Research Article
Oversampling Imbalanced Data Based on Convergent WGAN for Network Threat Detection
Table 1
The symbols are used in the formulas from (
1) to (
15) in Section
3.3.
| No. | Math symbols | Meaning of the math symbols |
| 1 | | The training set | 2 | | The size of the training set | 3 | | The majority class of the training set | 4 | | The size of the majority class in the training set | 5 | | The minority class of the training set | 6 | | The size of the minority class in the training set | 7 | | The imbalance ratio of the imbalanced data sets | 8 | | The oversampling rate of the minority class | 9 | | The false data generated by the generator of GAN | 10 | | The output data of the generator | 11 | | The output data of the discriminator | 12 | | The set of all possible joint distributions of the combination of and | 13 | | One joint distribution of all possible joint distributions of the combination of and | 14 | | The function of neural network model | 15 | | The Lipschitz constant | 16 | | The Wasserstein distance | 17 | , | The iterations in the training process of CWGAN | 18 | , | The training epoch in each iteration | 19 | | The training epoch in iteration | 20 | | The loss function of the discriminator | 21 | | The training stride in the training process of CWGAN | 22 | | The optimal solution of | 23 | | The optimal value of | 24 | | The Wasserstein distance between generated data and real data when the discriminator is convergent at the -th iteration. |
|
|