Research Article
Effect Improved for High-Dimensional and Unbalanced Data Anomaly Detection Model Based on KNN-SMOTE-LSTM
Table 5
Comparison of detection results of basic classifier.
| Basic classifier | Error rate | Accuracy | Precision | Recall | F-score | AUC |
| Gaussian naive Bayes (GaussianNB) | 0.0176 | 0.9824 | 0.1026 | 0.8537 | 0.1832 | 0.9163 | Logistic regression | 0.0008 | 0.9992 | 0.8492 | 0.6740 | 0.7516 | 0.8296 | AdaBoost classifier | 0.0008 | 0.9992 | 0.8113 | 0.7143 | 0.7597 | 0.8611 | k-nearest neighbor classifier (kNN) | 0.0007 | 0.9993 | 0.9228 | 0.6726 | 0.7781 | 0.8373 | BP neural network | 0.0007 | 0.9993 | 0.8894 | 0.7189 | 0.7952 | 0.8591 | Gradient boosted decision tree (GBDT) | 0.0006 | 0.9994 | 0.9175 | 0.7154 | 0.8040 | 0.8604 | Support vector machine (SVM) | 0.0007 | 0.9993 | 0.8170 | 0.8022 | 0.8095 | 0.9012 | Random forest (RF) | 0.0005 | 0.9995 | 0.9313 | 0.7850 | 0.8519 | 0.8902 | LSTM | 0.0005 | 0.9995 | 0.8723 | 0.8255 | 0.8483 | 0.9132 |
|
|