Sentiment Analysis on COVID-19 Twitter Data

Abstract

Corona virus first appeared in December 2019 in Wuhan, China which eventuaUy lead to a catastrophic impact allover the world. The entire world had been fighting this pandemic and expressing their feelings, sharing their opinions on various social media platforms. Substantially Twitter had been the better medium to express their opinion and share updates about the situation. This paper analyzes the sentiments of the public based on positive, negative, and neutral tweets. This analysis eventually helps in the prediction of the covid-19 situation in the world. The dataset was collected from Kaggle which was uploaded by Gabriel Prada containing more than 1,70,000 tweets. Based on tbese tweets posted on Twitter a sentiment analysis was performed. Data Collecting, Data cleaning (Removing RL, #, @ and various types of punctuation), Tokenization, Stemming, removing stop- words were performed, and to fmd the polarity, two types of analyzers were used that is TextBlob and Mmn. 1,79,108 tweets were manually analyzed and comparing it with both the analyzers shows Afmn is more accurate. To evaluate the accuracy a few Machine learning Algorithms had been applied (Logistic Regression, aive Bayes, Decision Tree, and Linear Regression) for predicting the sentiment of the tweet.