OPTIMIZED FEATURE SELECTION AND MSVM-BASED FRAMEWORK FOR AUTOMATED FAKE NEWS DETECTION

Abstract

A significant issue in the contemporary interconnected world is the fast dissemination of disinformation and fraudulent material across many internet platforms. Disinformation may sway public perception, hinder democratic functions, and instigate pervasive uncertainty or anxiety. This paper presents an enhanced Multi-class Support Vector Machine (MSVM) classification method, using feature-based improvements to detect fraudulent news, therefore tackling the recognized difficulty. The system thoroughly comprehends incoming material by examining news articles for lexical patterns, syntactic structures, semantic representations, and contextual cues. To improve model efficacy and reduce computing complexity, these properties are refined by sophisticated selection methods. The MSVM classifier is taught to recognize not just binary outcomes but also several categories, including satire, partly true, and deceptive information, hence facilitating more nuanced and precise classifications. This hybrid method enhances detection precision, scalability, and flexibility across many data sources and languages. The system offers strong safeguards against the spread of disinformation via the use of machine learning and optimal feature engineering. It may be used in real-time contexts such as social media, factchecking tools, and news monitoring systems. The suggested solution improves intelligent misinformation detection systems and addresses the shortcomings of current models.