Table 4.
Number of features | Accuracy in training set | Accuracy in validation set | % in which the prediction is better than randomly assigned labels | Accuracy without selected features |
---|---|---|---|---|
1 | 75% | 67% | 98.3% | 50% |
2 | 81% | 70% | 99.9% | 50% |
3 | 83% | 70% | 97.7% | 50% |
4 | 84% | 70% | 95.2% | 50% |
5 | 86% | 70% | 95.7% | 50% |
6 | 86% | 70% | 99.7% | 50% |
7 | 88% | 69% | 98.8% | 50% |
8 | 89% | 69% | 98.7% | 50% |
9 | 90% | 69% | 99.9% | 50% |
10 | 90% | 69% | 99.7% | 50% |
11 | 90% | 69% | 98.8% | 50% |
12 | 90% | 70% | 96.2% | 50% |
13 | 90% | 69% | 98.3% | 50% |
14 | 90% | 70% | 99.8 5 | 50% |
15 | 90% | 70% | 100% | 50% |
16 | 90% | 70% | 99.9% | 50% |
17 | 90% | 70% | 99.7% | 50% |
18 | 90% | 70% | 100% | 50% |
19 | 90% | 70% | 100% | 50% |
20 | 90% | 70% | 100% | 50% |
All accuracy values are cumulative.