Situation-aware multi-criteria recommender system: using criteria preferences as contexts

Recommender systems (RSs) have been successfully applied to alleviate the problem of information overload and assist users' decision makings. Multi-criteria recommender systems is one of the RSs which utilizes users' multiple ratings on different aspects of the items (i.e., multi-criteria ratings) to predict user preferences. Traditional approaches simply treat these multi-criteria ratings as addons, and aggregate them together to serve for item recommendations. In this paper, we propose the novel approaches which treat criteria preferences as contextual situations. More specifically, we believe that part of multiple criteria preferences can be viewed as contexts, while others can be treated in the traditional way in multi-criteria recommender systems. We compare the recommendation performance among three settings: using all the criteria ratings in the traditional way, treating all the criteria preferences as contexts, and utilizing selected criteria ratings as contexts. Our experiments based on two real-world rating data sets reveal that treating criteria preferences as contexts can improve the performance of item recommendations, but they should be carefully selected. The hybrid model of using selected criteria preferences as contexts and the remaining ones in the traditional way is finally demonstrated as the overall winner in our experiments.