kwinkunks
commented
7 years ago Hello! Thank you for the PR, and welcome to the contest! Always good to see another team.
The CSV scores 0.444 against the blind data. But... I noticed that your model predicts a lot of facies 9 and not much of facies 8; if I swap those predictions around, I get 0.539. In fact, if I just turn all the 9's into 8's I get 0.544. So I wonder if something happened with y somewhere to get these muddled up?
I'll hold off entering a score for a bit, until you have a chance to look it over and see if you can spot anything.
Cheers! Matt
Find attached ipython notebook and csv file. First attempt at something like this so hopfully it works!
Scroll all the way to the bottom for a section called "Reproducibility and result." Using the last couple of cells the submission should be recreated. It looks like you will need to enter a new path to load the training data and test data files.
CSV currently contains "predicted_y", which is the result.