<![CDATA[The Mitra Darat application is an Android-based application available on the Google Play Store, developed by the Directorate General of Land Transportation (Ministry of Transportation). User reviews on this platform enable direct communication with developers, offering valuable feedback for service enhancement and future development. For that reason, aspect-based sentiment analysis is needed to help organizations monitor product sentiment in user feedback, and understand user needs. This research aims to implement aspect-based sentiment analysis on Mitra Darat application user reviews to generate insights via a system dashboard. Comparing Naive Bayes (NB) and Support Vector Machines (SVM) for machine learning models with the addition of pre-trained Indobert as word embedding, SVM showed superior performance with an accuracy score of 94% for aspect classification and 90% for sentiment classification, compared to Naive Bayes with scores of 84% and 78% respectively. The trained Support Vector Machine model (SVM) was then utilized to analyze 967 reviews of the Mitra Darat application for 2023. The results of the analysis are presented on a dashboard page with summary information, which shows that the overall user sentiment is 52.3% positive, 10% neutral, and 37.7% negative. In terms of sentiment polarity by aspect, the system aspect is 29% positive, 8% neutral, and 63% negative, meaning that some bugs and issues have been found in the application, so it can be evaluated for future system development. The service aspect is 62% positive, 27% neutral, and 11% negative, which means that the free mudik service is quite well organized]]>
Copyrights © 2024
