A METHOD FOR DETERMINING k‐NEAREST NEIGHBOURS

Abstract

All the modified Nearest Neighbour methods of pattern classification2–6 developed to reduce the amount of computer storage and time needed for the implementation of a NN classifier require prohibitively costly data preprocessing which involves detailed examination of the neighbouring points to the elements of the reference set. In this paper a method for determining k‐nearest neighbours to a given point is described. The method uses the computationally efficient city block distance to select candidate points for the set of k‐nearest neighbours. In this way the preprocessing time is considerably reduced.

Citation

KITTLER, J. (1978), "A METHOD FOR DETERMINING k‐NEAREST NEIGHBOURS", Kybernetes, Vol. 7 No. 4, pp. 313-315. https://doi.org/10.1108/eb005497