Sentiment Analysis of COVID-19 Vaccine Tweets in Indonesia Using Recurrent Neural Network (RNN) Approach

The issue in social media Indonesia about the COVID-19 vaccine is unsafe to use caused concerns among the public. This issue has become a topic of discussion and contention. This work aims to conduct sentiment analysis on the COVID-19 vaccine issues in Indonesia using Recurrent Neural Networks (RNN) variant and traditional machine learning. Data collection is obtained using Tweet's data crawling from the Twitter APL Traditional Machine Learning Methods are used to make a comparison. Compared to other algorithms, the Support Vector Machine has the greatest accuracy, precision, and recall, with an RMSE value of 0.117. We tested types of Recurrent Neural Network (RNN) algorithms using LSTM approach such as simple RNN, Bidirectional LSTM, and Gated Recurrent Unit (GRU). LSTM, BLSTM, and GRU have the same good performance with 91% Accuracy, 91% Recall, 91% Precision, and 0.085 RMSE. Dataset comparisons are also performed for each method. As a result, when more datasets are utilized, good results are produced. Furthermore, the World cloud Tweet reveals that the word "sehat" frequently appears in positive sentiment, whereas the word "daftar" frequently appears in negative sentiment.