Machine Learning based Sentiment Analysis of Coronavirus Disease Related Twitter Data

Abstract

Newly discovered coronavirus has resulted in an infectious disease named Coronavirus disease (COVID-19). The outbreak of Coronavirus has caused many nations to exercise lockdowns to curb the spread of this virus. In this situation, there has been an upsurge in the usage of the internet and social media platforms. Therefore, amidst this pandemic, another crisis has also evolved; caused by the spread of incomplete and often inaccurate information which keeps leading to mass fear and anxiety. Hence it is high time to address this informational crisis and find the polarity of people's sentiments towards COVID-19 situation so that authorized bodies can take suitable actions to cope with the situation and also review the prevalent systems. Currently, most of the sentiment analysis focus on texts, therefore we extracted sentiment from real-time tweets as well as from individual input image/video. By making use of PySpark and the libraries within it, we enabled scalable and fault-tolerant stream processing of live data stream. This study uses classification algorithms such as logistic regression, decision tree, random forest, Naive Bayes. After evaluation of all the models, logistic regression model was seen to have the highest values for accuracy and F_score. Hence this model has been used for the final sentiment analysis.