Characterizing and Detecting Similar Twitter Users

The vast emerging and growth of Twitter raises the need to identify similar users. This may benefit in a variety of applications such as social recommendations, recruitments, and even for users profiling for security reasons. Several methods and techniques have been proposed to discover similar users that either integrate one measure or combine several ones. In this work, we investigate the measures of different existing frameworks and propose new measures that enhance the similarity detection precision. Our measures involve the use of both, the social graph and user profile contents. We build several classifiers based on real Twitter data and prove better performance of our measures as compared to existing ones.