Recommender systems are widespread services behind user oriented applications.
They serve to help the users to find the proper movie, music track or other products.
Recommender algorithms use information about the historical behavior of the users and items and their interactions to make the best estimation for the users current preferences.
Context-aware recommender systems use not only user-item interactions, but the circumstances of these events to make an even better recommendation.
The main focus of our research is to improve the performance of current recommender systems using contextual information.
In this thesis, we present our results based on [Pálovics, Szalai,Kocsis, Pap, Frigó, Benczúr 2015] extended with further models and experiments.
In our work, we focused on the usage of geolocation information in recommenders.
We present location-aware models using GPS coordinates and attribute based models using various type of meta information about the items.
In order to evaluate our models, we made our experiments on the publicly available 30M Last.fm dataset and on a Twitter crawl that contains geolocation information about each tweet.