A Simple Noise-Tolerant Abstraction Algorithm for Fast k-NN Classification

Abstract

The k-Nearest Neighbor (k-NN) classifier is a widely-used and effective classification method. The main k-NN drawback is that it involves high computational cost when applied on large datasets. Many Data Reduction Techniques have been proposed in order to speed-up the classification process. However, their effectiveness depends on the level of noise in the data. This paper shows that the k-means clustering algorithm can be used as a noise-tolerant Data Reduction Technique. The conducted experimental study illustrates that if the reduced dataset includes the k-means centroids as representatives of the initial data, performance is not negatively affected as much by the addition of noise.