Profiling Fake News Spreader on Twitter Using Stochastic Gradient Descent

A fake news spreader is a person who deliberately shares misleading information to misguide people. Fake news spreaders can be considered a potential threat to our society. Any misleading information floating in social media can create a serious misconception in society. It can damage the reputation of a person and create cybersecurity threats. To ensure a safe and sound social networking environment for everyone, it is inevitable to detect fake news spreaders. Numerous state-of-the-art approaches have been done to address this issue. They employed several classification algorithms to detect the fake news spreaders. Usually, social media data often lack proper grammar due to noise and informal language. In most of the prior work, the authors did not use the proper text preprocessing techniques which leads to poor performances. To solve this problem, we describe our proposed framework which includes extensive text preprocessing techniques and statistical features along with the stochastic gradient descent (SGD) classifier on top to detect the fake news spreader efficiently. We choose SGD because of its capability to handle large datasets as well as the stochastic phenomenon helps to escape local minima and performs better generalization in complicated optimization scenarios. Experimental results on the benchmark PAN 20 fake news spreader dataset show that our proposed approach outperformed most of the prior works with a 76.50% accuracy thus validating its effectiveness.