ANASTASIA : recommandation de séquences d'activités spatiotemporelles

As amount of activities available for users and their variety have grown, personalised recommendation of activities sequences has become an important challenge. However, most of recommender systems do not consider temporal constraints of activities, making the recommendation hard for user to follow. In this article, we describe a novel approach for recommendation of competing activities limited in time. It makes use of historical records of users’ activities in order to mine users’ behavioral patterns, and combines different contextual elements (popularity, demographic and spatio-temporal information). We present an evaluation framework and a dataset that will allow us to evaluate our approach. MOTS-CLÉS : système de recommandation, séquence d’activités, score personnalisé, contexte d’utilisateur.