EMOJI-AWARE AND CONTEXTUAL SPAM FILTERING ON SOCIAL MEDIA USING ADVANCED ENSEMBLE MACHINE LEARNING

Abstract

The proliferation of spam comments on social media poses a growing challenge to user engagement, content authenticity, and platform trustworthiness. Traditional spam detection models primarily rely on textual features, often overlooking the expressive and contextual elements embedded in social media interactions. This study proposes an emoji-aware and context-driven spam filtering framework that integrates emotional cues from emojis and contextual information from post-comment pairs to enhance spam detection capabilities.
By analyzing not just the comment text, but also the semantics of the corresponding post and the emojis used, the model captures nuanced patterns commonly exploited by spammers. To effectively model these diverse features, we utilize advanced ensemble machine learning techniques, including Random Forest, Gradient Boosting, and a hybrid Voting Classifier, which collectively improve accuracy and reduce false positives.
Experimental results on curated social media datasets demonstrate the superiority of the proposed method over conventional approaches, showing significant gains in precision, recall, and F1-score. The findings underscore the value of incorporating non-textual and contextual indicators in spam detection and pave the way for more intelligent, adaptive, and user-aware content moderation systems.