Adaptive K-Nearest Neighbor Classifier Based on Features Extracted by Nonparametric Model

In general there are two main approaches for overcoming the highdimensional and small sample size (SSS) problem. One is to apply feature extraction or selection to reduce the dimensionality, and then applying the reduced-dimensionality data set to classifier. The other is to modify the classifier design to be suitable for SSS problem. This study integrates the two approaches into a new K-nearest neighbour (KNN) classifier, namely adaptive KNN (AKNN). One remotely sensed hyperspectral benchmark image data set is included for investigating the effectiveness of AKNN. Experimental results demonstrate that the proposed AKNN can perform better than KNN and support vector machine (SVM) classifier. Keywords— K-nearest neighbor classifier, dimension reduction, feature extraction, small sample size problem, curse of dimensionality.