Abstract: A collaborative-filtering based combination model is used to select advertisements to serve to a current network user. The combination model comprises a plurality of vectors, each vector comprising ratings for entertainment items and ratings for advertisements (or advertisement categories) from a previous network user. A vector may also include ratings for demographic information of the previous network user. Upon receiving one or more entertainment item ratings (and optionally demographic ratings) of a current user, the combination model is used to select one or more advertisements to serve to the current user. A representative vector in the combination model is determined having entertainment item ratings (and optionally demographic ratings) that are most similar to the corresponding entertainment item ratings (and optionally demographic ratings) of the current user.