A unified framework for flickr group recommendation based on tetradic decomposition

Different from current researches on Flickr group recommendation approaches that recommend groups to either users or images, this work proposes a unified framework that recommends groups to both users and images. Four types of entities in the Flickr system (users, tags, images, and groups) are integrated into a tetradic model, and then we uses tetradic decomposition to discover the latent semantic association among these entities and recommend groups to images and to users simultaneously. The innovation of this design can be summarized as follows. 1) The design is convenient to users because many Flickr users aim to recognize not only groups in which images should be shared but also groups that might interest them. 2) Experiments proof that the consideration of the semantic relations among users, images, tags, and groups enhances the performance of both two kinds of recommendations in terms of mean average precision.