Enhancing Fake News Detection through Clustering with Convolutional Neural Networks

Abstract

In this study, a deep learning-based clustering algorithm was applied for the detection of fake news. In the initial stage, true and fake news articles were collected from three different news websites using web scraping method. Subsequently, these collected news data were merged with two publicly available datasets. This resulted in the creation of a dataset encompassing approximately 130206 text samples of true and fake news, providing a wide range of topics. A Convolutional Neural Network (CNN) architecture based on Bidirectional Encoder Representations from Transformers (BERT) was developed to obtain the deep representations of the dataset. Then, clustering using the K-Means algorithm was performed on the deep features of the data to detect fake news. This study presented a higher accuracy rate while reducing the cost and computational process by utilizing fewer features. Additionally, this work contributes to the field as it introduces a new dataset and addresses the scarcity of studies on fake news detection using the deep learning-based clustering approach, thereby serving as a valuable resource for future research in this area.