ClickBait News Classification for Indonesian Headlines Using the Mamba Selective State Space Model

ClickBait headlines remain a persistent problem in Indonesian digital journalism, where sensational or intentionally misleading titles conceal essential information to inflate page traffic, steadily eroding public trust in online media and weakening readers’ information literacy. This study introduces a real-time, web-based ClickBait detection system that adopts the Mamba architecture, a Selective State Space Model, as a single end-to-end method for judging the consistency between an Indonesian news headline and its body. In contrast to conventional Transformer encoders, whose self-attention grows quadratically with sequence length, Mamba carries out sequential modelling in linear time, and its selective scanning mechanism adaptively keeps semantically relevant tokens while discarding uninformative ones, allowing long-range dependencies between title and content to be captured efficiently. The text is tokenized with the pretrained IndoBERTweet model, whose vocabulary mirrors the informal Indonesian online ecosystem of slang and abbreviations, and punctuation is intentionally retained as an additional linguistic cue. The model was trained and validated on the labelled CLICK-ID corpus of headline-content pairs and assessed through three-fold cross-validation together with a confusion matrix spanning accuracy, precision, recall, and F1-score. Averaged over the three folds, the Mamba classifier obtained an accuracy of 73.43%, precision of 71.31%, recall of 61.15%, and F1-score of 65.65%, with the best result on the second fold, indicating stable generalization to unseen news data. The system is deployed as an interactive web application that classifies a news item directly from its URL, showing that a single state-space sequence model is a viable and computationally efficient option for Indonesian ClickBait detection.

Keywords: ClickBait Detection, Deep Learning, IndoBERTweet, Mamba, Selective State Space Model

Copyright (c) 2026 Journal of Dinda : Data Science, Information Technology, and Data Analytics

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.