homework results

[ptrBodnar]

Hope I am doing everything correctly)

[vikua]

Well done.

Things I liked:

• Doing custom type inference. This is very tricky part which nevertheless is super important in AML space. We never have types provided for us in real world.
• Using ordinal encoder for tree based models! Good job on solving unknown category problem
• Playing with different model types (linear, trees, gb)

Minor comments:

• Target for this dataset is loss. It is fine that you picked different column in dataset as target, but I'm not sure why doing so.
• Maybe different optimization approach would lead to better scores (e.g. Random search, Bayesian)
• [off topic] You don't need to commit files from lectures.

[ptrBodnar]

Thank you!

About comments:

1. Honestly i just missed what was the target from task setting therefore picked yhe one that looked probable.
2. Yes, I haven't done that due ti the lack of time mostly.
3. Oh, sure haven't thought about that.