Anomaly Detection on MSL (F1-score)
94.08F1 ScoreDGHL
Evaluation Results
| Method | Links | |
|---|---|---|
| DGHL2025.01 | 94.08 | |
| Anomaly2025.01 | 93.59 | |
| Mutual InfoBandwidth=50%, Classifier=RF n=100, Interpolation method=Forward-fill, Tuning protocol=No per-dataset tuning, Number of seeds=52026.04 | 93 | |
| PCA-TriageBandwidth=50%, Classifier=RF n=100, Interpolation method=Forward-fill, Tuning protocol=No per-dataset tuning, Number of seeds=52026.04 | 92.1 | |
| VarianceBandwidth=50%, Classifier=RF n=100, Interpolation method=Forward-fill, Tuning protocol=No per-dataset tuning, Number of seeds=52026.04 | 91.7 | |
| Random DropoutBandwidth=50%, Classifier=RF n=100, Interpolation method=Forward-fill, Tuning protocol=No per-dataset tuning, Number of seeds=52026.04 | 91.4 | |
| ThresholdBandwidth=50%, Classifier=RF n=100, Interpolation method=Forward-fill, Tuning protocol=No per-dataset tuning, Number of seeds=52026.04 | 91.3 | |
| UniformBandwidth=50%, Classifier=RF n=100, Interpolation method=Forward-fill, Tuning protocol=No per-dataset tuning, Number of seeds=52026.04 | 91.2 | |
| THOC2025.01 | 89.69 | |
| Omni2025.01 | 87.67 | |
| TimesNet2025.01 | 85.15 | |
| ETSformer2023.02 | 85.03 | |
| Dlinear2025.01 | 84.88 | |
| DLinear2023.02 | 84.88 | |
| Pyraformer2023.02 | 84.86 | |
| Reformer2023.02 | 84.4 | |
| Informer2023.02 | 84.06 | |
| Anomaly Transformercriterion=reconstruction error2023.02 | 83.31 | |
| GPT2(6)layers=62023.02 | 82.45 | |
| TimesNetreproduced=true2023.02 | 81.84 | |
| LogTrans2023.02 | 79.57 | |
| Autoformer2023.02 | 79.05 | |
| LightTS2025.01 | 78.95 | |
| LightTS2023.02 | 78.95 | |
| PatchTST2023.02 | 78.7 | |
| Transformer2023.02 | 78.68 | |
| FEDformer2023.02 | 78.57 | |
| FITS2025.01 | 78.12 | |
| Stationary2025.01 | 77.5 | |
| Stationary2023.02 | 77.5 | |
| OCSVM2025.01 | 70.82 | |
| IForest2025.01 | 66.45 | |
| SWIFT2025.01 | 58.51 |