Hi,
Solved, March/22. Details in my other post.
=== Old post ==
cool group and lots of fascinating stuff to discover. I just started working on a time series project with a fairly naive approach to replicate the “adult” tabular data example.
The data pre-processing and model building was straightforward but obviously, the prediction didn’t worked out.
Somebody pointed out to use LSTM or GRU instead but I just can’t find a working example in fast.ai. Can anyone help me to find a starting point on how to do a minimal time series model in fast.ai?
Thank you
Marvin
Gist:
Details here:
https://forums.fast.ai/t/auto-regression-with-tabularmodel-zero-accuracy-why/