0

I've been using XGBoost to predict student retention, and have been using eli5 to provide explanations for the individual predictions. For a few different reasons I decided to give deep learning a try, and it performed surprisingly well on the data. However, the explanation bit is a requirement but eli5 only does explanation for Keras on images. Is there a way to get explanations for a Keras model built for regular classification? Having skimmed the LIME paper I don't really see a reason why it shouldn't work, but it seems like it is only available for image and text classification?

My data is just plain old numerical values (gpa, age, activity, test scores etc) and I am trying to predict 0 or 1 (pass/fail).

Thanks

L Xandor
  • 1,659
  • 4
  • 24
  • 48

1 Answers1

0

LIME works for Text, Tabular and Image data. So it would work in your case as well. You should check out LIME's documentation and explore the given examples for better understanding. https://github.com/marcotcr/lime

atinjanki
  • 483
  • 3
  • 13