Our median values range : 120K-265K, and avg.error is 157 K (result of RMSE).

How I can interpret that is overfitting, underfitting , or a good model?

I thought if I divide as like that: 157000/120000 and 157000/265000...I can get some inferences from them.

from sklearn.metrics import mean_squared_error as mse
inst_pred = lin_reg.predict(instance_prepared)

inst_mse = mse(instance_label, inst_pred )

inst_rmse = np.sqrt(inst_mse)


(It is cited from Hands-on ML book.)


1 Answer 1


No you can't, the value alone is meaningless. What you need is to compare the performance on the training test to performance on test set, that could give you some idea about potential overfitting. As about general model quality, to interpret this number you would need to compare it to performance of another model, the most trivial one would be to predict the mean for all the observations.

See the following threads for more details: How do you Interpret RMSLE (Root Mean Squared Logarithmic Error)?, How to interpret root mean squared error (RMSE) vs standard deviation?, Overfitting and Underfitting, and Techniques to detect overfitting.

  • $\begingroup$ By the way, it was from Hands-on Machine Learning Book. And there was no reason why it is overfitting. Thank you, Tim. $\endgroup$
    – Zehra N.
    Jan 28 at 13:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.