Cyberbullying, a pervasive issue in online social networks, can have catastrophic effects on a person’s mental health and overall well-being. This study analyzes the feasibility of applying different machine learning methods to automatically detect cyber bullies in social media text. We evaluated multiple machine learning models and thoroughly validated them on a large-scale social media interactions dataset. Our results indicated that machine learning has tremendous potential to fight cyberbullying especially when fine-tuned through an exhaustive grid search. With regularization strength (so-called ‘C’ value) 10 and ‘liblinear’ solver the model realized an F1 score of 0.9825610022978444. We also examined the results of the grid search and noted that the higher the regularization strength the better the results. This could be indicative of overfitting being present therefore requiring some regularization for increased generality. Also, ‘liblinear’ performed better than ‘lbfgs’ for larger ‘C’ values. Although our work was conducted on a particular set of data, our results are encouraging and show that machine learning holds great potential for constructing precise automated detection systems that are robust. Such systems enable social media platforms to take preemptive action against cyberbullying events, making the online space safer and more accessible for everyone.