Sentiment Analysis on Twitter Using Machine Learning Approach

Abstract

On social media Twitter, the handling of COVID-19 in Indonesia has been a highly discussed issue. The handling carried out by the Indonesian government raises pros and cons among the public. Public opinion contained on Twitter can be used as a source of decision support in making appropriate policies in evaluating government performance. In dealing with this phenomenon, the sentiment analysis method can be used to analyze public opinion on Twitter. The purpose of this research is to understand public opinion about COVID-19 in Indonesia. Tweet data regarding COVID-19 in Indonesia is processed to classify sentiment into positive and negative. Preprocessing is done to eliminate duplicate and irrelevant data. In addition, text feature extraction is compared to perform sentiment analysis on the new data. The machine learning algorithm is tested using several methods and is subjected to 10-fold validation. The best results were obtained in the second scenario with the best classifier being the Support Vector Machine, which obtained an accuracy of 84.03%.