Next Point Of Interest (POI) category prediction: An overview of predictive approaches

Summary

Having reached 2020, the use of smartphones seems to be one of the most dominant daily activities in terms of time. User activity is tracked automatically from various mobile applications (asking for user’s consent) or it is actively reported by users who offer their location footprint in order to get back rich location based services. Vast amounts of accumulated location data is used to acquire useful insights regarding user’s activity as well as for recommendation purposes. In certain location-based social networks (LBSNs) such as Facebook, Foursquare (Swarm), etc. users are checking-in to various places for social interaction purposes or in order to keep a personal visit history. Daily, a trajectory is formed for each user containing all the places he visited along with a corresponding category based on a predefined taxonomy (e.g. transport, restaurant, park, etc.). This thesis focuses on the task of predicting the next POI category to be visited daily based on historical check-in data.