An Ensemble of Machine Learning Algorithms for Detecting Fake News

The collection of false updates propagated through various media in disguise of genuine information is considered fake news. Society might be negatively impacted by the unmonitored propagation of these updates. A recent area of interest for machine learning and natural language processing academics is the identification of fake news. The main objective of this research is to develop systems to determine the news items that are fake in the large collection of news updates everyone receives in their day-to-day life. There have been a good number of research works in the area, but most of them are done using classical machine learning algorithms. This work aims to solve this problem using an ensemble of three machine learning algorithms: SVM, random forest classifier, and extra tree classifier. An open-source data set was taken to train the ensemble of the classifiers. Before training the model, the data set was passed through a sequence of state-of-the-art data preprocessing steps. The data set was found to be imbalanced and for that, the standard class balancing was done to make sure that the models were not biased to a particular class. The accuracy obtained through the ensemble classifier is 94.6%. A comparison of the result was done with the pre-existing research works in this field and it was found the current work produces better accuracy.