Experimenting Datasets and Machine Learning Techniques for Enhancing Cyberbullying Detection
May Fen Gan, Hui Na Chua, Muhammed Basheer Jasser, Richard T.K. Wong · 2023
Social media has become an essential aspect of everyday life for most individuals worldwide and continues to expand. In social media, short text communication has become more common, and the ease of accessibility has also contributed to the increase in cyberbullying cases. Numerous research studies have been conducted by researchers in cyberbullying detection using different datasets and machine learning methods for cyberbullying prediction. However, different data characteristics may influence the predictive model performance, and there is a need to examine how the impact is. Therefore, this research aims to compare different datasets, including merging existing ones with different machine learning techniques, shedding light on their relative effectiveness and contributing to a deeper understanding and better decision-making of dataset usage and techniques deployment. Our research finding provides scientific evidence that combining various cyberbullying datasets to create balanced data and Bi-LSTM and BERT improve prediction model performance.