Fusion of Pairwise Nearest-Neighbor Classifiers Based on Pairwise-Weighted Distance Metric and Dempster-Shafer Theory

Overview & Synopsis

The performance of the nearest-neighbor (NN) classifier is known to be very sensitive to the distance metric used in classifying a query pattern, especially in scarce-prototype cases. In this paper, a pairwise-weighted (PW) distance metric related to pairs of class labels is proposed. Compared with the existing distance metrics, it provides more flexibility to design the feature weights so that the local specifics in feature space can be well characterized. Base on the proposed PW distance metric, a polychotomous NN classification problem is solved by combining several pairwise NN (PNN) classifiers within the framework of Dempster-Shafer theory to deal with the uncertain output information. Two experiments based on synthetic and real data sets were carried out to show the effectiveness of the proposed method.