Performance Analysis of Cyberbullying Threats in Social Media using Machine Learning Techniques

Cyberbullying is the use of internet media to harass, threaten, or hurt others, has become a widespread problem on social media platforms. As social media becomes more prevalent, so does cyberbullying, which hurts victims' mental health and well-being. This study examines the performance of multiple Machine Learning (ML) algorithms in detecting cyberbullying tweets on social media. Racism and sexism-based Twitter dataset is taken for study obtained from Kaggle repository consisting of 28,357 samples with 5 attributes. The dataset which consists of raw tweets applied for data preprocessing using a Natural Language Toolkit (NLTK). To detect and classify cyberbullying (sexism/racism) tweets, five Supervised Machine Learning (SML) models are developed: a Naive Bayes, a Random Forest, a K-Nearest Neighbor, a Support Vector Machine, and a Decision Tree model for classification. Precision, Accuracy, F1-Score, Recall, Mean Squared Error (MSE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and R-squared (RS) are used to evaluate the designed models. Based on comparison of the models, it is discovered that the Random Forest and Support Vector Machine algorithms outperforms with maximum accuracy of 97% in identifying cyberbullying tweets.