Nearest cluster center decision in hierarchical classification process

Abstract

The hierarchical classification method based on SVM multi learning process has proved its abilities to classify large databases. However, some misclassification can be detected through the hierarchy levels. To minimize these errors and improving the rate of classification, we propose to use Nearest Cluster centers (NCC) in the hierarchy structure to correct the classification ambiguities. Obtained results on IRIS database prove the given improvement of global classification rate.