Cyberbullying Detection System on Social Media Using Supervised Machine Learning

The use of digital and social media is growing every day as technology advances. People in the twenty-first century are growing up in a social media and internet-enabled society. Digital media offers a lot of opportunities, but people frequently tend to misuse them. On social networking sites, people spread anger toward a person. People are affected by cyberbullying in various ways. It has an impact on more than just health; numerous other factors put life in danger. Cyberbullying is a widespread modern phenomenon that people cannot completely avoid but can prevent. The author proposes a system for automatic cyberbullying detection and prevention using supervised machine learning. The system considers key characteristics of cyberbullying, such as the intention to harm, repeated behavior, and the use of abusive language. Support vector machines and logistic regression are employed to identify cyberbullying and related themes/categories such as race, physical, sexuality, and politics. This proposed method offers a novel theory for the detection of cyberbullying: texting has evolved over time due to changes in context usage, and language. In the dataset that includes tweets, Support Vector Machine (SVM), Naïve Bayes, and Logistic Regression (LR) models were tested along with different Natural Language Processing methods. The accuracy of the system is improved by sentiment analysis, N-gram analysis, and other non-traditional feature extraction methods like Term Frequency-Inverse Document Frequency (TF-IDF) and profanity detection.