Sentiment Analysis of Covid-19 Handling in Indonesia Based on Lexicon Weighting

Abstract

Covid-19, a contagious disease, has been classified as a global pandemic. Indonesia, as one of the ASEAN countries, has taken various measures to combat the spread of this disease. One of the government's initiatives to tackle the pandemic is the PeduliLindungi application, through which the public provides feedback on government policies. However, analyzing and comprehending public opinions in a non-subjective manner poses a challenge in objectively evaluating government services. This study aims to address this issue by conducting a sentiment analysis of Covid-19 handling in Indonesia, using a lexicon-based weighting system that includes SentiStrengthID and InSet. The decision tree (DT) machine learning algorithm is utilized to evaluate the polarity results provided by the lexicon. The results indicate that the sentiment polarity towards Covid-19 handling in Indonesia is negative based on both SentiStrengthID and InSet weights. Evaluating machine learning performance with the SentiStrengthID lexicon, the DT-entropy and DT-gini models achieved an accuracy of 82% and 83%, respectively. Similarly, evaluating machine learning performance with the InSet lexicon, the DT-entropy and DT-gini models achieved an accuracy of 81% and 82%, respectively.