Clickbait in Social Media : Detection and Analysis of the Bait

Taking advantage of visual-centric social media's rising popularity, content creators have started using enticing images to lure users into clicking on bothersome clickbaits, in place of previously used text-based baits. In addition, the development of a single model to detect clickbait on multiple image-centric social media platforms is largely an unexplored problem. Therefore, we introduce a novel model that can detect visual clickbaits on both Instagram and Twitter posts. The proposed model consists of a stacking classifier framework composed of six base models (K-Nearest Neighbors, Support Vector Machine, XGBoost, Naive Bayes, Logistic Regression, and Multilayer Perceptron) and a meta-classifier (Random Forest). The developed classifier achieved an accuracy of 88.5% for Instagram posts and 85% for Twitter posts, which is an improvement over previous separate state-of-the-art models for both platforms. Additionally, the stated classifier does not use meta-features (e.g., the number of likes or followers) for classification, which helps to detect potential clickbaits right away, enhancing its applicability in real-time clickbait detection use cases. Furthermore, based on our analysis, we have drawn essential conclusions about the telling characteristics of clickbaits.