Fig. 2

Aggregated results of the feature rankings. Borda list of the 700 rankings obtained applying seven ranking methods on 100 instances of 70% training subsets of D. We ranked the Borda list by importance, quantitatively expressed as the Borda count score, corresponding to the average position across all 700 lists. The lower the score, the higher the average rank of the feature in the 700 lists and thus the more important the feature. We highlight the top two features with red circles