利用未标记数据解决不平衡二分类问题: 标签传播与重采样算法结合

This article presents a method for solving imbalanced binary classification problems using unlabeled data, combining the label propagation algorithm with the resampling algorithm. Firstly, the unlabeled data is added to the training set, and the label propagation is used to obtain pseudo-labels for the unlabeled data, resulting in samples with pseudo-labels as positive examples. These samples are then combined with the labeled training set to form a new dataset. The new dataset is then resampled to balance the positive and negative samples, and the inclusion of the new minority class increases the diversity of the minority class. Finally, experiments were conducted on the imbalanced binary classification dataset KEEL, and the results showed that using unlabeled samples to expand the training set and enrich the minority class data can effectively improve the classification performance of imbalanced data.