hi @rmwkwok
can you please elaborate more on what you mean here? what do you mean by trying higher order features? like polynomial features?
and especially when there are too many features, how do you choose the features for higher order?
quote=“rmwkwok, post:2, topic:255475”]
However, it is more usual that we cannot inspect the dataset in your way when the number of features are large, then we would need to try some higher-order features, and then evaluate how much better those extra features are delivering to us.
[/quote]
hi @Christian_Simonis
what do you mean by applying transformation ? what does it involve? and how do you do it?
I mean: polar coordinate transformation as described in this post where I hope your question is fully answered: Using correlation matrix for feature selection - #9 by mehmet_baki_deniz
All the info can be found in this thread if you just scroll up! 
Best regards
Christian
1 Like
Hello @mehmet_baki_deniz,
It simply means to, for example, add some higher order term like x_1^2 for x_1, then train a model, then evaluate it with a predefined metric (such as squared loss) on a cv dataset, then we compare to find out whether it does a better job before or after the additional feature.
Raymond
2 Likes