Table 8 Performance classification for the proposed Hybrid Feature Selection with tenfold validation
From: A novel time efficient learning-based approach for smart intrusion detection system
Machine learning algorithm | Feature selection | Model training time (in secs) | Prediction time (in secs) batch prediction for test set | Percentage decrease in model building time using hybrid approach (%) | Percentage decrease in prediction latency using hybrid approach (%) |
|---|---|---|---|---|---|
Histogram Based Gradient Descent | PCA | 62.91809869 | 0.387566 | 34.74 | 22.46 |
Hybrid (RF + PCA) | 41.05721924 | 0.300503 | |||
ExtraTrees | PCA | 27.95780349 | 0.353288 | 17.94 | 2.25 |
Hybrid (RF + PCA) | 22.94263474 | 0.345334 | |||
RandomForest | PCA | 126.6453106 | 0.352546 | 27.93 | 4.03 |
Hybrid (RF + PCA) | 91.27173629 | 0.33835 | |||
XGBoost | PCA | 156.5729681 | 0.329168 | 52.68 | 44.52 |
Hybrid (RF + PCA) | 74.09030104 | 0.182613 | |||
KNN | PCA | 129.2284153 | 310.8692 | 18.58 | 40.03 |
Hybrid (RF + PCA) | 105.2205418 | 186.4265 | |||
LightGBM | PCA | 7.37608552 | 0.143057 | 31.36 | 3.53 |
Hybrid (RF + PCA) | 5.062662244 | 0.138008 |