Abstract
Effective partitioning of feature space for high classification accuracy with due attention to rare class members is often a difficult task. In this paper, the border feature detection and adaptation (BFDA) algorithm is proposed for this purpose. The BFDA consist of two parts. In the first part of the algorithm, some specially selected training samples are assigned as initial reference vectors called border features. In the second part of the algorithm, the border features are adapted by moving them towards the decision boundaries. At the end of the adaptation process, the border features are finalized. The method next uses the minimum distance to border feature rule for classification. In supervised learning, the training process should be unbiased to reach more accurate results in testing. In the BFDA, decision region borders are related to the initialization of the border features and the input ordering of the training samples. Consensus strategy can be applied with cross validation to reduce these dependencies. The performance of the BFDA and Consensual BFDA (C-BFDA) were studied in comparison to other classification algorithms including neural network with back-propagation learning (NN-BP), support vector machines (SVMs), and some statistical classification techniques.
Keywords
Decision region borders, BFDA, data classification, remote sensing, consensual classification
Date of this Version
February 2007