A NEW ALGORITHM FOR MULTI-MODE RECOMMENDATIONS IN SOCIAL TAGGING SYSTEMS

Social tagging is one of the most important characteristics of web 2.0 services. Different from traditional recommendation algorithms, in social tagging systems, recommendation algorithms involve the ternary relations between users, items and tags. And algorithms that support integrated multi-mode recommendations are very appealing. We propose a multi-mode recommendation algorithm based on higher-order singular value decomposition, and our algorithm handles not only the existing triplets {user, item, tag}, but also the pairs {user, item} with no tags in social tagging system. Meanwhile. We propose a measure for user recommendations. We empirically show that our algorithm outperforms a state-of-the-art algorithm for multi-mode recommendations with a Last.fm dataset.