Analysis of Changes in Sentiment towards COVID-19 Vaccination in Indonesia Using the Convolutional Neural Network

Vaccination against COVID-19 is one of the efforts to prevent transmission of COVID-19 in accordance with recommendations from the World Health Organization (WHO) and supported by the government through Presidential Decree 99 of 2020. Responding to the vaccination program, the public gave various opinions on Twitter. During the vaccination process, there are many changes in public opinions that can be analyzed over time. In this research, a time series analysis of sentiment changes was conducted using the Convolutional Neural Network (CNN) algorithm to find the best model. The data obtained will be weighted using the FastText method. Changes in sentiment will be divided into three phases, each being tested nine times with three variations of epochs, namely 10, 50, and 100. Each epoch has three variations of split data performed with percentage ratios of 80:20, 70:30, and 60:40. The study results show that the model can achieve the highest accuracy of up to 87.7%. Sentiment results in the first time series had positive sentiments of 1,454, while negative sentiments were 1,541, with an analysis confidence level of 83.1%. Sentiment results in the second time series had positive sentiments of 1,379, while negative sentiments were 1,265, with an analysis confidence level of 87.7%. Furthermore, the sentiment results in the third time series had 1059 positive sentiments, while 872 negative sentiments with an analysis confidence level of 85.43%. From the three-time series negative sentiment towards the COVID-19 vaccination policy is decreasing over time.