An Empirical Study and Analysis of the Machine Learning Algorithms Used in Detecting Cyberbullying in Social Media

Regardless of the demography, social media has become an integral part of our everyday lives. Nowadays, it is the most popular platform people use for staying connected with their friends and family. As a consequence, the likelihood and growth of cyber threats have increased rapidly. To mitigate this situation, we proposed a system that can detect cyber crimes such as blackmail, fraud, impersonation, spam etc. from the social media network Twitter. This type of study can help people to detect early threats and possible criminal activity and the types of accounts to stay alert of in real time thereby, creating a more secure social media experience. Our main goal is to compare various sentiment analysis approaches for detecting bullying or threats from social media. We used two supervised machine learning algorithms to form a comparison and determine which among the two gives out the highest accuracy in order for us to decide how to detect cyberbullying activity on the Internet and be alert of threats in both the real and virtual world.