Deciphering COVID-19 Narratives: A Comparative Study of ML Models (RF, MNB, GB, LR, SVM) and DL Models (CNN, Bi-LSTM) for News Article Classification

Abstract

The COVID-19 pandemic has provided an unprecedented amount of information in news outlets, which include scientific, health-related, political, economic, and social narratives. This study compares the effectiveness of machine learning and deep learning algorithms for classifying text data, with a certain emphasis on how well the former handle COVID-19 news narratives. The study dataset contains news articles regarding COVID-19. To achieve the primary purpose of this research is to classify COVID19 related news, we integrate multiple datasets. The analysis reveals machine learning models exhibit superior performance in text data classification. In particular, the Random Forest model reaches a 98% accuracy rate. In contrast, with regards to deep learning models, the Bidirectional Long Short-Term Memory model with FastText integration turns out to be the best option due to its exceptional accuracy. Exploratory data techniques such as topic modeling and word cloud approaches are incorporated to uncover hidden patterns in the data. Pre-trained (e.g., deep learning) and non-pre-trained ML models are implemented highlighting the versatility of ML in text classification tasks. The specific purpose to compare to the deep learning and machine learning algorithm to classification of the new article. Notably, a predictive model employing Bi-LSTM with the FastText pre-trained model achieved an impressive 94% accuracy in classifying COVID-19 news reports.