Abstract: To improve the performance of recommender systems in a practical manner, many hybrid recommendation approaches have been proposed. Recently, some researchers apply the idea of ranking to recommender systems which yield plausible results. Collaborative ranking is a popular ranking based method, it regards that unrated items have lower rankings than rated items for a user. Unfortunately, the existing collaborative ranking approaches only focus on the partial associations with users and items, and thus fail to detect some features that could potentially improve the performance of the recommender systems. For this reason, these methods continue to suffer from data sparsity and do not work well for recommending an interesting item to an individual user. To address these issues, we present an Assembled Collaborative Ranking with Random Walk (ACR-RW) approach based on the combination of collaborative ranking and random walk method, which can be used to rank items according to expected user preferences by detecting both absolute and relative correlative information, in order to recommend top-ranked items to potentially interested users. On the basis of ACR-RW, we can improve the collaborative ranking approaches by adding absolute relationship information and defining the partial order relationship in assemblages rather than the global, so as to better describe and predict one user’s preference. Finally, we implement experiments on three real-world datasets, and the results show that our approach consistently outperforms all other comparative approaches, demonstrating its effectiveness for recommendation tasks.
0 Replies
Loading