Abstract
We study the domain of dominant competence of six popular classifiers in a space of data complexity measurements. We observe that the simplest classifiers, nearest neighbor and linear classifier, have extreme behavior of being the best for the easiest and the most difficult problems respectively, while the sophisticated ensemble classifiers tend to be robust for wider types of problems and are largely equivalent in performance. We characterize such behavior in detail using the data complexity metrics, and discuss how such a study can be matured for providing practical guidelines in classifier selection.