Analisis Sentimen Terhadap Pemilihan Umum Indonesia di Media Sosial Twitter dengan Metode Naïve Bayes Classifier

Abstract

This research aims to analyze public sentimen regarding the election in Indonesia using the Naïve Bayes Classifier method, chosen for its speed, accuracy, and ability to classify text data. The results show that the choice of sentimen lexicon, such as Indonesian Sentimen (InSet), significantly affects the accuracy of the analysis. A larger data set improves the model’s performance, although imbalanced class distribution remains a challenge. Preprocessing techniques such as tokenization with NLTK Word Tokenizer, feature extraction with CountVectorizer, and Hold-Out validation play key roles. With an
accuracy of 67%, this study highlights the importance of developing an Indonesian sentimen lexicon and strategies to handle data imbalance