2

I split my data into training set and test set and am running linear regression on it. I am using Python's "scikit" library and I am getting an $R^2$ score of 0.31 and an RMSE value of 0.037. The response variable in my regression consists of values between 0 and 1 so the RMSE value is giving me an error rate of 3.7%. However, the low $R^2$ score indicates that the model is not working well. I'm not sure if I'm misunderstanding these metrics and not sure how to interpret these results. I would appreciate any guidance.

Richard Hardy
  • 54,375
  • 10
  • 95
  • 219
covfefe
  • 1,089
  • 2
  • 10
  • 9
  • http://stats.stackexchange.com/questions/38631/rmse-vs-coefficient-of-determination This thread might help. – tho_mi Dec 05 '15 at 21:15
  • 3
    Depending on the task, an R^2 of 0.31 may be pretty good. – George Dec 05 '15 at 21:28
  • 1
    What makes that $R^2$ "poor"? What makes some larger $R^2$ "good"? – Glen_b Dec 06 '15 at 04:43
  • *"The response variable in my regression consists of values between 0 and 1 so the RMSE value is giving me an error rate of 3.7%"* - I don't think that's a valid interpretation of RMSE... Have you tried just plotting the data? – naught101 Sep 02 '16 at 05:50

0 Answers0