Topic Modeling on Covid-19 Vaccination in Indonesia Using LDA Model

ABSTRACT

According to data from covid19.go.id, there is a lot of hoax news about Covid-19 vaccinations spread across various social media in Indonesia. Meanwhile, the ability to monitor and track misinformation and trends regarding Covid-19 and its spread is an important part of the response process by the media and government to dealing with fake news about Covid-19. Twitter is a social media that is actively used in spreading issues. Twitter users in Indonesia reached 18.45 million users as of January 2022. To find out useful information on Twitter social media comments regarding the Covid-19 Vaccination, a method, namely Topic Modeling, can be used. This study aims to obtain the distribution of the Covid-19 Vaccination topic on Twitter Data in Indonesia to assist the government in knowing the trend of topics related to Covid-19 vaccination and the trend of changing the topic. The dataset on Twitter used is 10,140 pieces about Covid-19 vaccinations in Indonesia in the period August 2021 to April 2022. Based on the Latent Dirichlet Association (LDA), the 5 most popular topics were obtained for each model, and spread in the fields of health, religion, society. Based on the alpha and beta hyperparameter tuning, it was found that the topic with K=5, alpha=asymmetric, and beta=0.61 was a relevant LDA topic model for the research dataset because good in topic diversity and have coherence value=0.590. © 2022 IEEE.