Many existing state-of-the-art top-N recommendation methods model users and items in the same latent space and the recommendation scores are computed via the dot product between those vectors. These methods assume that the user preference is consistent across all the items that he/she has rated. This assumption is not necessarily true, since many users can have multiple personas/interests and their preferences can vary with each such interest. To address this, a recently proposed method modeled the users with multiple interests. In this paper, we build on this approach and model users using a much richer representation. We propose a method which models the user preference as a combination of having global preference and interest-specific preference. The proposed method uses a nonlinear model for predicting the recommendation score, which is used to perform top-N recommendation task. The recommendation score is computed as a sum of the scores from the components representing global preference and interest-specific preference. A comprehensive set of experiments on multiple datasets show that the proposed model outperforms other state-of-the-art methods for top-N recommendation task.
|Original language||English (US)|
|Number of pages||8|
|Journal||IEEE International Conference on Data Mining Workshops, ICDMW|
|State||Published - Jan 1 2015|
- Data mining
- Database Applications
- Mining methods and algorithms