Skip to main content

Table 4 Comparative analysis with respect to competing feature selection / extraction techniques using the same classifier (LDA)

From: A novel feature extraction methodology using Siamese convolutional neural networks for intrusion detection

Exp. Preprocessing Feature Selection / Extraction Dimensionality of the resulted feature space Training (%) Testing (%) Execution time (s)
8. Fuzzification Vec2im – Siam 1 98.69 86.64 113.736
9. Normalization Vec2im – Siam 1 90.20 80.64 113.245
10. Normalization Wr-FS (best feature) 1 82.84 77.91 1.991
11. Fuzzification Wr-FS (best feature) 1 82.84 77.91 1.953
12. Normalization Wr-FS (2 first features) 2 90.79 82.04 3.048
13. Fuzzification Wr-FS (2 first features) 2 90.79 82.04 3.161
14. Normalization Wr-FS (3 first features) 3 92.20 82.98 4.370
15 Fuzzification Wr-FS (3 first features) 3 91.24 85.81 4.788
16. Normalization Wr-FS (10 first features) 10 92.33 84.06 17.091
17. Fuzzification Wr-FS (10 first features) 10 92.03 85.93 17.214
18. Normalization ILFS 39 95.44 78.44 0.307
19. Normalization InfFS 13 93.91 81.94 0.063
20. Normalization Relief-F 13 92.96 81.75 122.105
21. Normalization LS 33 95.42 78.50 158.381
22. Normalization Fisher 13 93.91 81.94 0.415
23. Normalization CFS 24 93.14 81.51 0.071
24. Normalization UFSOL 39 95.44 78.44 312.436
25. Normalization Lasso 22 93.61 82.04 14.505
26. Normalization Chi2 33 95.44 78.48 1.113
27. Normalization mRMR 26 94.70 81.24 1.060
28. Normalization FuzCoC 25 94.61 81.94 1.160
29. Normalization PCA (1 pr. component) 1 89.69 75.64 0.144
30. Normalization PCA (2 pr. components) 2 89.47 76.30
31. Normalization PCA (3 pr. components) 3 89.48 76.32