Analisis Sentimen Publik Terhadap Elektabilitas Ganjar Pranowo di Tahun Politik 2024 di Twitter dengan Algoritma KNN dan Naïve Bayes

Abstract

The political year for 2024 has now increasingly entertained all Indonesian people to hold a democratic party. Various political parties have become quite dramatic in expressing their coalitions and declaring their alignment with several presidential candidates that are known to the whole community. The electability of each presidential candidate that is determined is increasingly interesting and hotly discussed, which often makes anyone take action to voice their partisanship between the pros and cons. One of them is Ganjar Pranowo, who is a political figure for the governor of Central Java. Recently, in the middle of 2023, a political party has proposed him to advance to the seat of head of state as a presidential candidate for the upcoming 2024 election. With the existence of various polemics of opinion from various layers of society, this is the right moment to carry out an analysis as a form of polarization unanimity which is presented from various public opinions as a general description and an outline in sentiment in the form of information on the conclusions of public opinion. The stages in this research began with conducting a literature study and exploring studies related to opinions and alignments with public sentiment regarding the electability of Ganjar Pranowo as a presidential candidate, and then collecting opinion data from Twitter on the electability of Ganjar Pranowo. At the experimental stage, the authors divided the data with a percentage of 80% training data and 20% testing data. The modeling used is K-Nearest Neighbor (KNN) and Naïve Bayes to classify text data as well as make comparisons of the two. In the implementation process, the author uses python as a programming language in building the model. Confusion Matrix is used for every performance evaluation related to model accuracy in each algorithm. The results showed that the division of training data and testing data and the value of k in the K-Nearest Neighbor (KNN) model greatly affect the accuracy of the model. From the test results on the comparison of the two models, the K-Nearest Neighbor model has the best accuracy with an accuracy value of 99% of the K-Nearest Neighbor with an accuracy value of 96%. The percentage of sentiment with a comparison of 96.6% positive sentiment and 3.4% negative sentiment concluded that most people still dominate positive sentiment.