在将分类变量转换为虚拟变量后，如何从 sklearn api 中找到特征重要性？

Question

After we have converted our categorical variables to dummy variables for training the model. We tend to find feature importance. But sklearn's model.feature_importance_ object returns feature impotance for every dummy variable, rather than the original categorical variable. How to fix this?

Answer 1

Because the dummy variables are used to train the model, you cannot find the importance of the original categorical variable. It is mathematically impossible thing.