Truth in the Age of Misinformation: A Dual-Approach Model for Improved Fake News Detection

The inspiration that drives this study is the increase in the prevalence of misinformation, and the need to solve the big challenge of detecting fake news. With the possibility of uncontrolled misinformation disastrously impacting our society, the objective of our study is to make contributions towards the body of knowledge which aims to improve the efficiency and accuracy of current methodologies to detect false information. Although there has been an improvement in the fake news detection systems, their adaptation to real-time applications remains sub-optimal due to the subtle and complex processing involved with handling misinformation. The purpose of this study is to bridge this gap and come up with a better detection model using the Recurrent neural network (RNN) models like Long Short-Term Memory (LSTM) and Bi-directional LSTM (BiLSTM) that will enhance the performance of the existing models, especially in terms of detection accuracy, as the existing systems are weak in the “disagree” category and mostly misclassify it. We use a mix of machine learning algorithms and natural language processing techniques. Integrating LSTM with BiLSTM will analyze textual content at a granular level to find patterns representing fake news. The preliminary results have indicated that our model is performing significantly better than existing ones, with an overall improvement of up to 86.25% in accuracy [15]. However, in the “disagree” category, there are a few challenges faced by the system, reflecting the scope for many improvements in the future. Our research has made a valuable contribution to the knowledge and tools in the area of digital information credibility by providing a strong model that improves detection rates of more subtle cues of misinformation thereby, combating the spread of fake news.