Abstract
The rise in popularity of social interacting websites such as Facebook, Twitter, and Snapchat has been challenged by the upsurge of unwelcomed and troubling bodies on these systems. This includes spam senders, malware systems, and other content contaminators. It is noted that highly automated accounts with 450 tweets per day produced almost 18% of entire Twitter circulation in the 2016 U.S. Presidential election. It is also observed that those disruptive systems called bots are inclined more towards circulating negative news than positive information. This paper introduces a novel framework named Associative Affinity Factor Analysis (AAFA) designed for stance detection and bot identification. Using AAFA, the proposed framework identifies real people from bots and detects the stance in bipolar affinities. The 2016 U.S. Presidential election campaign was used as a test use case because of its significant and unique counter-factual properties. The results show that our proposed AAFA framework achieves high accuracy when compared to several existing state-of-theart methods.
