Fig 1.
The p-values of the different properties of DTPs and contaminated NDTPs according to the Kolmogorov-Smirnov test.
Fig 2.
Flowchart of the proposed framework.
Fig 3.
Structure of the auto-encoders.
Fig 4.
Structure of the stacked auto-encoders.
Table 1.
Parameter settings in the SAE.
Table 2.
The optimal parameters for BSVM (SAE), BSVM (Wrapper), and BSVM (Origin).
Table 3.
Statistical results of the average of 10 iterations of the three models.
Figures in parentheses are the corresponding variance of the 10 independent results.
Fig 5.
The F1 scores of DTPs in 10 independent iterations for the training dataset.
Fig 6.
The F1 scores of DTPs in 10 independent iterations for the testing dataset.
Fig 7.
The Recall Ratios of DTPs in 10 independent iterations for the training dataset.
Fig 8.
The Recall Ratios of DTPs in 10 independent iterations for the testing dataset.
Fig 9.
The Precisions of NDTPs in 10 independent iterations for the training dataset.
Fig 10.
The Precisions of NDTPs in 10 independent iterations for the testing dataset.