Friends Recommendation Algorithm Based on Graph Mining and Collaborative Filtering

Abstract

As an open, free, flexible social platform, microblog develops rapidly recent years. State of art research on friends recommendation has attract both industrial and academical concerns. Compared with traditional social networks, microblog contains both strong social relations based on the real relationship, and weak social relations based on interests, locations and other incidental factors. How to utilize these relationships and characters in personalized friends recommendation is still under research. This paper presents a new hybrid recommendation model, considering both the relationship strength and interest similarity in microblog, using the social graph mining algorithm to find strong social relations and the item-based collaborative filtering algorithm to mine weak social relations. Experimental results show that the proposed hybrid algorithm outperforms the traditional algorithm.