Asymmetric neighbourhood selection and support aggregation for effective classification

The k-Nearest-Neighbours (kNN) is a simple but effective method for classification. The success of kNN in classification depends on the selection of a "good value" for k. To reduce the bias of k and take account of the different roles or influences that features play with respect to the decision attribute, we propose a novel asymmetric neighbourhood selection and support aggregation method in this paper. Our aim is to create a classifier less biased by k and to obtain better classification performance. Experimental results show that the performance of our proposed method is better than kNN and is indeed less biased by k after saturation is reached. The classification accuracy of the proposed method is better than that based on symmetric neighbourhood selection method as it takes into account the different role each feature plays in the classification process.