Perbandingan Logistic Regression, SVM, dan Random Forest untuk Analisis Sentimen Ulasan Aplikasi Gopay

Abstrak

The expansion of Indonesia's digital financial landscape has triggered a surge in the adoption of e-wallets, most notably GoPay. Within this context, feedback available on application platforms such as the Google Play Store serves as a crucial metric for assessing user sentiment and service quality. Sentiment analysis based on machine learning algorithms allows for systematic and objective identification of public opinion. This study used 3,000 user reviews collected through web scraping from the Google Play Store, received up to April 21, 2025, with initial labeling based on a lexicon approach. Although many studies have compared sentiment classification algorithms, there has been no research specifically comparing the performance of Logistic Regression, Support Vector Machine (SVM), and Random Forest in the context of GoPay user reviews with lexicon based labeling. This paper aims to fill the existing void by evaluating the comparative performance of three algorithms based on sentiment classification metrics. Preprocessing procedures encompassed cleaning, case-folding, stemming, slang normalization, tokenizing, filtering, and labeling to ensure data quality. The models, built within the Scikit-learn environment, were tested for accuracy, precision, recall, and F1-score. Empirical results confirm that Logistic Regression outperformed the alternatives, securing 88.16% accuracy while maintaining stability across all sentiment categories. SVM recorded 87.5% accuracy but was weak in detecting negative sentiment. Random Forest showed the lowest performance with 79.33% accuracy and less consistent classification results. Thus, Logistic Regression is recommended as the most effective algorithm for GoPay user sentiment analysis. Future research can explore deep learning-based approaches to handle higher sentiment complexity.