This projects built 5 models by training & testing the most efficient Feature extraction techniques (Word2Vc, One-Hot Encoding & TF-IDF) and Deep Learning models (LSTM, Bi-LSTM, CNN, DNN) for various combinations of train-test data split ratios (80:20, 70:30, 64:33, 30:70, 20:80). Out aim is to provide 360 analysis for fake news detection for researchers to understand the trends for each model and then select the model that offers the best accuracy on their specified dataset.
For this reason, we selected all our specified techniques based on the findings for recently published paper in Emerald Publishing Limited, under the title "A systematic survey on Deep Learning and Machine Learning approaches of fake news detection in the pre- and post-COVID-19 pandemic". This paper reviews the existing fake news detection technologies by exploring various machine learning and deep learning techniques and trends pre- and post- pandemic.
You can access the full paper here: https://lnkd.in/d4UyS48f
Dataset Avalaible at : https://www.kaggle.com/datasets/mrisdal/fake-news