D, 2047–2785 (2000)Google Scholar