An approach to tree-classifier design based on hierarchical clustering

Abstract

The tree classifier has proved to be an effective method of statistical pattern classification, but the designing of the tree structure is not usually easy. In this paper, a new design approach based on hierarchical clustering is proposed. The design principle and the algorithm of the method, as well as the experimental results for both test data and remote-sensing data, are discussed. The design process of the method is simple and the performance of the classifier designed by the method has been proved to be satisfactory.