Cyber Bullying and Toxicity Detection Using Machine Learning
Ranjana S. Jadhav, Naman Agarwal, Srushti Shevate, Chinmayee Sawakare, Piyush Parakh, Snehankit Khandare · 2023
The increased use of online platforms for communication has made cyberbullying and toxicity detection a critical issue in recent times. This paper explores the topic of cyberbullying and toxicity detection and proposes potential solutions for identifying cyber violence and offensive language more effectively. According to the study of the algorithms in examined research papers on cyberbullying and toxicity detection, this research study presents a novel approach that achieved 90% accuracy in identifying bully text in social media comments. This is done by using machine learning algorithms such as SVM, Logistic Regression, Naive Bayes, KNN, and Random Forest, with SVM and Random Forest exhibiting the best performance. Additionally, the system improved the accuracy of identifying bully images in social media posts to 84.5% by using the MobileNetV2 model (DNN), which is superior to other approaches. The system is trained using a large, labeled dataset of text data to identify and classify different types of cyberbullying and toxic content. The findings suggest that the proposed models hold promise in detecting instances of cyberbullying and offensive content effectively. These results have significant implications for the development of cyberbullying and toxicity detection systems. The proposed approach can be integrated into various social media platforms and online communities to identify and mitigate cyberbullying and toxic content more efficiently. The study also highlights the need for continued research and collaboration among stakeholders to address cyberbullying and toxicity effectively.