Title :
Theoretical and Empirical Criteria for the Edited Nearest Neighbour Classifier
Author :
Ludmila I. Kuncheva;Mikel Galar
Author_Institution :
Sch. of Comput. Sci., Bangor Univ., Bangor, UK
Abstract :
We aim to dispel the blind faith in theoretical criteria for optimisation of the edited nearest neighbour classifier and its version called the Voronoi classifier. Three criteria from past and recent literature are considered: two bounds using Vapnik-Chervonenkis (VC) dimension and a probabilistic criterion derived by a Bayesian approach. We demonstrate the shortcomings of these criteria for selecting the best reference set, and summarise alternative empirical criteria found in the literature.
Keywords :
"Prototypes","Training","Data mining","Bayes methods","Upper bound","Electronic mail","Probabilistic logic"
Conference_Titel :
Data Mining (ICDM), 2015 IEEE International Conference on
DOI :
10.1109/ICDM.2015.36