Table 2 The performance of special versions of our deep learning models and other architectures
From: Network intrusion detection using feature fusion with deep learning
Model | Train | Validation Performance (%) | Test |
|---|---|---|---|
Early-fusion (w/o processing layer) | Â | Â | Â |
 Accuracy | 84.27 | 84.33 | 73.55 |
 Precision | 92.41 | 92.18 | 83.45 |
 Recall | 78.39 | 78.32 | 66.04 |
Late-fusion (w/o processing layer) | Â | Â | Â |
 Accuracy | 84.32 | 84.48 | 74.09 |
 Precision | 93.8 | 94.02 | 84.07 |
 Recall | 77.32 | 77.37 | 64.8 |
Late-Ensemble (w/o processing layer) | Â | Â | Â |
 Accuracy | 84.19 | 84.26 | 73.97 |
 Precision | 92.38 | 92.46 | 82.05 |
 Recall | 78.62 | 78.58 | 65.48 |
Traditional FCN (w/o processing layer) | Â | Â | Â |
 Accuracy | 71.88 | 68.25 | 31.93 |
 Precision | 75.24 | 71.98 | 31.93 |
 Recall | 68.56 | 64.94 | 31.90 |