My solution - thoughts on using R2?

Pretty pleased with my solution to this one!

https://www.kaggle.com/crimsoneer/bike-hire-prediction/

Interestingly, I note a couple of things:

  1. we threw every feature at the model - I’m surprised we don’t work up to that
  2. We tend to use MSE instead of R2 or adjusted R2, which I’m more comfortable using

Given our error with the test set is so low, I assume that’s not a problem, but found it interesting none-the-less?

1 Like