MobHet: Predicting human mobility using heterogeneous data sources.

•We propose MobHet, a new family of data-driven models for human mobility prediction.•MobHet exploits region popularity, inter-region transitions and user contacts.•MobHet is evaluated in various scenarios of homogeneous and heterogeneous data.•Evaluation uses large real-world datasets of mobile phone calls and tweets.•Results attest MobHet effectiveness when compared to single-source baseline models.