Analisis Sentimen Terkait Pelaksanaan Wifest pada Instagram dengan Metode Naïve Bayes

Abstract

Instagram comments capture public responses to an event, but reading them manually is slow and subjective. This study measures the sentiment tendency toward Wifest and assesses how well Naïve Bayes classifies it. A total of 120 @wifest comments were collected with IGCommentsExport, labeled positive or negative, and preprocessed through data cleaning, case folding, normalization, tokenization, stopword removal, and stemming. The text was then weighted with TF-IDF and classified using Naïve Bayes in RapidMiner Studio with a 70% training and 30% testing split. Of the 120 comments, 84 (70%) were positive and 36 (30%) were negative. The model achieved an accuracy of 86.11%, a precision of 85.71%, a recall of 96.00%, and a macro-average F1-score of 82.01%, exceeding a baseline that labels every comment positive (69.44%). However, negative-class recall was only 63.64%, so the model recognizes positive comments more reliably. It is concluded that Wifest drew mostly positive responses and that Naïve Bayes suits sentiment mapping of comments on this account. Organizers should use the model to estimate sentiment proportions and read negative comments directly, and future studies should add negative data and balance the classes.