An ensemble machine learning approach through effective feature extraction to classify fake news

Abstract There are numerous channels available such as social media, blogs, websites etc., through which people can easily access the news. It is due to the availability of these platforms that the dissemination of the fake news has become easier. Anyone using these platforms can create and share the fake news content based on personal or professional motives. To address the issue of detecting fake news, numerous studies based on supervised and unsupervised learning methods have been proposed. However, all those studies do suffer from a certain limitation of poor accuracy. The reason for poor accuracy can be attributed due to several reasons such as poor selection of features, inefficient tuning of parameters, imbalanced datasets etc. In this article, we have proposed an ensemble classification model for detection of the fake news that has achieved a better accuracy compared to the state-of-the-art. The proposed model extracts important features from the fake news datasets and the extracted features are then classified using the ensemble model comprising of three popular machine learning models namely, Decision Tree, Random Forest and Extra Tree Classifier. We achieved a training accuracy of 99.8% and testing accuracy of 44.15% on ISOT dataset and training and testing accuracy of 100% on Liar dataset.