A Machine Learning Approach for Sentiment Analysis to Nurture Mental Health Amidst COVID-19

During the pandemic, when fresh news content is generated every minute about the widespread of the virus, many conversations revolve around the spread and cure of the contagion. At the hands of a commoner who posts news about COVID-19 on social media, the news may manifest itself to accommodate the said person's fear or negative propaganda which can potentially trigger a mass panic outbreak or can disrupt the mental health of a reader. This research discusses the application of Machine Learning in Sentiment Analysis to classify Tweets about Coronavirus as fear sentiment or panic sentiment. It proposes the idea of a web-based application that caters to filter out the fear-inducing sentiment from a user's daily Twitter feed, thus giving the user accurate and well-spirited information. Textual analysis is performed along with necessary textual data visualization. A substantial accuracy of 91% is achieved in the classification of brief Tweets using the Naïve Bayes method. An accuracy of 74% is achieved using the Logistic Regression classification method for brief tweets. This depicts the advancements in the field of sentimental analysis and sheds light on how it can be employed amidst a challenging situation like the pandemic to preserve mental health.