MACHINE LEARNING APPROACHES FOR FAKE NEWS IDENTIFICATION AND CLASSIFICATION

Abstract

With the rapid growth of social media, nearly 70% of people now rely on these platforms as their primary news source. However, social media has also become a major channel for spreading misinformation and fake news. This paper introduces a semantic-based approach for fake news identification, designed to capture the depth of misinformation and enable dynamic decision-making. The proposed Fake News Identification Ontology (FNIOnt) categorizes news content into fictitious classes and performs semantic analysis on dataset content. The extracted features are evaluated using three machine learning classifiers—Random Forest (RF), Logistic Regression (LR), and Long Short-Term Memory (LSTM). Experimental results demonstrate that the proposed framework outperforms existing methods, achieving a 99% accuracy rate.

Despite these promising results, the adaptability of fake news creators presents an ongoing challenge. Techniques such as half-truths, subtle linguistic manipulation, and evolving writing styles make detection increasingly complex. Therefore, machine learning systems must evolve dynamically, learning from new data and adapting to emerging patterns of deception. Deep learning models—including Recurrent Neural Networks (RNNs), Convolutional Neural Networks (CNNs), and Transformers—further enhance detection by automatically capturing semantic and syntactic relationships in unstructured text. Unlike traditional models that rely on manual feature engineering, these neural approaches enable context- and sentiment-aware classification, strengthening robustness against nuanced manipulations.

This study underscores the potential of combining semantic analysis with advanced machine learning and deep learning methods to build adaptive, accurate, and scalable systems for fake news detection.