Research Article
A Hybrid Feature Selection Method Based on Rough Conditional Mutual Information and Naive Bayesian Classifier
Table 5
Number and accuracy of selected features based on CART.
| Number | Unselect | RCMI | Filter wrapper | Accuracy | Feature number | Accuracy | Feature number | Accuracy | Feature number |
| 1 | 72.35% | 279 | 75.00% | 16 | 75.89% | 9 | 2 | 79.35% | 19 | 85.16% | 8 | 85.81% | 3 | 3 | 89.74% | 34 | 91.74% | 6 | 91.74% | 6 | 4 | 96.15% | 19 | 95.54% | 3 | 95.80% | 3 | 5 | 74.52% | 60 | 77.40% | 15 | 80.77% | 6 | 6 | 92.53% | 35 | 91.36% | 24 | 91.95% | 10 | 7 | 95.63% | 16 | 96.09% | 2 | 95.63% | 1 | 8 | 93.50% | 16 | 94.90% | 4 | 94.90% | 4 | 9 | 94.94% | 12 | 94.94% | 5 | 93.82% | 3 | 10 | 92.08% | 16 | 93.07% | 10 | 93.07% | 8 |
| Average | 88.08% | 50.6 | 89.52% | 9.3 | 89.94% | 5.3 |
|
|