Machine Learning Based Anxiety Prediction of General Public from Tweets During COVID-19

ABSTRACT

In this crisis of COVID19, everyone is staying in touch with the world through social media. This has led to social media becoming a significant source of new information for many people and unfortunately this phenomenon has given birth to a lot of misinformation, chaos and fear in people’s minds. This fear is often due to the inadequate and wrong information. Therefore, there is a important need to understand this crisis. Patterns need to be established between popular tweets and its effect on the public’s sentiments, especially their fear. So, tweets of three different countries namely United States of America, Federative Republic of Brazil and Republic of India. Sentiment analysis reveals that fear of this unknown and mysterious nature of the coronavirus is dominant among the public. Predominant analysis of tweets within past two months will be done and then a model will be built to predict future reaction of the general public based on the crisis level in the country. Machine Learning algorithms such as ‘Logistic Regression (LR)’, ‘Multinomial Naïve Bayes’ and ‘Support Vector Machine (SVM)’are used for classification purpose preceded by the pre-processing steps of raw data from each country. 90% of accuracy has been achieved from sentiment classification result. Insights to the fear, sentiments have also been provided. Tweets with negative sentiment and emotion indicates the cases for the pandemic outbreak. © 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.