This work presents a granular K Nearest Neighbor, or grKNN for short, classifier in the metric lattice of Intervals’ Numbers (INs). An IN here represents a population of numeric data samples. We detail how the grKNN classifier can be parameterized towards optimizing it. The capacity of a preliminary grKNN classifier is demonstrated, comparatively, in four benchmark classification problems. The far-reaching potential of the proposed classification scheme is discussed.
V.T. Tsoukalas, V.G. Kaburlasos, C. Skourlas, “A granular, parametric KNN classifier”, 17th Panhellenic Conference on Informatics (PCI 2013), Thessaloniki, Greece, 19-21 September 2013, pp. 319-326.