Research Article
An Efficient Communication Intrusion Detection Scheme in AMI Combining Feature Dimensionality Reduction and Improved LSTM
Table 9
Evaluation results of SAE and LSTM with different structures.
| Structures of SAE and LSTM | Accuracy | Precision | Recall | FAR |
| {128, 64, 32, 32} | Attention-BiLSTM | 0.9941 | 0.9914 | 0.9952 | 0.0069 | LSTM | 0.9859 | 0.9778 | 0.9905 | 0.0178 |
| {128, 64, 32} | Attention-BiLSTM | 0.9893 | 0.9822 | 0.9938 | 0.0142 | LSTM | 0.9809 | 0.9682 | 0.9888 | 0.0252 |
| {128, 32, 32} | Attention-BiLSTM | 0.9861 | 0.9763 | 0.9924 | 0.0189 | LSTM | 0.9767 | 0.9595 | 0.9880 | 0.0320 |
|
|