Analisis Sentimen Terhadap Ulasan Aplikasi Info BMKG Menggunakan Logistic Regression dan XGBoost

Abstract

Public opinion through user reviews plays an important role in evaluating the quality of digital public services, especially in mobile-based government applications. This study aims to classify sentiment from 5,000 user reviews of the Info BMKG application using machine learning techniques. The text data underwent pre-processing, including lowercasing, cleaning, normalization, stopword removal, and stemming to improve data quality. Feature extraction was performed using Term Frequency–Inverse Document Frequency (TF-IDF), while class imbalance was handled through the application of the Synthetic Minority Oversampling Technique (SMOTE). Two classification models were tested: Logistic Regression and Extreme Gradient Boosting (XGBoost). Tuning of hyperparameters was conducted through Grid Search combined with five-fold cross-validation to find the optimal model configuration. Evaluation results show that Logistic Regression achieved the best performance with 86.4% accuracy and a weighted F1-score of 0.86, while XGBoost achieved 85.7% accuracy. Both models demonstrated consistent ability in detecting positive and negative sentiment. The findings suggest that Logistic Regression is a suitable and reliable approach for sentiment analysis in short-text reviews within digital public service applications.