Implementation of CNN Algorithm for Indonesian Hoax News Detection on Online News Portals

The Spread of hoax news in the Industrial Revolution 4.0 era has occurred in the world’s society, including Indonesia. Therefore, an effective method is needed to detect it. The purpose of this research is to apply deep learning with the Convolutional Neural Network (CNN) algorithm in detecting text-based hoax news in Indonesian. The dataset is taken from Kaggle, which has been scraped from CNN Indonesia, Tempo, and Turnbackhoax, which will be labeled as valid and hoax. The implementation of the dataset goes through several processes that include input dataset, data pre-processing using pre-trained embedding GloVe, data processing, model evaluation, also model deployment into the simple web. Data is divided into 80% training data and 20% test data for CNN model development. The results show that the CNN model can achieve high accuracy in detecting hoaxes with training accuracy values reaching 99.65% and validation accuracy reaching 99.88% with a loss of 0.0477 and 0.0435, which means that the model is effective in classifying text-based hoax news to the maximum. The model is evaluated using a confusion matrix, precision, recall, and heatmap as a visualization of results. For further research, it is recommended to increase additional variations for training data so the model can understand patterns well.