Detecting Fake News with Machine Learning

Published in International Conference on Deep Learning, Artificial Intelligence and Robotics, 2019

Recommended citation: Nagender Aneja, Sandhya Aneja "Detecting Fake News with Machine Learning." International Conference on Deep Learning, Artificial Intelligence and Robotics, 2019. pp. 53--64 doi: 10.1007/978-3-030-67187-7_7 https://link.springer.com/chapter/10.1007/978-3-030-67187-7_7

(Conference Publication)

Access paper here

Abstract: Fake news is intentionally written to influence individuals and their belief systems. Fake news detection has become extremely important since it negatively impacts society and politics. Most existing works have used supervised learning but have given importance to the words used in the dataset. The approach may work well when the huge dataset covers a wide domain. However, getting the labeled dataset of fake news is a challenging problem. Additionally, the algorithms are trained after the news has already been disseminated. In contrast, this research emphasizes content-based prediction based on language statistical features. Our assumption of using language statistical features is relevant since the fake news is written to impact human psychology. A pattern in the language features can predict whether the news is fake. We extracted 43 features that include Parts of Speech and Sentiment Analysis and showed that AdaBoost gave accuracy and F-score close to 1 when using 43 features. Results also show that instead of all 43, the top ten features give an accuracy of 0.85 and an F-Score of 0.87.