I am using XGBoost as a classifier, and one of my important variables has values ranging from 500 to 20000. In the training data, there are very few observations where this variable is above 15000 AND the label given is 1. (2 successes, 400~ failures, compared to the entire data's 900 successes, 7900~ failures.)

This variable is something that a customer chooses the value for, and it is likely that they will choose a value within this range (15000 to 20000) so my main question about this is, in a generalisation setting, would tree-based algorithms such as XGBoost just almost always classify observations within this range as a 0?



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.