如何在多类分类中使用Univariate roc_auc选择特征

Question

我正在上在线课程Feature Selection in Machine Learning。导师教我如何选择我清楚理解的二进制分类中的特征。伪代码如下

- First, builds one decision tree per feature, to predict the target
- Second, makes predictions using the decision tree and the mentioned feature
- Third, ranks the features according to the machine learning metric (roc-auc)
- Fourth, selects all features with roc_auc>0.5

尽管，他没有提到多类分类的roc_auc条件，但我认为应该将其概括如下

Fourth, selects all features with roc_auc>(1/#classes to predict)

有人能在这方面提供更多见识吗？