Cyberbullying Prevention Mechanism: Integrating AI and Social Media Platforms
B. Jaison, D Rajashekar, R. Gayathri, V. Sabaresan, Julia Oinam, G. Yamini · 2024
Background: Originally used in the modern era to describe the behavior of one person or a small group purpose-fully and forcefully harassing a victim using technology meth-ods, the word cyberbullying (CB) first surfaced. Given CB, which is growing increasingly common, the victim could experi-ence either psychological or physical agony. Though very few studies on these technologies have been done, this creates the av-enue for automated detection systems. This is thus because low datasets or significant feature reduction in CB identification have made these investigations impossible. Methodology: This work presents an integrated model combining engines of social media engine feature extraction and classification. For this model, raw text is the input. Seeking for normal behaviors (CBs), the feature extracting engine considers context, psycho-logical aspects, user comments. The classification engine of a re-current neural network (RNN) orders an output. An evaluation system that either rewards or penalizes the output classified helps one to obtain a categorization. Reinforcement learning (RL) is performed to support the evaluation since it increases classification performance. Findings: To find whether the RNN-RL model produces accurate results, several measures—includ-ing f-measure, recall, precision, and accuracy—will be com-pared to each other. Simulations will assist us success in this re-gard. The results of the simulation show that in degree of classi-fication accuracy the RNN-RL beats traditional machine learn-ing classifiers.