Classification of fake news using machine learning and deep learning

Abstract

The rapid spread of fake news through digital channels is a major problem. In this study, after processing the texts with natural language processing techniques, machine learning methods and deep learning methods, the style-based detection of fake news was investigated with text analysis. After the necessary text processing on the open-source dataset ISOT, different models were built using word representations (TF-IDF, word2Vec) and different machine learning (K nearest neighbor Naïve Bayes, logistic regression) and deep learning Long Short-Term Memory (LSTM) methods. Acc, P, R and F were used to evaluate the performance of these models. On the fake news dataset, the LSTM model performed best with 99.2% Acc. Improving state-of-the-art methods on word representations and classification steps, including preprocessing in text classification processes, and making them usable in a practical environment can significantly reduce the amount of fake news.