<![CDATA[Social media usage has been steadily increasing, with Instagram emerging as one of the most prominent platforms. As of January 2023, Instagram had 1.318 billion users, predominantly aged 18-24. While teenagers report enhanced self-confidence and diminished feelings of loneliness, social media also facilitates cyberbullying, impacting 35% of adolescents with low emotional well-being. This research seeks to develop a model for detecting cyberbullying in Instagram comments, classifying them into negative, positive, and neutral categories using SVM, Naïve Bayes, and Random Forest algorithms. The methodology encompasses data collection, preprocessing, text transformation via TF-IDF, and a comparative analysis. Grid search is employed to optimize algorithm parameters. Initial results indicated that Naïve Bayes and SVM achieved an accuracy of 75.47%, while Random Forest reached 69.88%. Following parameter tuning, SVM's accuracy improved to 97.79%, whereas Random Forest's decreased to 66.51%. The findings underscore the superior performance of SVM with parameter tuning in detecting Instagram cyberbullying.]]>
Copyrights © 2024
