Sentiment Analysis of Social Media Twitter with Case of Large Scale Social Restriction in Jakarta using Support Vector Machine Algorithm

Abstract

When the Large-Scale Social Restrictions (LSSR or PSBB in Indonesian) policy was implemented it the policy was not entirely obeyed by the community which then reaped various opinions and responses on various social media, especially on Twitter. This study aims to conduct a sentiment analysis to find out the cause or phenomena that occur based on the opinions or views of Twitter. The Tweet data about the implementation of LSSR both part 1 and part 2 in Jakarta were obtained as many as 1080 opinions using the crawling method then the data is manually labelled with two labels, which are positive and negative after labelled the data is cleaned after and the data is processed by being weighted using the Bag of Words and TF-IDF extraction feature. The classification process is carried out with four splitting data scenarios, with 60:40, 70:30, 80:20, 90:10 then classified using the Support Vector Machines algorithm. The final result of this study shows that the classification accuracy results using the Support Vector Machine algorithm with 90:10 data splitting ratio using the TFIDF extraction feature is superior with an accuracy value of 85.185% and F1-Score 72.413%, which is better when compared to the Bag of words extraction feature which produces an accuracy value of 83.333% and F1-Score 66.666%. As for this study, Twitter users tend to give opinions with negative sentiments, which contain complaints and discomfort regarding the implementation of the LSSR policies, both the first LSSR and the second LSSR. Finally, the results of this research are also expected to be input for the government when making better policies in the future.