Research Article
Keratoconus Severity Classification Using Features Selection and Machine Learning Algorithms
Table 4
Classification accuracy of RF model with respect to different feature selection methods.
| | Model | Feature selector | Keratoconus classification accuracy | | 2 classes | 4 classes |
| | Random forest | All features | 98.0% | 95.32% | | MI | 97.15% | 90.54% | | ANOVA | 97.03% | 90.83% | | Embedded | 97.63% | 91.21% | | Embedded and filter | 96.81% | 89.79% | | Filter and RFE | 97.63% | 93.17% | | RFE | 97.91% | 93.83% | | Filter and HRFA | 88.9% | 77.9% | | Filter and SFS | 97.76% | 93.64% | | SFS | 98.1% | 95.32% | | Genetic | 98.04% | 94.34% | | Filter and SBS | 97.34% | 93.71% | | SBS | 98.07% | N.A |
|
|
