TF-IDF, chi-square, and ANN machine learning techniques for fake news anomaly detection in online social networks

Abstract

Platforms such as Facebook, WhatsApp, Twitter, and Telegram have a substantial influence on the dissemination of information in modern society. Numerous individuals rely on them without verifying the veracity of their information or the sources of their knowledge. False information is referred to as "Fake News" and is disseminated through both traditional and non-traditional media channels including social media. Contemporary media consumers uncritically embrace the news they read online, rather than taking the time to critically assess it. This method has led to a heightened trust in internet content, which has facilitated the spread of misinformation. This article describes a machine learning (ML)-based approach to classifying and detecting fake news anomalies in online social media data. This framework accepts social media data as input. Features are chosen using the term frequency identification term frequency–inverse document frequency (TF-IDF) and chi-square tests. The categorization model is then built utilizing approaches such as artificial neural network (ANN), support vector machine (SVM), and random forest (RF). The classification model is trained and tested on a preprocessed data set. Finally, fake news is discovered, and the suggested framework is evaluated using measures such as accuracy, precision, recall, and F1-score. ANN is performing better for fake news classification and detection.