0
$\begingroup$

I was working on a dataset about some credit card fraud detection. The box plot for the "Transaction Amount col." showing a lot of outliers. Can anyone suggest what to do? Should i cap the outliers, remove them or any ML model that's robust to outliers.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
0
$\begingroup$

Are there fraudulent transactions in these outliers? If so then removing them would be bad as your model would be missing these points to train on.

If the outliers are important and deviate a lot, you could use scaling to have the outliers not impact the model as much and still retain all data.

You could also separate the outliers, treat them as a second dataset and build a second model. Then use both models together in an ensemble

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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