An Evaluation Methodology for Collaborative Recommender Systems
Paolo Cremonesi, Roberto Turrin, Eugenio Lentini, Matteo Matteucci · 2008
Recommender systems use statistical and knowledge discovery techniques in order to recommend products to users and to mitigate the problem of information overload. The evaluation of the quality of recommender systems has become an important issue for choosing the best learning algorithms. In this paper we propose an evaluation methodology for collaborative filtering (CF) algorithms. This methodology carries out a clear, guided and repeatable evaluation of a CF algorithm. We apply the methodology on two datasets, with different characteristics, using two CF algorithms: singular value decomposition and naive bayesian networks.