Analisis Topic Modelling pada Ulasan Aplikasi Shopee di PlayStore Menggunakan Latent Direchlet Allocation (LDA)

Abstract

Shopee is a popular e-commerce platform in Indonesia. To improve service quality, data analysis is
needed to understand user responses and preferences. This study aims to conduct data analysis and topic
modelling using the Latent Dirichlet Allocation (LDA) method on 33,896 user review data of the Shopee
application on PlayStore. LDA modelling was performed by considering the parameters passes and iterations.
Passes with values of 5, 10, 15, and 20 were tested with a combination of iterations values of 50 and 100 and a
random state of 142. During the testing, it was found that the highest average coherence score was achieved by
using passes 10 and iterations 50. Therefore, passes 10 and iterations 50 were selected as the final LDA model
parameters, resulting in seven topics. The topics that emerged from the analysis include user satisfaction, shipping
and product suitability, user trust, user experience, service and support, application, and promotional offers. The
"application" topic was found to be the most critical and received the highest number of rating 1. High-weighted
words on the application topic such as "slow," "heavy," and "crash" indicate that issues related to the
application's system performance and functionality need to be addressed and improved.