HYBRID ENSEMBLE MODEL FOR SOCIAL MEDIA SPAM DETECTION INCORPORATING EMOJI SEMANTICS AND CONTEXTUAL POST ANALYSIS

Abstract

The rapid expansion of social media platforms has made them a prime target for spam activities, including promotional content, phishing attempts, and misleading information. Traditional spam detection methods primarily rely on textual features, often overlooking the rich contextual and emotional cues embedded in posts and comments. This paper introduces a Hybrid Ensemble Model designed to enhance spam detection by integrating emoji semantics and contextual post–comment analysis into the learning process. The proposed model begins with comprehensive data preprocessing that captures both textual and non-textual elements such as emoji usage, comment relevance, and interaction patterns between users. Emojis are treated as sentimentbearing tokens and encoded through semantic embeddings to capture emotional intent, while contextual post–comment relationships are modeled to understand how spam messages interact within conversation threads. Multiple base classifiers, including Support Vector Machines, Random Forests, and Gradient Boosting models, are combined using an ensemble learning strategy to maximize classification accuracy and reduce false detections. Experimental results on real-world social media datasets show that the hybrid ensemble framework significantly outperforms traditional single-model and text-only approaches. The inclusion of emoji semantics enhances the model’s ability to interpret user tone and intent, while contextual post analysis provides deeper insight into message relevance and authenticity. Together, these improvements lead to a more resilient, adaptive, and accurate spam detection system capable of handling the dynamic nature of social media communication. This research establishes a foundation for developing intelligent, context-aware content moderation tools that can effectively manage spam and maintain the integrity of online interactions across diverse social platforms.