An Intelligent System for Cyberbullying Detection in Facebook Interactions
Samia Sabrin, Mymona Akter, Mahbub Rabbany Rouien, Tanvir Ahmed · 2025
One sort of online bullying that particularly happens on social media sites is called cyberbullying. Cyberbullying is used on various online platforms like social media, messaging apps or email to harm or intimidate others. Since 3.07 billion people on earth use Facebook, our model aims to ensure a safe online platform for preventing cyberbullying. Our model is all about detecting cyberbullying in an effective way on Facebook interactions. To identify cyberbullying, a large number of studies in English and other languages have been carried out. On the other hand, very little research on cyberbullying has been conducted in Bangla. In this paper, we have worked with both Bangla and English datasets. Using several approaches for cyberbullying detection, three models-BERT, BiLSTM and CNN-for deep learning were put into practice and assessed. Model strengths and shortcomings are clarified by performance analysis, which directs iterative improvements for increased accuracy. We have converted BangIa dataset into English dataset using Google Translator via Python code. Comparing all the results, the BERT model shows the highest accuracy 95% for Bangla dataset and 90% for Translated English dataset. In addition to providing insights for future study and applications in digital safety and well-being, this paper offers a strong foundation for cuberbullvina detection in online social networks.