Fig. 4
From: Unsupervised encoding selection through ensemble pruning for biomedical classification
The kappa-error diagram depicts pairs of base classifiers (dots) using the kappa diversity on the x-axis and the average error on the y-axis (left column). Each classifier pair denotes two particular encodings. Rows depict the base classifiers. The dark blue line indicates the convex hull, and the red line the Pareto frontier. Pruning methods are represented as colors. The left column shows the first fold. The right column groups the results of all folds in a 2D histogram. The darker the color, the more classifier pairs are binned in one group. The dashed line in all panels depicts the theoretical boundary (refer to the method section for more details). The plot shows the example of the avp_amppred dataset