Perbandingan Performa Ensemble Classifier dan Model Klasifikasi Tunggal dalam Clickbait Detection

Abstract

The rapid advancement of the digital era has led to a significant increase in online news content, accompanied by a growing issue of clickbait—sensational headlines designed to attract readers. This study aims to develop a clickbait detection system to classify Indonesian news headlines as clickbait or non-clickbait. Building upon previous research, this work explores the performance of ensemble classifiers compared to single classifiers such as Multinomial Naïve Bayes, Support Vector Machine (SVM), and Random Forest. Using a dataset of 4,500 headlines with 51.6% clickbait and 48.4% non-clickbait ratio. With an 80-20 train-test split, four machine learning models were applied. The best-performing model was the ensemble classifier using hard voting, achieving an accuracy of 84.22%, precision of 85.81%, recall of 82%, and an F1-score of 83.86%. These results indicate that the ensemble classifier outperforms single classifiers in identifying clickbait in Indonesian news headlines. The findings suggest that ensemble classifiers are a promising approach for improving clickbait detection in digital news media.