Sentiment Analysis on COVID-19 Tweets: Machine Learning Approach

Analysis of tweets accompanying a catastrophic situation is a crucial chore. Sentiment analysis is the field of study to analyze the varied opinions shared by diverse users on social networking platforms on various social phenomena. In this paper, an analysis of the sentiments on thousands of tweets collected from Kaggle on the ongoing pandemic of COVID-19 is carried out. Data preprocessing technique followed by TF-IDF approach for uni-gram and bi-gram features is extracted. Three different supervised machine learning classifiers such as Bernoulli’s Naïve Bayes (BNB), Gaussian Naïve Bayes (GNB), and Random Forest (RF) models are applied. Experimental results suggest that on both the feature extraction models, i.e., uni-gram and bi-gram feature extraction techniques, RF classifier has performed better than the other two models. With 70%-30% train–test set, RF has achieved an accuracy of 90.06% to classify the tweets into negative, neutral, and positive classes.