Klasifikasi Komentar Online Untuk Deteksi Dini Cyberbullying Menggunakan Metode Support Vector Machine

This research focuses on addressing the issue of cyberbullying prevalent in YouTube and TikTok comment sections, which has become a significant threat to the mental health of internet users in Indonesia. The high volume of comments along with the use of slang, abbreviations, and non-standard terms on both platforms is the primary reason this research developed an early detection system using the Support Vector Machine (SVM) algorithm. By utilizing 8,000 comment data collected in a balanced manner through web scraping techniques, the system applies text cleaning processes and the TF-IDF method to accurately recognize word patterns. The testing results show that this model achieved optimal performance at a 70:30 data split, with an accuracy rate of 80.75% and a precision of 83% for the bullying category. These findings prove that the synergy between the SVM algorithm and the TF-IDF method is highly effective in handling the complexity of social media language and can be relied upon as an automated tool to prevent the spread of bullying in the digital space.