Enhancing Feature Selection for Fake News Detection Using A Hybrid Genetic Algorithm and Particle Swarm Optimization Approach

Abstract

Fake news has emerged as a major concern in today’s digital era, spreading rapidly and influencing various aspects of society, including mental well-being and public opinion. This study focuses on addressing this issue by proposing an effective framework for detecting fake news. The research utilizes the FakeNewsNet dataset, beginning with thorough data pre-processing to ensure quality and consistency. Features were then extracted using the TF-IDF (Term Frequency-Inverse Document Frequency) technique, which is widely used for textual data analysis. To further enhance the dataset, a hybrid approach combining Genetic Algorithm and Particle Swarm Optimization was employed for feature selection. This hybrid method ensures that only the most relevant and impactful features are retained, optimizing the overall performance of the detection system. The selected features were used to train different machine-learning models. Among these, Logistic Regression delivered 99% accuracy, a significant result. However, three other machine learning algorithms outperformed it, achieving even higher accuracy, highlighting their superior ability to classify fake news accurately. This research aims to contribute to the ongoing fight against fake news by providing a reliable and efficient detection system that can help mitigate its negative effects on society.