Enhanced Fake News Classification Using Feature-Based Optimized MSVM

Abstract

The rapid spread of misinformation and deceptive content across online platforms has become
a pressing concern in today's digitally connected world. Fake news can influence public
opinion, disrupt democratic processes, and cause widespread panic or confusion. In response
to this challenge, this project proposes a robust and intelligent fake news detection system
utilizing a feature-based optimized Multi-class Support Vector Machine (MSVM)
classification approach. The system extracts a rich set of features from news articles,
including lexical patterns, syntactic structures, semantic representations, and contextual cues,
ensuring a comprehensive understanding of the input text. These features are then optimized
using advanced selection techniques to enhance model performance and reduce computational
complexity. The MSVM classifier is trained to identify not only binary outcomes (fake or
real) but also multiple categories such as satire, partially true, and misleading content,
offering more nuanced and accurate classifications. This hybrid approach improves detection
accuracy, scalability, and adaptability across different data sources and languages. By
integrating machine learning with optimized feature engineering, the system presents a
powerful solution for combating the spread of fake news and can be deployed in real-time
environments such as social media platforms, fact-checking tools, and news monitoring
applications. The proposed method addresses limitations of existing models and significantly
contributes to the advancement of intelligent misinformation detection systems