A Location Independent Machine Learning Approach for Early Fake News Detection

The spread of fake news on the Internet is presenting increasing threats to national security, with the potential to incite public unrest and violence. However, detecting fake news is challenging as they are intentionally written to mislead. Some current methods cannot detect fake news early and require external information like the source to assess articles. To tackle these challenges and improve the generalizability of the models, we adopted a text-based location-independent machine learning approach. It employed two types of machine learning models. The first is the bag-of-words model, made more robust by stacking two levels of models. The second is neural networks that utilize pre-trained GloVe word embeddings, including (a) one-dimensional convolutional neural network (CNN) and (b) bidirectional long short-term memory network (BiLSTM). All models were assessed on various metrics (accuracy, recall, precision and F1), and achieved over 90% on the test set, making this an effective location-independent approach to detect fake news at an early stage without reliance on external information.