So I have a dataset where I have a continuous variable for only about 10% of the entries. How would you incorporate this in a model. Imputing does not make much sense to me, because there are so few values, however for those entries where I do have it, it is quite an important feature.

  • $\begingroup$ What kind of models do you use? The best answer might vary depending on whether you use a decision tree based model, some kind of linear approach or something else. $\endgroup$ – NiklasvMoers May 29 '20 at 10:30
  • $\begingroup$ Currently I use Catboost. But if there are other models/packages I would use them instead $\endgroup$ – Hans Geber May 29 '20 at 10:48
  • $\begingroup$ Definitely try out @Uday's approach. Another one would be to use two seperate models: One for data points with and one for data points without a missing entry. $\endgroup$ – NiklasvMoers May 29 '20 at 10:53

Where ever you don't have an entry just make that to a null value or a random unique value. an then create another column, make that column data equal to 1 if data is present 0 otherwise so, the model may learn about that unique imputation and data using another column we are adding.

  • $\begingroup$ thats what I was thinking too, but is this the only game in town? $\endgroup$ – Hans Geber May 29 '20 at 10:51

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.