A User-centric Evaluation of Recommender Algorithms for an Event Recommendation System

While several approaches to event recommendation already exist, a comparison study including di↵erent algorithms remains absent. We have set up an online user-centric based evaluation experiment to find a recommendation algorithm that improves user satisfaction for a popular Belgian cultural events website. Both implicit and explicit feedback in the form of user interactions with the website were logged over a period of 41 days, serving as the input for 5 popular recommendation approaches. By means of a questionnaire users were asked to rate di↵erent qualitative aspects of the recommender system including accuracy, novelty, diversity, satisfaction, and trust. Results show that a hybrid of a user-based collaborative filtering and content-based approach outperforms the other algorithms on almost every qualitative metric. Correlation values between the answers in the questionnaire seem to indicate that both accuracy and transparency are correlated the most with general user satisfaction of the recommender system.