Sentiment Analysis of Cyberbullying Detection on Social Networks using the Sentistrenght Method

Abstract

In today's swiftly changing digital realm, social media has emerged as a pervasive means of communication, yet it has also fostered the rise of cyberbullying, especially among young demographics. This research strives to develop an application that assesses public sentiment on Instagram regarding cyberbullying instances, categorizing sentiments as positive, negative, or neutral. Drawing data from Instagram accounts such as kumparandotcom, merdekadotcom, and okezonedotcom, the approach combines lexicon-based text labeling and sentiment analysis employing Sentistrength. Findings demonstrate the method's effectiveness, achieving accuracy, precision, and recall rates exceeding 85% while offering precise visualization of predictions. This study contributes to combatting cyberbullying, aiming to improve victims' mental well-being by providing clearer insights into social sentiment. The dataset comprises 4500 comments collected through web crawling, categorized into positive (735 entries), negative (2478 entries), and neutral (1288 entries) sentiments. The evaluation highlights the commendable performance of Sentistrength, achieving the highest accuracy at 93.85%.