Unmasking Deception: A Comprehensive Comparative Analysis of Machine Learning Methods for Detecting Fake News in Social Media

In the current scenario, recognition of fake news on Internet community is challenging, prompting the need for an automated system. This study proposes a Machine Learning (ML) model aimed at distinguishing fake news from authentic sources efficiently, addressing the pressing issue of rapid misinformation dissemination. This study employs the InFake Source Open Type (ISOT) dataset, which contains articles include 23,481 fake news and 21417 correct news sourced from online. In pre-processing, various text transformations including lowercase conversion, removal of brackets, URLs, HTML tags, and non-word characters are incorporated. Further, text vectorization is performed using the TfidtVectorizer representing the data numerically through Term Frequency-Inverse Document Frequency (TF-IDF) weighting. Then, the vectorized text data is used to train ML models such as logistic regression, gradient boosting, decision tree and random forest. Then, the combined predictions from different models are obtained, classifying the news as either real or fake. The experiments demonstrated the suggested technique outperforms current techniques in terms of classification accuracy.