An Ensemble Technique to Detect Fabricated News Article Using Machine Learning and Natural Language Processing Techniques

Abstract

Fake news or fabricated news, refers to false information published under the guise of being authentic news, often to influence political views. Fabricated news articles are a threat to people’s trust in the government and in effect, one of the biggest threats that modern-day democracies are facing. As the menace of fake news is growing with each passing day, so is the research community getting more actively involved in curbing this issue. This paper reviews the current progress of the advancements done to solve the issue. The paper also presents various ensemble techniques to perform the binary classification of news articles. Additionally, the paper focuses on sources of articles to widen misclassification tolerance and make more accurate predictions. Evaluation metrics such as accuracy score, precision, recall, f-1 score have been used to make comparison among various models. The best performing model has been an ensemble of Decision Tree, Logistic Regression, Bagging Classifier used with a hard-voting ensemble technique, which gives the accuracy of over 88%.