Detecting Violent Extremists in Social Media using Machine Learning Approaches

Abstract

In order to find an efficient way to detect the violent-radical accounts in social media networks. A new-labeled Arabic ISIS related tweets and writings dataset was gathered from various sources, annotated cleaned, normalized and firmly analyzed in order to prepare a proper dataset that can undergo various machine-learning algorithms in experiments to create Pro-ISIS content/twitter-accounts detectors which are mandatory to the proposed autonomous online ISIS community detector. The thesis is organized as follows: Chapter 1 gives an introduction to the research in this thesis. Chapter 2 discusses some topics that are relevant to this thesis subject such as radical extremism, social media and machine learning besides addressing some of the latest researches related to the thesis work. Chapter 3 clarifies dataset gathering, preparing and analysis. Chapter 4 presents the proposed architecture, which is used to achieve the detection system. Chapter 5 presents implementations and the results for different experiments to reach the best detection accuracy. Finally, Chapter 6 gives a conclusion for the thesis’ work, and potential directions for future work.

Keywords: Machine learning, ISIS, Daesh, Extremism, Data mining, Social media, Twitter.