Prediksi Sentimen dan Pemodelan Topik dari Ulasan Aplikasi Identitas Kependudukan Digital

Abstract

In this digital era, the Indonesian government is digitizing services, for example, the Digital Population Identity (IKD) in 2023 as a continuation of the Identity card (KTP). The IKD application has received mixed reviews on the Google PlayStore and Apple AppStore. From these various reviews, sentiment analysis can be carried out to determine the level of user satisfaction, as well as topic modeling to determine topics that are frequently discussed by users. This study aims to predict sentiment using the Naive Bayes and PNN methods, and topic modeling using the LDA method. The results of sentiment prediction show that the fourth scenario has the highest accuracy with 97.12%, and positive and negative f1-scores of 97% each. The results of topic modeling show the best coherence for positive reviews on topic 2 (0.4076) and for negative reviews on topic 10 (0.5564). Interpretation of the results shows that positive reviews include identity verification, application development suggestions, ease of use, KTP digitization, and user experience, while negative reviews are related to technical constraints, access and installation, network constraints, features and data security, and application registration.