Pursuit for Authentic News using Machine Learning Models

Abstract

Fake news spreads like a wildfire and this is a big issue in this era. The Online world contains lots of fake news from various sources and channels by political parties, influential peoples, and bots. People are not good for easily able to distinguish fake news from real one. This will negatively impact people's lives, society, and the world. To maintain stability from the harm done by fake news, we will tackle the topic of fake news detection. we are investigating the effectiveness of machine learning technologies and also introducing a deep learning model for fake news detection with better accuracy. We are using Natural language processing with the above technologies for our purpose. In investigating the effectiveness of machine learning technologies, we had applied various classification algorithms on our data processed by NLP to obtain their accuracy for the problem. Then using new emerging technologies like deep learning, we are proposing solutions with better accuracy. In the machine learning technologies for fake news detection, we have found that AdaBoost classifier, Gradient boosting classifier, and Logistic regression are better in terms of accuracy than other classifiers like decision tree, KNeighbors classifier, Random Forest classifier, and MultinomialNB. But These technologies are more prone to error when a different category of data comes. Deep learning technology used here is the Long short-term memory deep learning model which gave us an accuracy of more than 0.993 and the Bi-directional LSTM model with accuracy near 0.99 taking more time in training than the LSTM model. Through this research, we conclude that machine learning technologies perform worse than deep learning technologies. And proposed LSTM model is better than the Bi-directional LSTM model for fake news detection.