Political Fake News Detection from Different News Source on Social Media using Machine Learning Techniques

Abstract

People are more dependable on online news systems
than ever in this modern time and day. The more people depend
on online news, magazines, and journals, the more likely it will
have more significant consequences of fake news or rumors. In
the era of social networking, it has become a significant problem
that negatively influences society. The fact is that the internet has
become more accessible than ever, and its uses have increased
exponentially. From 2005 to 2020, overall web users have
increased from 1.1 billion to 3.96 billion [16]. As most
individuals' primary sources are microblogging networks, fake
news spreads faster than ever. Thus it has become very
complicated to detect fake news over the internet. For that
purpose, we have used four traditional machine learning (ML)
algorithms and long short-term memory (LSTM) methods. The
four traditional methods are as follows logistic regression (LR),
decision tree (DT) classification, k-nearest neighbors (KNN)
classification, and naive bayes (NB) classification. To conduct this
experiment, we first implemented four traditional machine
learning methods. Then we trained our dataset with LSTM and
Bi-LSTM (bidirectional long-short term memory) to get the bestoptimized result. This paper experimented with four traditional
methods and two deep learning models to find the best models for
detecting fake news. In our research, we can see that, from four
traditional methods, logistic regression performs best and
generate 96% accuracy, and the Bi-LSTM model can generate
99% accuracy, which outbreaks all previous scores.