Geo-Sentiment Analysis as a Location-Based Opinion Analysis System on Public Opinion Data about Governor Candidates

Abstract

Ahead of governor elections, there were a lot of news and opinions related to the candidates through social media. The candidates could map the positive public opinions as their political supports that need to be strengthened, and the negative opinions that need for correction. To map those opinions, it is necessary for an opinion classification system from textual opinions. It became the focus of this research. The system was designed to work on textual opinions in Bahasa since the proposed case study was the opinion of East Java governor candidates mainly written in Bahasa. Classification method that was used to classify the opinions in this system, is Naive Bayes Classifier (NBC). The opinions would be classified into 2 classes, negative and positive opinion. The classified opinions then grouped by region. It would make users easier to map the opinion in each region. The visualization became more user-friendly since the count of classified opinion displayed as a pie chart on a geographical mode or a map. After testing on the classification results, the accuracy value that we got was 78%. It indicated that NBC could perform very well as a simple text classification method with a good result.