Hybrid CNN-LSTM Model for Fake News Detection

Abstract

In recent years, the way people access information has changed because of the increasingly digital world. Social media has begun to replace traditional news sources such as television and newspapers. Most people reach news about social, economic, and political developments worldwide through social media. Its fast, easy access and cost advantage have made social media widely used among users. In addition to these advantages, social media has become a suitable platform for disseminating fake news. Fake news can have hazardous consequences for individuals, societies, and governments. Therefore, detecting fake news on social media must be necessary. This research created a hybrid CNN-LSTM model for detecting fake news. The CNN component is responsible for analyzing subsequences, which serve as inputs to the LSTM, and extracting relevant features. While the CNN captures critical features from the input data, the LSTM is employed for the classification. The created model was tested with LR, RF, SVM, MLP, and LSTM. The experiments showed that the created model is more successful than the others, with 99.91% accuracy, 99.93% precision, and 99.89% recall. In addition, according to our research, more successful results were obtained in this study than in all studies in the literature using the ISOT dataset.