Sentiment analysis and topic modelling of 2018 central java gubernatorial election using twitter data

Twitter as a social media widely used by the public at this time. Public opinion divided into two sentiments, namely positive or negative sentiments. Public opinion on social media about a figure who runs in an election is not always in line with the actual results of the general election. This research aims to analyze sentiment and model the topic of public opinion in the 2018 Central Java Gubernatorial Election from social media twitter. So, the sentiment classification of opinions is done to predict which candidate pairs will win in this election, by looking at the most positive sentiments. The dataset used is a dataset of Indonesian tweets with the keywords ganjarpranowo and sudirmansaid for a classification model of 1600 tweets and implementation of a classification model of 1000 tweets. Naïve Bayes and SVM to develop the sentiment classification model. Latent Dirichlet Allocation (LDA) to identify patterns and find a topic from the relationship between Twitter sentiment data. The results of sentiment analysis show that SVM has the highest accuracy value than Naïve Bayes of 92.9%. The prediction results from the SVM classification model's implementation were won by the pair Ganjar Pranowo-Taj Yasin with 826 positive tweets and found two dominant topic that appeared on the positive and negative sentiments of each candidate.