Sentiment Analysis of Digital Korlantas Polri Apps Service Based on LSTM and SVM Methods

Abstract

Advancements in digital technology have encouraged numerous innovations in public services, one of which is the Digital Korlantas Polri app. This application makes it easier for the public to access traffic services such as driver’s license issuance and renewal, vehicle data checking, and accident reporting. However, despite the convenience it offers, there are still various user reviews that point to technical issues and dissatisfaction with the quality of service. This study applies sentiment analysis to understand public perception of the Digital Korlantas app, providing a basis for improving its quality. The collection of the dataset was achieved by web scraping 2,000 user reviews from the Google Play Store spanning the period from December 2023 to March 2025. The phases of the research encompass gathering data, pre-processing text, assigning sentiment labels based on lexicons, applying TF-IDF for word weighting, and performing classification using the Long Short-Term Memory (LSTM) and Support Vector Machine (SVM) algorithms. The performance of the model was assessed through a confusion matrix, utilizing accuracy, precision, recall, and F1-score as evaluation metrics. The findings indicated that, out of 2,000 reviews, 1,402 were identified as positive, 538 were categorized as negative, and 60 were considered neutral. The SVM model demonstrated the highest performance, obtaining an accuracy of 96.8%, a precision of 65.6%, a recall of 50.0%, and an F1-score of 55.0%. At the same time, the LSTM model attained an accuracy of 94.5%, with a precision of 31.5%, a recall of 33.3%, and an F1-score of 32.4%. These results show that SVM is superior at handling high-dimensional data, while LSTM remains effective at capturing long-term context patterns in review texts.