Rule-based Sentiment Analysis and Evaluation using Twitter Data in The 2024 Indonesian Presidential Election

Sentiment analysis, also known as opinion mining or polarity classification, has been used to determine public perception of an entity, such as a brand, company, country, person, or program. Currently, sentiment analysis is also employed in the political domain to predict the winner of general elections, determine candidates’ reputation and electability, or recommend candidates in a general election. The rule-based method is one of the most frequently used methods in sentiment analysis. However, not many works of literature have evaluated its effectiveness and accuracy often. This research contributes to discussing some evaluations of the rule-based method by comparing the prediction vs actual results. The method applied in this research is rule-based sentiment analysis with the Natural Language Tool Kit (NLTK) to predict which candidate will get a higher vote based on the Twitter data toward presidential candidates in the general election held in Indonesia in February 2024. The rule-based method is applied in this research by using predefined rules to automatically label input data and identify the polarity of user sentiments. The data in this research was taken from Twitter during 11-31 October 2023. Based on the sentiment analysis results, we predict that candidate Ganjar Pranowo has slightly higher positive sentiment than candidate Anies Baswedan. Neutral sentiment is very dominant, with a percentage of 70% for both candidates. The result indicates that most citizens are still undecided and swing voters. However, based on the official results of the presidential election released on March 2024, candidate Anies Baswedan got a higher vote than Ganjar Pranowo. This proves that the number of neutral sentiments does not contribute much to the prediction, reducing the accuracy of sentiment analysis. We discover that the amount of data, the period of the data taken, and the method used extensively determined the accuracy of the prediction results.