Afi Ghufran Yuda, -
(2024)
COMPARISON OF SERVICE AND EASE OF E-COMMERCE USER APPLICATIONS USING BERT.
Jurnal Sistem Cerdas.
ISSN 2622-8254
Abstract
The development of e-commerce has transformed shopping patterns by harnessing the internet, enabling consumers to shop online. In Indonesia, e-commerce has experienced rapid growth, with numerous options such as Tokopedia, Shopee, and Lazada, leading to intense competition. Sentiment analysis using machine learning techniques has become crucial for understanding consumer views on these e-commerce services. This study analyzes user comments on Tokopedia, Shopee, and Lazada e-commerce platforms from Instagram social media, totaling 3600 data comments, using the Bidirectional Encoder Representations from Transformers (BERT) model with 10, 20, and 30 epochs, a batch size of 32, and several splitting data. Sentiment analysis utilizes 3 types of labels: positive, neutral, and negative. The final results of the study include the performance analysis of the BERT model, as well as comparisons for each predefined category, namely Promotions & Offers, and Services. The final results of the model indicate good performance, with accuracy rates of 95%, 97%, and 99%, respectively.
Actions (login required)
 |
View Item |
