Research Article
Anomaly Detection Collaborating Adaptive CEEMDAN Feature Exploitation with Intelligent Optimizing Classification for IIoT Sparse Data
Table 3
Training and test classification accuracies of traditional ABC-OCSVM and PSO-OCSVM anomaly detection classifiers.
| Attack power | ABC-OCSVM | PSO-OCSVM | Training accuracy | Test accuracy | Training accuracy | Test accuracy |
| 1 | 96.0% | 86.5% | 98.0% | 80.5% | 2 | 93.5% | 88.0% | 96.5% | 83.0% | 3 | 91.0% | 90.0% | 98.0% | 85.0% | 4 | 96.5% | 87.0% | 100.0% | 84.0% | 5 | 98.5% | 93.5% | 98.0% | 83.5% | Average | 95.10% | 89.00% | 98.00% | 83.20% |
|
|