Sentiment Analysis of State Capital Relocation of Indonesia using Convolutional Neural Network

Abstract

The current government policy led by President Joko Widodo regarding relocating the capital city from Daerah Khusus Ibukota (Capital Special Region) of Jakarta to East Kalimantan has drawn a variety of comments, ranging from praise, criticism, suggestions, innuendo to hate speech. This is supported by many Indonesian political figures who have Twitter accounts to provide support or opinions on this policy. This study aims to analyze the sentiments about this issue. There is very varied sentiment from this issue, either positive or negative responses. This research used Convolutional Neural Network (CNN) algorithm as a part of the Deep Learning method to classify sentiments towards government policy on moving capital city of Indonesia with data obtained from Twitter. The process begins with text pre-processing containing case folding, tokenizing, stop-words removing, stemming, and changing the emoticon to word. Then, the word embedding process used Word2Vec. The result of experiment of CNN algorithm with 1,515 tweets in the Indonesian language and 15 times of experiment shows that the average accuracy is 66.68% with the highest accuracy is 70.3%. The experiment used five training and testing data splitting variations, with three epochs, among others: 10 epochs, 30 epochs, and 100 epochs.