In the last decade, collaborative filtering approaches have shown their effectiveness in computing accurate recommendations starting from the user-item matrix. Unfortunately, due to their inner nature, collaborative algorithms show their limits when they deal with sparse matrices and, in these cases, encoding user preferences only by means of past ratings may lead to unsatisfactory recommendations. Hybrid approaches have been proposed to cope with this issue by exploiting side information about the items within the catalog. In this paper, we propose to exploit past user ratings, and Linked Open Data to evaluate the relevance of every single feature within each user profile thus moving from a user-item to a user-feature matrix. Here, each value is a pair representing both the popularity of the feature in the user profile and its estimated rating. We then propose two computationally efficient content-based approaches (CB-fsr, and Jaccard-fsr) and a hybrid one (Hybrid-fsr), that make use of matrix factorization techniques to compute recommendations. The evaluation has been performed on three datasets referring to different domains (movies, music, and books) and experimental results show that the proposed method outperforms state of the art approaches in terms of accuracy, novelty and diversity of results.
If you publish research that uses CB-fsr, Jaccard-fsr or Hybrid-fsr please use:
This work is currently under review
The full paper describing the overall approach WILL BE available here PDF
This algorithm has been developed by Vito Walter Anelli while working at SisInf Lab under the supervision of Tommaso Di Noia.
Tommaso Di Noia, tommaso [dot] dinoia [at] poliba [dot] it
Vito Walter Anelli, vitowalter [dot] anelli [at] poliba [dot] it