Recommender systems model user preferences by exploiting their profiles, historical transactions, and ratings of the items. The quality of the recommendations heavily relies on the availability of the data. While typical recommendation methods such as collaborative and content-based filtering can be effective in a wide range of online shopping and e-commerce applications, they suffer from the cold-start problem in settings where new users enter the system and ratings are sparse for new or low-volume items. To this end, we present a pairwise association rule-based recommendation algorithm that builds a model of collective user preferences by utilizing mined associations at both the item and the category levels. In the meantime, the model allows an individual user’s in-session activities to be integrated at the category level to further improve the recommendation quality. Experimental results show that the proposed method improves recommendation performance, as compared to similar approaches.