Automated Detection and Analysis of Cyberbullying Behavior Using Machine Learning
Rejuwan Shamim, Mohamed Lahby · 2023
Cyberbullying is an issue that is getting worse on social media sites and online forums. It can cause serious mental health problems and even suicide. The present study introduces a novel methodology that employs machine learning techniques to identify and examine instances of cyberbullying on digital platforms. The approach utilized in this study entails the utilization of natural language processing methodologies to extract pertinent characteristics from textual data, followed by the application of diverse machine learning algorithms to categorize the messages into either cyberbullying or non-cyberbullying. The present study assesses the effectiveness of our methodology on a corpus sourced from Twitter and Reddit. Our results indicate a noteworthy accuracy rate of 0.92, an F1-score of 0.88, and a recall number of 0.89 for the identification of cyberbullying. In addition, we conduct a thorough examination of the findings and pinpoint the most notable characteristics that contribute to the occurrence of cyberbullying. The approach we propose has the potential to be utilized by social media enterprises, virtual communities, and guardians to efficiently detect and alleviate instances of cyberbullying.