Next-POI Recommendations for the Smart Destination Era

Abstract

A novel Recommender System exploiting behavioural users' data in order to identify and recommend relevant and novel points of interest (POIs) is here presented. The proposed approach applies clustering to users' sensed POI visit trajectories in order to identify like-behaving users and then it learns a distinct behaviour model for each cluster. The learnt behaviour model is used to generate novel and relevant recommendations for next POI visits that optimise the user reward, which is inferred from the data. In a live user study it is assessed, along different dimensions, how users evaluate recommendations produced by the proposed method in comparison with a traditional one. The results illustrate the differences between the compared approaches and the benefits of the proposed one.