Regularization is used to handle over-fitting problem.
Similarly can we have some methodology to overcome under fitting problem or merely adding new features or training data will help us in reducing under-fitting issue?
$\begingroup$
$\endgroup$
2
-
$\begingroup$ Why do you want to underfit? having better data, more features and modelling as per proper requirements $\endgroup$– AdityaCommented Mar 15, 2020 at 8:42
-
$\begingroup$ I dont want to underfit, I just want to know the methods for handling it. $\endgroup$– user91244Commented Mar 15, 2020 at 13:38
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
As such we don't have much ways to handle under-fitting issue. Generally I follow following ways to handle it:
- Add more features to training data. That includes deriving new features front the existing one.
- Make a complex model.
- Increase iterations.
Adding new training data only, I believe will not solve this problem.
-
$\begingroup$ Thanks for your response. $\endgroup$– user91244Commented Mar 15, 2020 at 13:38