I am currently working on an lbsn (localization-based social network) system and i need to predict the user's age and gender.

Every time a user enters a venue, the system creates a "check-in" with the user, the venue and the datetime.

Every venue is categorized using Foursquare Venue Categories.

The system generate a Weigthed Concept Hierarchy to represent the interest level between a user and a Venue Category.

Is it possible to predict the user's age and gender using the mentioned data?

  • $\begingroup$ You could average the neighbors demographics to obtain a user's demographics. The social network serves as a regularizer, so given a model that estimates the age, sex, etc. you want the prediction to be close as possible to that of the user's neighbors (penalize by the error). There are many papers on this; e.g., Collective Semi-Supervised Learning for User Profiling in Social Media Welcome to the site! $\endgroup$
    – Emre
    Oct 25, 2017 at 20:27
  • $\begingroup$ Ty for your comment @Emre the paper was really helpful, i didn't even know there was semi-supervised learning but i'm trying to avoid adding social networks as much as posible, because i have a very tight develop timeline. I just edited my question and i hope it's more clear the problem. Is there a way to reopen it? or should i ask again? $\endgroup$ Oct 26, 2017 at 20:39
  • $\begingroup$ Clarify and re-open. What do you mean by "adding social networks" when your question is all about them? Your problem is a joint regression/classification task. Model the joint density of the age and sex, and let the loss be the sum of the regression and classification losses. There will be a coefficient that trades off accuracy in one against the other. Or you could use a regression loss for both, but I think that's less motivated theoretically. $\endgroup$
    – Emre
    Oct 26, 2017 at 20:49
  • $\begingroup$ @Emre, The lbsn it's based on this paper link the social network they used it's meant to provide social knowledge in the form of Expert Users to recommend the best venues in a city. I'm not using a social network like Twitter or Facebook $\endgroup$ Oct 26, 2017 at 20:55

1 Answer 1


Possibly - If you frame it as supervised learning problem, you would need a dataset that has those features and is labeled with age and gender. Then you could build a classification model.


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.