COVID-19 Vaccination Sentiment Analysis on Twitter Using Random Forest and Information Gain

Abstract

Covid-19 in Indonesia has increased from January 2021 unti February 2021 there were 1,217,468 people who were confirmed positive for the corona virus. As a result the increase in the number, the government has taken preventive measures, one of which is the distribution of vaccines or vaccinating the Indonesian people, which has been started since January 13,2021. The government’s covid-19 vaccination efforts had a broad influence on the community through social media (especially Twitter) which then led to pros and cons. Therefore, sentiment analysis is needed to predict the tendency of public opinion regarding the Covid-19 vaccination policy which is classified into positive opinions, neutral opinions, and negative opinions. Random Forest Classifier has high performance compared to other machine learning methods. But the Random Forest Classifier is weak in the level of accuracy and stability of data, so it requires a selection feature to increase its accuracy by applying Information Gain which can increase accuracy by optimizing data features. Measurement of accuracy and sentiment prediction is measured by confusion matrix and classification report. The results show that the application of Information Gain can improve accuracy with the highest accuracy obtained in experiment 1 of 0.00747, that is 0.94776 from 0.94029 with a precision value of 0.65, recall 0.43 and f1-score 0.47 and have a tendency to have a neutral opinion on public tweets about the Covid-19 vaccination on Twitter