In follow up to my previous question, I was able to label survey responses based on 70 categories using RTextTools. Each survey gets one label while if manually coded some would have up to 8 labels.

The good news is my training data set does have multi-labels. So my question is, would be possible to train on 8 labels all at once? And then on the test data, get probabilities associated with each label and sort based on probability to get the top 8 labels?

Or can anyone suggest a better approach?

  • $\begingroup$ Comment: It looks like the mlr package in R allows you to perform multi-label classification for anyone who is interested. My personal challenge is in data preparation and if anyone has any tips it would be very useful. Thanks $\endgroup$
    – econstat
    Feb 8, 2016 at 19:13


Your Answer

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