You wouldn't want to do this when using a tree-based model, right? There is a lot of information lost from binning, and the learning algorithm can't find the best split points in the original data.
I can see the utility for a linear model, as we are getting additional degrees of freedom and we can inject domain knowledge, but the split points should be meaningful.
You wouldn't want to do this when using a tree-based model, right? There is a lot of information lost from binning, and the learning algorithm can't find the best split points in the original data.
I can see the utility for a linear model, as we are getting additional degrees of freedom and we can inject domain knowledge, but the split points should be meaningful.
Still over my head, but beginning to see patterns. Thanks love the learning aspect, so much appreciated.