Improved Classification Performance with Label Propagation and SMOTE-ENN

From the results in Table 1, it can be seen that our method has a higher AUC for classification on all datasets compared to other methods, or is on par with them. Similarly, in Table 2, our method achieves higher G_mean scores on most datasets compared to other methods. While AUC shows a slight improvement, G_mean significantly improves on some datasets, indicating that our method is effective in classifying both positive and negative instances. Comparing with the use of only the SMOTE-ENN algorithm, it can be seen that the classification performance improves after incorporating the newly labeled positive instances obtained through label propagation, especially with a higher improvement in the G_mean metric. This is due to the introduction of pseudo-labeled positive instances, which increases the diversity of positive samples and facilitates the generation of more diverse positive instances using the SMOTE algorithm.