A Graph Based Approach for Cyberbullying Classification Using Machine Learning Algorithms

Abstract

Using Social Network Services becomes a huge necessary, especially for teenagers. They can get closer in the real time. With more than 237 million daily active users [1], Twitter allows sharing short messages named « tweets ». Some users can employ offensive words, pictures or videos against another person via electronic devices. This phenomenon is calling cyberbullying. Because of timid, fear or weak cyberbullying attacks teens in darkness. Victims suffer in silence and can be driven to lower self-esteem, depression or suicide. In that aims many researches in computer science, sociology and psychiatry take the challenge and many initiatives are developed in order to detect offensive, abusive and negative comments.Because of the huge number of users, the tweets make a raw dataset containing a lot of information that must be purified. Some information are misspelled, missing, duplicated or inuseless. Our work is based on Natural Language Processing; we will discuss the importance of the data pre-processing and feature engineering before proceeding to the text classification using Machine Learning Algorithms.