Research Article
Effect Improved for High-Dimensional and Unbalanced Data Anomaly Detection Model Based on KNN-SMOTE-LSTM
Table 7
Comparison of detection results of unbalanced algorithm.
| Model | Precision | Recall | F-score | AUC |
| LSTM | 0.8723 | 0.8255 | 0.8483 | 0.9132 | ADASYN + LSTM | 0.0301 | 0.7718 | 0.0580 | 0.9246 | SMOTE + LSTM | 0.1300 | 0.9128 | 0.2276 | 0.9283 | Borderline-SMOTE + LSTM | 0.8095 | 0.7987 | 0.8041 | 0.9021 | Svm SMOTE + LSTM | 0.7669 | 0.8389 | 0.8013 | 0.9073 | SMOTEENN + LSTM | 0.1367 | 0.8523 | 0.2372 | 0.9275 | SMOTETomek + LSTM | 0.1373 | 0.8725 | 0.2356 | 0.9387 | kNN-SMOTE-LSTM (this work) | 0.9496 | 0.8859 | 0.9167 | 0.9296 |
|
|
Note. “+” indicates the combination of models and “−” indicates the structural fusion of models.
|