Resource recommender system based on tag and time for social tagging system

Recently social tagging systems are increasingly becoming popular. These systems allow users to manage, organize and search their required resources. Most of the available modules in recommender systems, don't consider information such as time that could be effective on the users preferences of a resource. So, one of the existing challenges in resource recommendation systems is the appropriate integration of tag and time information to discover user's taste considering time drift and improving accuracy. In this paper data from social tagging systems are modeled with four-mode tensors to capture four-way correlations between users, items, tags and time information. Then by applying multi-way analysis, latent correlations among users preferences based on time information are revealed, which help to improve the quality of recommendations. The advantages of proposed system are that it considers simultaneous relations among elements, discovers latent associations and improves the quality of recommendations accuracy. The experimental results on a real-world dataset, Citeulike, indicate significant improvements in recommendation quality and better performance compared with the existing systems.