Laporkan Masalah

A Hybrid Approach for Clickbait Detection

SHAQINA YASMIN, Afiahayati, S.Kom., M.Cs, Ph.D; Yunita Sari, S.Kom., M.Sc., Ph.D

2023 | Skripsi | S1 ILMU KOMPUTER

Media berita online menggunakan berbagai macam strategi yang menarik perhatian untuk menghasilkan pendapatan. Salah satu strategi yang digunakan adalah clickbait. Clickbait merupakan tajuk menarik yang dengan sengaja dibuat untuk menarik perhatian pembaca. Sering kali tajuk tersebut menggunakan bahasa yang menarik dan berlebihan. Namun, konten dari tajuk clickbait umumnya tidak sesuai apa yang dituliskan dalam tajuk tersebut. Keberadaan clickbait mengganggu bagi para pembaca. Terdapat banyak cara untuk mendeteksi clickbait, salah satunya menggunakan traditional machine learning. Metode traditional machine learning sering digunakan untuk mendeteksi keberadaan clickbait. Namun, traditional machine learning bergantung erat kepada rekayasa fitur. Riset ini menggunakan metode yang menggabungkan antara traditional machine learning dengan deep learning untuk menangani berita clickbait. Terdapat delapan model yang dibandingkan satu sama lain, mulai dari logistic regression, SVM, CNN1D, CNN2D, CNN1D dengan logistic regression, CNN1D dengan SVM, CNN2D dengan logistic regression dan CNN2D dengan SVM. Menggabungkan Convolutional Neural Network (CNN) dengan model traditional machine learning menunjukan bahwa metode gabungan atau metode hibrida dapat mengungguli model yang menggunakan model traditional machine learning dan deep learning tunggal untuk berita berbahasa Indonesia dari 12 media berita online. Kombinasi CNN dengan Support Vector Machine (SVM) meraih performa tertinggi di bidang akurasi sebesar 91,2%.

To generate revenue, online news media employ a variety of attention getting strategies. Clickbait is one of the techniques used. It is an enticing headline whose purpose is to attract readers’ clicks. The headlines of clickbait articles frequently use catchy or exaggerated language. Clickbait articles mostly contain poor content. Due to the poor quality of the content, the promises made in the headlines cannot be met. The presence of clickbait is disruptive to readers. A variety of approaches, such as traditional machine learning, have been used to overcome clickbait issues. Traditional machine learning methods, however, rely significantly on feature engineering. This research presents a hybrid technique that combines traditional machine learning with deep learning to overcome this issue. There are eight models compared against each other, from logistic regression, SVM, CNN1D, CNN2D, CNN1D with logistic regression, CNN1D with SVM, CNN2D with logistic regression, CNN2D with SVM. Combining Convolutional Neural Network (CNN) with traditional machine learning models such as logistic regression or Support Vector Machine (SVM) demonstrates that the hybrid approach outperformed the stand-alone traditional machine learning and deep learning approaches in the Indonesian news headlines from twelve online news sources. The CNN and SVM combination achieves the highest performance in terms of accuracy of 91.2% compared to other models

Kata Kunci : clickbait detection, machine learning, deep learning, hybrid approach, convolutional neural network, support vector machine, logistic regression

  1. S1-2023-429297-abstract.pdf  
  2. S1-2023-429297-bibliography.pdf  
  3. S1-2023-429297-tableofcontent.pdf  
  4. S1-2023-429297-title.pdf