如何在scikit-learn中生成自定义交叉验证生成器？

I have an unbalanced dataset, so I have an strategy for oversampling that I only apply during training of my data. I'd like to use classes of scikit-learn like I have an unbalanced dataset, so I have an stra