I’ve just joined the competition late in the game. This is my first Data Driven competition; I’ve done a few Kaggles.
On Kaggle I know there are a lot of openly shared notebooks an insights, but I’m not sure they do that here (?). Just thought I’d double check if I’ve missed anything that’s been revealed thusfar.
I’ve tried a few interactions of key variables, but nothing really jumps out. I noticed that - at least in the training data - the original row order seems important, but I’m not sure if what information that might be a proxy for or if it’s generalizable to the scoring data set (I haven’t made a LB submission yet to test it out that way, but it seems to hold up someone in cross-validation and hold out samples).