A Social Network Based Approach for Detection of Fake News on Twitter Data Using Machine Learning

Abstract

The availability of social media together with
its enhanced accessibility has led to increasingly fast
spread of deceptive information while creating serious
troubles for society and its citizens. The nature of fake
news within modern digital settings creates doubts about
its effects on public belief and political decisions as well
as democratic functions. Fake news operations already
existed, but technological progress combined with social
media platform growth especially among YouTube and
Facebook and Twitter users has created ideal conditions
for fast spreading misinformation. The urgent need
exists to investigate how false information spreads
through multiple social media platforms because of its
concerning rate of growth. This study utilizes a social
networking detection method that depends on network
properties through the Communities through Directed
Affiliation (CoDA) algorithm.
Different experiments
were performed to validate the proposed approach
through evaluations on the FakeNewsNet dataset.
Experimental findings show an Random forest achived
best results with accuracy 0.83, F1-score 0.71,precision
0.78 and recall 0.64 among all the other models
from the proposed detection methods.
Research findings from this study expand the field related to fake news detection through network-based perspectives that
enhance existing methods using content-based and
linguistic analysis approaches.