Sentiment Analysis of COVID-19 Vaccines from Indonesian Tweets and News Headlines using Various Machine Learning Techniques

Abstract

COVID-19 vaccine is a hot topic in online platforms due to the ongoing pandemic. Most studies on sentiment analysis of COVID-19 vaccines on Indonesian social media posts only used one or two classifiers with few modifications. This research investigated sentiment analysis using seven machine learning techniques on Twitter dataset in which the one with the highest evaluation value will be used to predict on other unlabeled Twitter datasets as well as news headlines dataset. The same classifier is also used to build a visualization dashboard that reflect the result of the sentiments. The result from the sentiment classification is then used to identify the topics, by using word cloud. The experiment revealed that SVM classifier has the highest accuracy and micro average F1-measure, which is 84% and 0.76. This classifier managed to capture similar patterns of sentiments in Twitter and news headlines datasets, which is dominated by neutral sentiment. Some of the topics from each sentiment, managed to reflect the real condition when the datasets were collected.