Evaluating CNN-LSTM for Indonesian Hoax News Classification: A Comparative Analysis with Existing Models

Abstract

Abstract— The spread of Indonesian-language hoax news through digital platforms continues to increase along with the massive use of social media and online news portals. This condition poses a serious challenge to providing accurate and reliable information to the public, necessitating an effective AI-based system for identifying hoax news. This study aims to implement the Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) algorithm in an Indonesian-language hoax news identification system. The CNN-LSTM method was chosen because it combines the advantages of CNNs in extracting local text features and LSTMs in modeling contextual word-order dependencies. The dataset used consists of a collection of Indonesian-language news articles that have been labeled as hoax or non-hoax, after text preprocessing stages including cleaning, tokenization, stopword removal, and padding. Model performance was evaluated using the Accuracy, Precision, Recall, and F1-Score metrics. The test results show that the CNN-LSTM model can classify hoax and non-hoax news with good accuracy and consistency, demonstrating the effectiveness of the hybrid approach for hoax detection. Based on these results, it can be concluded that the implementation of CNN-LSTM has the potential to be a reliable solution to support the Indonesian-language hoax news identification system.

Keywords: hoax news, CNN-LSTM, text classification, deep learning, Bahasa Indonesia