Reversing/undoing categorical embeddings on structured prediction results

(Daniel Jue) #1

Hi, I’ve been following along with Lesson 3 Rossman structured data predictions, on my own analogous data.

I had to make my own ‘test’ dataframe from real data, which also went through the categorical embedding process like the training and validation df did.

Now that I’ve trained and created some predictions, how do I undo or reverse the categorical embeddings, so that the predictions df has the original string values?

I’m probably not using the right search terms, I’m sure that it’s covered some where but I haven’t found it for a few hours.