Evaluating the Effectiveness of an Ensemble Random Forest Machine Learning Algorithm in Detecting Cyberbullying in the 4chan Politically Incorrect Board Social

Abstract

The United States has seen a drastic increase in the occurrences of cyberbullying. Children are often on the receiving end of this horrible phenomenon. The topic of cyberbullying common area of research; however, the body of research on the automated detection of cyberbullying on social media using ensemble learning is still in its infancy. The purpose of this study was to determine if a random forest ensemble learning method is effective at the identification of cyberbullying on 4chan Politically Incorrect social media message board. 4chan is a unique social media platform where most members post anonymously and post without fear of retribution. The use of 4chan in this study represents an opportunity to research cyberbullying on social media platforms beyond those typically studied, such as Twitter and Facebook. A structured experiment was conducted. A labeled dataset was created and trained using a learning curve method to develop an ensemble learning model for the automated detection of a cyberbully. Feature extraction was performed using term frequency-inverse document frequency (TF-IDF) and the model training was performed using a random forest classifier. The experiment result indicated that the ensemble learning algorithm proves to be an effective tool in detecting cyberbullying on 4chan. The performance of the model trained with a random forest classifier was reported as 87% precision, 84% recall, 81% F1-score, and 83% accuracy.