Hello. I should take a look at your data to give you a correct answer to the first question.

For the second question, in an unsupervised approach, you can’t remove 2 features if you only have them. The algorithm keeps the couple of features with the lowest collinearity and then keeps adding features as long as the collinearity of the whole dataset satisfies the constraints on the threshold. For a supervised approach, the feature with the highest correlation with the target variable is kept.