Implementation Of Deep Learning For Fake News Classification In Bahasa Indonesia

Abstract

Fake news has become a serious threat in the digital information era. This research aims to develop a model for detecting fake news in Bahasa Indonesia using a deep learning approach, combining the Long Short-Term Memory (LSTM) method with word representations from Word2vec Continuous Bag of Words (CBOW) to achieve optimal results. Our main model is LSTM, optimized through hyperparameter tuning. This model can process information sequentially from both directions, allowing for a better understanding of the news context. The integration of Word2vec CBOW enriches the model's understanding of word relationships in news text, enabling the identification of important patterns for news classification. The evaluation results show that our model performs very well in detecting fake news. After the tuning process, we achieved an F1-Score of 97.30% and an Accuracy of 98.38%. 10-fold cross-validation yielded even better results, with an F1-Score and Accuracy reaching 99%.