A Critical Evaluation of Machine Learning Techniques for Fake News Detection Systems

Abstract

The rapid proliferation of digital media platforms has significantly transformed the way information is produced, disseminated and consumed. While these platforms have enhanced accessibility and communication, they have also facilitated the widespread circulation of fake news, posing serious threats to social harmony, political stability and public trust. In this context, machine learning has emerged as a powerful tool for identifying and mitigating the spread of misleading information. This study critically evaluates various machine learning techniques used in fake news detection systems, focusing on their performance, strengths, limitations and applicability in real-world scenarios. The paper examines traditional algorithms such as Logistic Regression, Support Vector Machines, Decision Trees and Random Forest, as well as advanced approaches including Neural Networks and hybrid models. It further explores feature extraction methods, dataset challenges and evaluation metrics that influence model effectiveness. The findings highlight that while machine learning models can achieve high accuracy, they often struggle with issues such as data imbalance, lack of contextual understanding and vulnerability to adversarial manipulation. The study concludes by emphasizing the need for more robust, interpretable and adaptive models that integrate linguistic, social and contextual features for improved fake news detection.