Fig. 4

Precision-recall plot comparing analyzed feature selection methods for all simulation settings. Precision corresponds to the ratio of relevant detected features divided by total amount of features in the model. Recall shows the ratio of relevant detected features divided by the total existing number of relevant features. The cost budget defines an upper limit for the recall in the simulations. It is highlighted by a green line. To assess the quality of the feature selection methods precision and recall for selecting features randomly is added to the plots as horizontal and vertical dashed lines. The plot boundaries are re-scaled to depict the area of interest between randomness and optimal values