Python sklearn 中的训练集

Question

如果我在子组大小方面有一个有偏见的数据集，例如，如果我试图预测 'a','b','c' 标签，但我的数据集有 1000 个 'a' 标签、500 个 'b' 标签和 250 个 ' c' 标签，sklearn 是否对此有任何解决方法，还是会自动考虑到这一点？

Answer 1

Sklearn provides the funtionality to deal with imbalanced data itself, you have to check out with the different metrics avalaible here and use it to find the best metric for which your model have higher accuracy rates. Scroing metrics is the parameter which can help you with imbalanced datasets. Just choose your scoring while training the model according to your problem and it will considerably help to deal it with.

When dealing with imbalanced data you can always do oversampling for the classes. This process can be done in data preprocessing.

More about dealing with imbalanced data you can read here and here