0
$\begingroup$

I am working on some LTE Data say of shape (100000, 10) .

Now suppose I have a column Jitter Avg. Now Jitter occurs only for any connected call and for call which could not be setup due to some reason Jitter is not applicable.

So for those rows representing failed called setup there is no value for Jitter i.e NaN. I cannot replace it with 0 since that means good quality. All in all any value would not be applicable here.

So how to deal with such situation.

$\endgroup$
1
$\begingroup$

You can use a model that can handle NaN values. Most commonly this would be something based on decision trees. I recommend LGBM and XGBoost.

Decision trees also work well when these kind of values are set outside of the range for this feature, i.e. setting it to something like -999. This alternative is better if you want to use models from scikit-learn since NaNs are not supported there.

$\endgroup$

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.