Everyone nowadays uses social media to comment and express themselves freely. Facebook, Instagram, YouTube, Snapchat, and Twitter are just a few examples of social networking platforms where users may instantly share photos and videos throughout the world. Regrettably, these additionally, as a result of technology advancements, cyberbullying has increased, leading to despair and even suicidal thoughts in children, adults, and both. Recent events have demonstrated the power of a single tweet to sway opinion, the market, and even politics. People now have a great deal of animosity, which they express through verbal abuse and harsh remarks. Finding a method to stop cyberbullying is essential. Through the use of a recent Twitter dataset from Kaggle, this research effort seeks to identify cyberbullying based on various demographic groups. The work also made use of SMOTE for data balancing and TF-IDF for feature extraction to achieve better results than the baseline strategy. Classical supervised machine learning algorithms, BiLSTM and BERT models, were applied on pre-processed dataset. The performance of the various models is evaluated using metrics. SVM and deep learning models fared better than the other models. The study also explains how age, gender, religion, ethnicity, and other forms of cyberbullying and non-bullying interact in prediction models.