Exploring Fake Pandemic News Detection Using Artificial Intelligence Techniques

According to research, fake news spreads faster than real and trustworthy ones. Since the start of COVID-19 pandemic in 2019, a plethora of rumors and false news circulated that caused significant problems in identifying the validity of facts. Disinformation harms every aspect of society in ways that can prove to be awful. Only when there is a reliable system for evaluating news is necessary for individuals to distinguish between true and false information. In times of uncertainty like the COVID-19 epidemic, invalid and false news manipulated the psychological health of individuals by instilling fears inside their minds. People had scarce knowledge about the novel infectious disease that had just been disseminated and so they trusted every piece of information without questioning its validity. Organizations, businesses, and governments suffered as one single fake news caused turmoil and disruption of peace at economic and financial levels single source of truth was needed in this case it was the development of a detector that aided in the verification of the news circulating. A fake pandemic news detector is developed using the deep learning model Long Short-Term Memory (LSTM). Moreover, a dataset is compiled by scraping data from different online sources such as social media, news, and websites. This scraped data will be used as a training dataset for the detector. Data preprocessing is carried out to remove noisiness and ambiguities in the data. The creation of a fake pandemic news detector will be a great development and would enlighten individuals to the extent of the veracity of information. A detector is developed for the verification of the validity of COVID-19 news and then its deployment is done as a mobile appointment for every commoner. This study aims to enhance fake news detection accuracy, providing a robust AI- based solution to combat misinformation during pandemics. The dataset used in this study comprise of 18285 samples which has given an accuracy of 91% using the LSTM model and 90% using logistic regression. We have used the logistics Regression model in training the dataset, this model is suitable as the system will be producing binary outputs i.e. either fake or real.