Trinity: Walking on a User-Object-Tag Heterogeneous Network for Personalised Recommendations

摘要: The rapid evolution of the Internet has been appealing for effective recommender systems to pinpoint useful information from online resources. Although historical rating data widely used as most important in recommendation methods, recent advancements have demonstrating improvement performance with incorporation tag information. Furthermore, availability annotations well addressed by such fruitful social tagging applications CiteULike, MovieLens and BibSonomy, which allow users express their preferences, upload resources assign own tags. Nevertheless, existing tag-aware approaches model relationships among users, objects tags using a tripartite graph, hence overlook within same types nodes. To overcome this limitation, we propose novel approach, Trinity, integrate towards personalised recommendation. Trinity constructs three-layered object-user-tag network that considers not only interconnections between different nodes but also Based on heterogeneous network, adopts random walk restart strength associations candidate objects, thereby providing means prioritizing query user. We validate our approach via series large-scale 10-fold cross-validation experiments evaluate its three comprehensive criteria. Results show method outperforms several including supervised restart, simulation resource allocating processes, traditional collaborative filtering.