Hope you're enjoying the hackathon. I threw together a quick starting notebook to help anyone wanting to get a submission up as fast as possible. It's hastily made, but should be useful in prepping the data, shaping submission files, seeing how you can get a score for F1 even with low predicted probabilities...
I'll be working on a more thorough tutorial, but figured this was most useful if shared asap. Comment with your questions! And good luck all :)
The link: https://colab.research.google.com/drive/1HjJhghj2b5JJnOFNTcojLfAicDf5QWWK
Super useful, thanks!
Heads up that the test data is dirty like a dodgy uncle (OK maybe not that dirty)