Editing Prototypes in the Finite Sample Size Case Using Alternative Neighbourhoods

The recently intro(hwed concept of Nearest Centroid Neight)orhood is applied to discard outlirrs and prototypes in cl,~s overlapping regions in order to improve the performance of the Nearest Neighbor rule through an etliting i)rocedure. This apl)roach is related to graph b~sed editing algorithms which also define alternatiw, neighborhoods in t[,rms of geometric relations. Cl,~si('al e([iting algorithms are compared to these alternative editing schemes using sew,ral synthetic and real data problems. The empirical results show that the proposed editing algorithm constitutes a good trade-off among performance and computational burden.