An Effective Nearest Neighbor Classification Technique Using Medoid Based Weighting Scheme

The k-nearest neighbor decision rule is a simple, robust and widely used classifier. The method puts a point into a particular class, if the class has the maximum representation among the k nearest neighbors of the point in the training set. However, determining the value of k is difficult. Moreover, nearest neighbor classification techniques put more stress on the data points that lie on the boundary region of individual classes. These methods rely upon those boundary points to decide the class label of a new data point, but, the boundary points may not be a good representation of a particular class. A method is thus proposed here in spirit of the nearest neighbor classification technique using a medoid based weighting scheme to overcome these limitations. The experimental results using various standard benchmark data sets have shown that the proposed method outperforms different state of the art classifiers.