Analisis Sentimen Berbasis Aspek pada Aplikasi LinkAja Berdasarkan Ulasan Pengguna dengan Menggunakan Latent Direchlet Allocation dan Attention Based-LSTM

Abstract

One of the financial technologies that is increasingly being used in Indonesia is the Digital Payment Systems. LinkAja, as one of the popular applications among the public, still faces various issues in its use, such as slow customer service response to complaints, failed transactions, long refund processes, and frequent errors due to application updates. It is crucial to understand user opinions about this application to improve service performance, providing a better experience for users. However, the large number of reviews makes manual analysis inefficient. Therefore, this study develops an automated system to process and categorize user reviews of the LinkAja application based on aspects of ease of use, customer service, quality, and speed using Latent Dirichlet Allocation (LDA) and Attention-Based Long Short Term Memory (LSTM) methods. This study shows that the Attention-Based LSTM model achieves an accuracy of 90% for ease of use, 92% for customer service, 85% for quality, and 92% for speed, with an overall average accuracy of 89,75%.