I have a project to model users with characteristic tags (e.g. runner, cyclist, swimmer, vegan, pianist) in order to correlate user behaviour to these labels. Obviously a user can have multiple characteristics (non-exclusive) and so the distance (or similarity) between users is determined by the amount of tag overlap.

  1. Is there a name for this kind of model?
  2. Are there any particular patterns/algorithms that are useful for measuring correlations in data represented by this kind of model?

I don't need a complete list; just to be pointed in the right direction.

  • $\begingroup$ You can also try Fuzzy Set Theory, which allows for partial membership to non exclusive classes. $\endgroup$
    – steadyfish
    Commented Feb 26, 2013 at 2:41

1 Answer 1


In answer to your first question, I think you are looking for the jaccard distance.

I'm not aware of any particularly special methods for this metric, but it is a proper metric so most methods should be fine. Hopefully knowing the name will help!


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.