Dependence of the f-measure on the number of features. Performances (precision, recall, f-measure) for different numbers of features. Starting from the full feature set, recursive feature elimination removes the features with the lowest weight vector and we measure the performance after each round.
Hakenberg et al. BMC Bioinformatics 2005 6(Suppl 1):S9 doi:10.1186/1471-2105-6-S1-S9