Analisis Sentimen Ulasan Aplikasi dengan Multinomial Naïve Bayes, Logistic Regression, dan SVM

Abstract

The swift uptake of mobile health applications has led to an increase in user-generated feedback, providing important insights into public satisfaction. To explore user sentiments, this study analyzes 9,848 reviews from a health-oriented application utilizing three machine learning methods: Multinomial Naïve Bayes, Logistic Regression, and Support Vector Machine (SVM). The feedbacks were classified as positive or negative. The methodology included standard preprocessing such as cleaning and stemming, feature extraction using Term Frequency-Inverse Document Frequency (TF-IDF), and addressing class imbalance with the Synthetic Minority Oversampling Technique (SMOTE). Models were fine-tuned and verified through 5-fold cross-validation. Effectiveness was measured by accuracy, precision, recall, and F1-score. Logistic Regression and SVM reached the greatest accuracy at 92%, while Naïve Bayes trailed at 86%. Logistic Regression showed strong precision (95%) and recall (94%) for positive reviews, with SVM performing comparably. These results emphasize the capability of sentiment analysis in enhancing digital health services through information-based user feedback.