Considering correlation retarded growth for personalized recommendation in social tagging

Due to the massive amounts of data, finding social media suited to their need is a challenging issue. To help such users retrieve useful social media content, we propose a new model of personalized recommendation system by using annotating information from relationship among users, tags, and items. However, the frequency of users' tagging has strong or weak correlation, which affects the dynamic interest mining of users. In this paper, CRGI is proposed to describe the correlation between users and tags or tags and items. Our approach has two phases, in the first phase, we describe the correlation between users, items and tags by CRGI and in the second phase, we build a tag-item weight model and a user-tag preference model on the basis of the first phase. Then we utilize the two models to find the suitable items with the highest scores. The experimental results demonstrate that the item recommendation performance is improved in both the accuracy and the diversity, and validate that the proposed personalized approach is effective for improving the social media recommendation.