Sentiment Analysis of Tweets During the COVID-19 Pandemic Using Multinomial Logistic Regression

Abstract

Recently, the research on sentimental analysis has been growing rapidly. The tweets of social media are extracted to analyze the user sentiments. Many of the studies prefer to apply machine learning algorithms for performing sentiment analysis. In the current pandemic, there is an utmost importance to analyze the sentiments or behavior of a person to make the decisions as the whole world is facing lockdowns in multiple phases. The lockdown is psychologically affecting the human behavior. This study performs a sentimental analysis of Twitter tweets during lockdown using multinomial logistic regression algorithm. The proposed system framework follows the pre-processing, polarity and scoring, and feature extracting before applying the machine learning model. For validating the performance of proposed framework, other three majorly used machine learning based models-- namely decision tree, naïve Bayes, and K-nearest neighbors-- are implemented. Experimental results prove that the proposed framework provides improved accuracy over other models.