While solving the questions for machine learning, I got two values for R square from 2 different regressors, i.e, 0.9999 and 0.9769. So, which should go for as both could lead to overfitting? Thanks in advance
1
-
$\begingroup$ Please provide us with more information. How many samples do you have in your dataset? What is the variance of your target variable? $\endgroup$– pythinkerCommented Apr 11, 2019 at 12:44
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
Are you asking about rmse or r-squared?
The size of rmse depends on the general range of your response feature(predicting target). The lower the better.
As to r-squared, both 0.9999 and 0.9769 are very likely to indicate overfitting. If you have to choose between these two models, my suggestion is to evaluate them using generalization error. You can also do cross-validation if no out-of-sample data are available.
