Table A.5.
% of feature space | Total # of features | Naïve Bayes
|
SVM (linear kernel)
|
SVM (RBF kernel)
|
MaxEnt
|
||||||||
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
P. | R. | F1 | P. | R. | F1 | P. | R. | F1 | P. | R. | F1 | ||
| |||||||||||||
75.00 | 34,272 | 0.761 | 0.635 | 0.688 | 0.739 | 0.771 | 0.752 | 0.727 | 0.821 | 0.770 | 0.746 | 0.767 | 0.754 |
50.00 | 22,848 | 0.738 | 0.624 | 0.674 | 0.738 | 0.767 | 0.749 | 0.728 | 0.825 | 0.772 | 0.749 | 0.763 | 0.753 |
25.00 | 11,424 | 0.819 | 0.609 | 0.693 | 0.761 | 0.767 | 0.760 | 0.717 | 0.810 | 0.759 | 0.753 | 0.775 | 0.761 |
10.00 | 4,569 | 0.792 | 0.581 | 0.660 | 0.753 | 0.779 | 0.763 | 0.701 | 0.779 | 0.737 | 0.742 | 0.748 | 0.741 |
5.00 | 2,284 | 0.699 | 0.488 | 0.567 | 0.746 | 0.799 | 0.770 | 0.684 | 0.799 | 0.735 | 0.717 | 0.763 | 0.738 |
1.00 | 1142 | 0.676 | 0.542 | 0.595 | 0.705 | 0.775 | 0.738 | 0.657 | 0.810 | 0.723 | 0.685 | 0.760 | 0.720 |
0.50 | 456 | 0.688 | 0.728 | 0.705 | 0.685 | 0.782 | 0.729 | 0.637 | 0.830 | 0.720 | 0.671 | 0.786 | 0.723 |
0.10 | 228 | 0.628 | 0.821 | 0.710 | 0.703 | 0.763 | 0.731 | 0.618 | 0.841 | 0.711 | 0.650 | 0.771 | 0.705 |
0.05 | 45 | 0.561 | 0.856 | 0.677 | 0.767 | 0.767 | 0.703 | 0.596 | 0.899 | 0.715 | 0.611 | 0.806 | 0.694 |
Note: Performance of shaded region is not significantly different from the top overall model (i.e., F1 of 77.4%).