Preparation to CISS-2022 competition in Singapour
SWaT.A1 _ A2_Dec 2015
| Метод | Гиперпараметры | Threshold | F1 | Accuracy | Precision | Recall |
|---|---|---|---|---|---|---|
| SVM | nu=0.00458 kernel="rbf" gamma=0.0008 scaler=Standard |
- | 0.28 | 0.47 | 0.17 | 0.86 |
| SVM | scaler=MinMax | - | 0.26 | 0.37 | 0.15 | 0.89 |
| LOF | n_components=5 n_neighbors=10 scaler=Standard |
-165.84 | 0.56 | 0.93 | 1.00 | 0.39 |
| LOF | -30 | 0.73 | 0.94 | 0.91 | 0.60 | |
| iForest | n_estimators=50 contamination=0.001 scaler=Standard |
- | 0.72 | 0.95 | 1.00 | 0.57 |
| iForest | n_estimators=200 contamination=0.005 |
- | 0.74 | 0.95 | 0.98 | 0.59 |
| LSTM-AE | 64-32-32-64 window_size=30 resampling=1m scaler=Standard |
5977 | 0.00 | 0.88 | 0.00 | 0.00 |
| LSTM-AE | 800 | 0.00 | 0.88 | 0.60 | 0.00 | |
| LSTM-AE | 400 | 0.74 | 0.95 | 1.00 | 0.59 | |
| MLP-AE | 16-8-4-4-8-16 | 0.74 | 0.95 | 1.00 | 0.59 | |
| SOM | 8x8 sigma=2.0 learning_rate=0.5 |
784 | 0.00 | 0.88 | 0.00 | 0.00 |
| SOM | 220 | 0.74 | 0.95 | 0.91 | 0.65 |
SWaT.A4 _ A5_Jul 2019
| Метод | Гиперпараметры | Threshold | F1 | Accuracy | Precision | Recall |
|---|---|---|---|---|---|---|
| SVM | nu=0.00458 kernel="rbf" gamma=0.0008 scaler=Standard |
- | 0.49 | 0.50 | 0.35 | 0.81 |
| SVM | scaler=MinMax | - | 0.54 | 0.52 | 0.38 | 0.93 |
| LOF | n_components=5 n_neighbors=10 scaler=Standard |
-4.5 | 0.49 | 0.62 | 0.41 | 0.60 |
| iForest | n_estimators=50 contamination=0.001 scaler=Standard |
- | 0.00 | 0.70 | 0.00 | 0.00 |
| iForest | n_estimators=200 contamination=0.05 |
- | 0.56 | 0.65 | 0.45 | 0.74 |
| LSTM-AE | 64-32-32-64 window_size=30 resampling=1m scaler=Standard |
200 | 0.49 | 0.49 | 0.35 | 0.81 |
| LSTM-AE | 80.5 | 0.48 | 0.49 | 0.35 | 0.78 | |
| MLP-AE | 16-8-4-4-8-16 | 1.89 | 0.45 | 0.63 | 0.40 | 0.52 |
| MLP-AE | 1.5 | 0.51 | 0.55 | 0.38 | 0.79 | |
| SOM | 8x8 sigma=2.0 learning_rate=0.5 |
142 | 0.36 | 0.67 | 0.44 | 0.31 |
| SOM | 100 | 0.37 | 0.62 | 0.37 | 0.37 |
WADI.A2_19 Nov 2019
| Метод | Гиперпараметры | Threshold | F1 | Accuracy | Precision | Recall |
|---|---|---|---|---|---|---|
| SVM | nu=0.00458 kernel="rbf" gamma=0.0008 scaler=Standard |
- | 0.10 | 0.45 | 0.05 | 0.51 |
| SVM | без "плохих" признаков | - | 0.39 | 0.95 | 0.71 | 0.27 |
| SVM | scaler=MinMax | - | 0.48 | 0.95 | 0.58 | 0.42 |
| LOF | n_components=5 n_neighbors=10 scaler=Standard |
-89,3 | 0.00 | 0.94 | 1.00 | 0.00 |
| LOF | -20 | 0.33 | 0.94 | 0.52 | 0.24 | |
| iForest | n_estimators=50 contamination=0.001 scaler=Standard |
- | 0.00 | 0.94 | 0.03 | 0.00 |
| iForest | n_estimators=98 contamination=0.04 |
- | 0.28 | 0.88 | 0.21 | 0.41 |
| LSTM-AE | 64-32-32-64 window_size=10 resampling=1m scaler=Standard |
1865 | 0.08 | 0.44 | 0.04 | 0.39 |
| LSTM-AE | без "плохих" признаков | 180 | 0.31 | 0.95 | 0.90 | 0.19 |
| LSTM-AE | scaler=MinMax | 14 | 0.45 | 0.95 | 0.68 | 0.34 |
| MLP-AE | 16-8-4-4-8-16 | 2.5 | 0.08 | 0.44 | 0.04 | 0.39 |
| MLP-AE | без "плохих признаков" | 2.15 | 0.11 | 0.95 | 0.99 | 0.06 |
| MLP-AE | 1.4 | 0.30 | 0.95 | 0.81 | 0.18 | |
| SOM | 8x8 sigma=2.0 learning_rate=0.5 |
800 | 0.00 | 0.94 | 0.00 | 0.00 |
| SOM | 400 | 0.18 | 0.94 | 0.49 | 0.11 |