Fig. 14
From: Graph clustering-based discretization of splitting and merging methods (GraphS and GraphM)
Ranking of predictive accuracies for standard datasets. Average predictive accuracies ranking with different classifiers of: a C4.5, b KNN, c NB, d SVM, and e all classifiers, obtain from the non-parametric Friedman test and the critical difference (CD) of the Nemenyi post hoc test. Note that, the ranking in a,b, c, and d are tested with 300 samples (30 standard datasets and tenfolds), while the ranking in e is tested with 1200 samples (30 standard datasets of 4 classifiers and tenfolds)