Distance-based classification methods

Given a set of points in a Euclidean space, and a partitioning of this "tra ining set" into two or more subsets ("classes"), we consider the problem of identifying a "reasonable" assignment of another point in the Euclidean sp ace ("query point") to one of these classes. The various classifications pr oposed in this paper are determined by the distances between the query poin t and the points in the training set. We report results of extensive comput ational experiments comparing the new methods with two well-known distance- based classification methods (k-nearest neighbors and Parzen windows) on da ta sets commonly used in the literature. The results show that the performa nce of both new and old distance-based methods is on par with and often bet ter than that of the other best classification methods known. Moreover, the new classification procedures proposed in this paper are: (i) easy to impl ement, (ii) extremely fast, and (iii) very robust (i.e. their performance i s insignificantly affected by the choice of parameter values).