A Multi-classifier Framework for Detecting Sp...

A Multi-classifier Framework for Detecting Spam and Fake Spam Messages in Twitter

Author Name : Shaik Amer, Syed Abrar Ali, Mohammed Muqtadir Qureshi, Imreena Ali

DOI: https://doi.org/10.56025/IJARESM.2023.1201243340

ABSTRACT Social media is integral for social gatherings, entertainment, communication, and knowledge sharing. Twitter, a prominent platform, connects millions for information sharing and has become crucial for promoting products and influencing decisions. However, the prevalence of spammers, who send duplicate messages for advertisements, phishing, and scams, poses a significant challenge. This paper introduces a novel spam detection mechanism to identify suspicious users on Twitter. The system utilizes a semi-supervised approach at the tweet level, transitioning to a supervised level for learning input tweets to detect spammers. It also identifies spammer types and removes duplicate tweets. We applied multiple classifier algorithms including Random Forest, KNearest Neighbors (KNN), Naive Bayes, Decision Tree, Stacking Classifier, and Voting Classifier. The ensemble method, combining predictions from individual models, yielded robust and accurate results. Notably, the Voting Classifier achieved 100% accuracy. Our findings suggest that ensemble techniques, particularly the Voting Classifier, significantly enhance spam detection performance, providing a reliable solution for maintaining the integrity of Twitter communications.