WOC: A New Weighted Ordinal Classification
Markus Zeindl, Christian Facchi · 2015
In ordinal classification problems, data objects are grouped into at least three different classes by an appropriate classification model, which can be arranged in a total ordering. Performance evaluation of such problems will actually be performed using imprecise evaluation metrics. This paper proposes WOC, a novel evaluation metric for ordinal classification problems and shows, that this metric acts as expected. As evaluation results confirm, the proposed metric provides more precise information about the quality of decision made by ordinal classification models.