Predictive Analysis of Government Application Comment on Playstore with Clustered Support Vector Machine
Shindi Shella May Wara, Andri Fauzan Adziima, Muhammad Nasrudin, Alfan Rizaldy Pratama · 2024
The rapid digital transformation in Indonesia has led the government to adopt mobile applications to enhance public services. Three such applications, "Identitas Kependudukan Digital" (Digital Identity), "Aplikasi Cek Bansos" (Social Assistance Check App), and "Digital Korlantas POLRI" (Digital Traffic Corps POLRI), aim to streamline various public services, from civil registration to social assistance and traffic management. This study employs the Clustered Support Vector Machine (CSVM) method to analyze user ratings and comments on these applications available on the Google Playstore. The analysis reveals key patterns in user feedback, categorizing them into technical issues, user satisfaction, improvement suggestions, and commendations. The CSVM method demonstrates a goodness-of-fit model comparable to SVM while also excelling in performance metrics such as accuracy, precision, and F1 score, achieving improvements on the order of 10-3. A particularly notable advantage of CSVM is its efficiency, achieving running times that are twice as fast as SVM, making it a suitable choice for processing large datasets efficiently.