Similarity-based data reduction and classification

Abstract

The κ-Nearest-Neighbors (κNN) is a simple but effective method for classification. The major drawbacks with respect to κNN are (1) low efficiency and (2) dependence on the parameter κ. In this paper, we propose a novel similarity-based data reduction method and several variations aimed at overcoming these shortcomings. Our method constructs a similarity-based model for the data, which replaces the data to serve as the basis of classification. The value of κ is automatically determined, is varied in terms of local data distribution, and is optimal in terms of classification accuracy. The construction of the model significantly reduces the number of data for learning, thus making classification faster. Experiments conducted on some public data sets show that the proposed methods compare well with other data reduction methods in both efficiency and effectiveness.© Springer-Verlag Berlin Heidelberg 2005.