Multinomial Naïve Bayes for sentiment analysis of Indonesian's local government performance

Abstract

Digitalization of government performance, in conveying information and getting criticism, suggestions, and complaints from the public, is currently being carried out using social media. The use of social media is a form of government responsibility and openness to society. The high number of Twitter users in Indonesia, which reaches 6.43 million, allows the government to get many responses from the public. This background provides an opportunity for the public to be able to measure government performance based on a number of criticisms, suggestions, and complaints that the government responds to. However, public sentiment towards government performance has not been used as an evaluation and benchmark for the government in determining policies. The purpose of this research is to build a social media twitter sentiment analysis system to measure public sentiment towards local government performance by implementing Multinomial Naïve Bayes. This research is divided into several stages including tweet grabbing, manual tweet filtering, tweet labeling, split tweets, preprocessing tweets, term frequency, classification, and evaluation. The tweet retrieval process was carried out on 1 June - 31 July 2020 with 2000 tweets used from the total tweets obtained after manual filtering was carried out. This study shows that the sentiment analysis carried out obtained an accuracy of 80%, a precision of 78%, and a recall of 82%.