基于标签传播的不平衡分类算法 - 提升少数类样本学习效果

The unbalanced classification algorithm based on label propagation mainly consists of two steps. Firstly, based on the samples and labels of the training set, pseudo labels are assigned to the unlabeled test set through label propagation, and the samples with pseudo labels as positive examples are combined with the training set to form a new dataset with more information from the minority class. Then, the new dataset is sampled using SMOTE-ENN. This is done to balance the number of positive and negative samples, as well as to remove some samples with incorrect pseudo labels. Finally, the classifier is trained using the obtained dataset.