Support Vector Machine (SVM) Classification: Comparison of Linkage Techniques Using a Clustering-Based Method for Training Data Selection

Abstract

Support vectors, which usually compose a subset of training sets, determine the decision function of support vector machine (SVM) classification. Selecting a subset including the support vectors through reducing a large training set is a challenge. This paper examines how different linkage techniques in a clustering-based reduction method affect classification accuracy for semiarid vegetation mapping. The investigated linkage techniques include single, complete, weighted pairgroup average, and unweighted pair-group average. Using a multiple-angle remote sensing data set, there is no loss of SVM accuracy when the original training set is reduced to 21%, 14%, 20%, and 20% for these four linkage techniques, respectively.