I have made regression selecting the best features (p-val<0.05). I now have the model with a 0.85 R-squared and residual of 2.64.
Well, sometimes when I try to predict some new instances, my predictions is way outside my residual deviation. I mean sometime my actual y-value is supposed to be 20 and I have 26,7.
I also fell the need to mention that when I explored my target with a boxplot; I had a maximum of 33 and outliers going up to 100 and the standard deviation for that target is nearly 7.
The good thing is, I have over a million of rows and the outliers represent about 5000 rows. Is taking them off improve my model? what can I do to improve such a model?


Your Answer

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

Browse other questions tagged or ask your own question.