Classification of metaverse-based sexual harassment and bullying risks with machine learning algorithms

The Metaverse is a network of virtual worlds that combine virtual reality (VR) and augmented reality (AR) to create immersive experiences. It represents the blending of real and virtual worlds and is considered the next evolution of the internet. The development of the Metaverse has been influenced by literary works, innovations in games and developments in virtual platforms. As the Metaverse gains popularity in sectors such as health, education, finance and gaming, it also brings risks such as financial crimes, data theft, sexual harassment and bullying. Measures such as authentication systems and user education can help mitigate these risks. Harassment and bullying, including physical, psychological, verbal, sexual and digital forms, is a concern in both Metaverse and real-world environments. It is therefore important to identify the risks of sexual harassment and bullying on Metaverse platforms. This study aims to predict and classify individuals' perspectives on sexual harassment and bullying incidents on Metaverse platforms, and machine learning methods were applied for the analysis. The dataset consists of survey responses obtained from associate, undergraduate, graduate and doctoral students at Aksaray University. Various metrics (accuracy, precision, sensitivity, etc.) were used to evaluate the analysis. In this context, the Logistic Regression algorithm demonstrated the highest accuracy (86.27%), AUC score (90.51%), precision (77.84%) and F1-score (80.98%) in identifying students who perceive harassment as a problem.