Abstract: A technique is disclosed for ranking venues and for providing a user with one or more of the ranked venues. To do so, a data processing system trains a model during a first, training sequence, by using a training dataset of past check-ins that were made by users visiting venues. The trained model, representing a learned hypothesis, represents venues that are ranked according to their relevance to users who are also represented by the model. During a second, run-time sequence, the system receives real-time geolocation data that represents users and generates a set of candidate venues for a user by determining venues that are nearby the geolocation of the user. The system then ranks the generated candidate venues by applying the learned hypothesis. One or more of the ranked venues can be provided to a user's wireless terminal, for example, where they can be presented as venue recommendations.