Enhanced Neighbourhood Specifications for Pattern Classification

Automatic classification is an active research area in artificial intelligence, pattern recognition, and machine learning. Its goal is to make a computer automatically label attribute vectors as examples of two or more different categories or classes. One of the most widely studied approaches corresponds to the well known Nearest Neighbour (NN) decision rule, proposed by Fix and Hodges [19], formally analysed by Cover and Hart [9], and then investigated both theoretically and experimentally by many researchers. On the other hand, a lot of successful applications have been reported in a multitude of real-world domains [10].