Abstract
User preferences acquisition plays a very important role for recommender systems. In a previous paper, we proposed a critique-based mobile recommendation methodology exploiting both long-term and session-specific user preferences. In this paper, we evaluate the impact on the recommendation accuracy of the two kinds of user preferences. We have ran off-line experiments exploiting the log data recorded in a previous live-user evaluation, and we show here that exploiting both long-term and sessionspecific preferences results in a better recommendation accuracy than using a single user model component. Moreover, we show that when the simulated user behavior deviates from that dictated by the acquired user model the session-specific preferences are more useful than the longterm ones in predicting user decisions.