Adding the sentiment attribute of nodes to improve link prediction in social network

Link prediction is an important tool for many social media sites to find the missing and future links among users. Understanding users´ sentiment and their social relationships are potentially valuable. In this paper, two new sentiment similarity measures have been proposed and an algorithm has been designed to do link prediction by incorporating the structure and sentiment attribute of nodes. In order to evaluate the proposed algorithm, links and tweets with regard to some hot topics of 2014 FIFA World Cup Brazil are crawled from Tencent Weibo, and the sentiment distributions of crowds are analyzed for each topic. The experimental results show that the number of users with the same emotion and the sentiment distributions of crowds will influence a user to link with another user, so the sentiment attribute of nodes in social network can help to improve the performance of link prediction.