Analisis Sentimen Ulasan Pengguna Layanan Publik Menggunakan Algoritma Naïve Bayes

This study aims to analyze the sentiment of user reviews of the Super App Polri application on the Google Play Store using the TF-IDF feature extraction method and the Multinomial Naïve Bayes classification algorithm. Super App Polri is a digital service of the Indonesian National Police providing SIM, STNK, SKCK renewals, and public complaint services. Data were collected via web scraping of 3,500 user reviews. Preprocessing stages—cleansing, case folding, tokenizing, stopword removal, and stemming—resulted in 3,478 valid data points. The data were extracted using TF-IDF and classified into positive, negative, and neutral sentiments. The results showed that the model achieved an 86% accuracy rate. Sentiment distribution was dominated by positive sentiment (49.7%), followed by negative (45.3%) and neutral (5.0%). The model performed well in classifying positive and negative sentiments but was less optimal for neutral sentiments due to an imbalanced dataset. This study provides an overview of user perceptions of service quality. These findings offer a strategic basis for developers to optimize payment, verification, and registration systems to increase user trust in digital police services.