A Comparative Study of Machine Learning Algorithm for Fake News Detection

Abstract: In today’s digital era, the widespread use of social media and online platforms has enabled rapid dissemination of information—but also facilitated the spread of misinformation, commonly known as fake news. Such false information can distort public opinion, disrupt political processes, and cause widespread confusion. This research addresses the growing challenge of fake news detection by developing a binary classification system using machine learning techniques. The study compares the performance of two supervised algorithms—Naïve Bayes and Logistic Regression—applied to the Constraint@AAAI 2021 shared task dataset on COVID-19 fake news. The dataset underwent rigorous preprocessing, including text normalization, noise removal, stopword elimination, and TF-IDF feature extraction. Experimental results demonstrate that both models perform effectively in classifying real and fake news, with Naïve Bayes achieving an accuracy of 92.37% and Logistic Regression slightly outperforming it with 93.85%. These findings highlight the potential of lightweight machine learning models for reliable and efficient fake news detection, contributing to the fight against online misinformation and promoting trustworthy digital communication.

Keywords: Fake news, Machine Learning, Naïve bayes, Logistic Regression.