A Study Towards Bangla Fake News Detection Using Machine Learning and Deep Learning

Verifying Bangla Fake news is challenging, especially if there are many updates from various sources such as social media or online news portals. This study aims to identify the Bangla fake news article; therefore, our Corpus is trained with 57,000 Bangla news items related to trustworthiness and counterfeit. In this study, 95% and 94% accuracy were found by applying K-fold cross validation on top of Bi-LSTM with Glove and FastText model. At the time, the research is also experimented with the state-of-the-art technique like Gated Recurrent Unit (GRU) and found the accuracy 77%. In sharp contrast, we tracked out the accuracy of 96% utilizing the Bi-LSTM which identically indicates our proposed model. A comparative study on existing work has been utilized in this research. Again, some experimental analysis based on the FEM is shown elaborately in this research. However, the proposed system can be adjustable in real-time news classification of Bangla Fake news.