Combining Topology-Based & Content-Based Analysis for Followee Recommendation on Twitter

Abstract

Twitter has become an important social platform for individuals and people share a high number of information about their personal lives, interests and viral news during emergencies. As of 2014, Twitter has 240 million active users and approximately 500 million tweets are shared every day. This information overload in Twitter has become a serious problem due to the growing volume of messages and increasing number of users. Recommender systems help to overcome this challenge.

Finding interesting users and getting useful information from micro-blogging sites has become difficult since the mass of the data contains irrelevant messages, promotions and spam. In this thesis we propose a followee recommender system to overcome this problem. Recommendation in Twitter has been studied by several researchers and promising results have been achieved. In this thesis, we combine topological approaches and content- based analysis within the scope of English and Turkish language to find relevant followees for Twitter users. We propose seven different strategies by using different aspects of Twitter. Personalized recommendations have been generated for 22 active Twitter users. In order to increase effectiveness of recommendations, real Twitter data has been used. The experimental results show that using retweet data gives better recommendations than favorite data and we have achieved 0.79 success rate when we combine the topological features of Twitter.