Sentiment Analysis and Topic Modeling of Citizen Satisfaction with the Indonesian Government in Handling a Pandemic

Abstract

When the COVID hit the world, many countries issued policies to stop the spread of the virus. In Indonesia, various opinions and surveys have emerged regarding citizen satisfaction with the government's performance in handling the pandemic. But the survey method still has many weaknesses, for example: bias from the researcher, lack of confidentiality, hello effect, and so on. The purpose of this study is to classify this phenomenon into positive and negative sentiments taken from Twitter data. Every record is pre-processed to clean the data. Data labelling using a lexicon consisting of positive or negative polarities. Sentiment classification using Support Vector Machine (SVM). Each positive and negative sentiment will be processed using Latent Dirichlet Allocation (LDA) method to find out the interpretation of the main topics that are often discussed, then made into a visualization using a word cloud. The best model obtained was the model with TF-IDF feature extraction with a precision value of 0.87, a recall of 0.95, an accuracy of 0.89, and an F1-measure of 0.93. Our findings indicate that people are more likely to be satisfied with the performance of the government's fight against COVID than with the policies they introduce. People are also satisfied because they can feel mudik (go back to hometown) again after two years of the pandemic. Dissatisfaction comes from people who think that there is a business game in vaccine policy as well as the government's lack of transparency regarding the number of COVID cases.