ANALISIS SENTIMEN MENGGUNAKAN NBC DAN SVM UNTUK MENGETAHUI RESPON TERHADAP KINERJA PRESIDEN MELALUI MEDIA INSTAGRAM

Abstract

The president's performance can be revealed through the public's response to various presidential programs in the economic or industrial sector which are uploaded on the President's official Instagram media. The public's response to each uploaded post is interesting data to analyze. This research aims to conduct sentiment analysis of the public's response to the president's performance, by uploading the president's activities on Instagram using the Naive Bayes Classifier (NBC) and Support Vector Machine (SVM) methods.  A total of 18,660 responses to the President's 5 Instagram posts on economic and industrial activities between March and September 2024. The data was then filtered to just 15,306 comments which were determined as a dataset. Dataset labeling before classification is carried out using the lexicon based method. The research results show that sentiment towards the president's performance is 28.3% positive, 36.8% neutral and 34.9% negative. Meanwhile, the NBC and SVM classification results obtained showed an accuracy of 68.7%, precision of 68.5%, recall of 67.8% and F1-score of 68.2% for the NBC algorithm. The SVM algorithm produces an accuracy of 80.5%, a precision value of 80.5%, a recall value of 79.6%, and an F1-score value of 79.6%. It can be seen that the SVM algorithm is superior to the NBC algorithm in all classification parameters. From several training and testing compositions, it was also found that the best composition was the 70:30 composition.