Penerapan Metode Ensemble Machine Learning Dalam Analisis Sentimen Berbasis Aspek Pada Ulasan Aplikasi Wondr By BNI

Hardiartama, Rendi (2025) Penerapan Metode Ensemble Machine Learning Dalam Analisis Sentimen Berbasis Aspek Pada Ulasan Aplikasi Wondr By BNI. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 21082010094-cover.pdf Download (1MB)
	Text (Bab 1) 21082010094-bab1.pdf Download (100kB)
	Text (Bab 2) 21082010094-bab2.pdf Restricted to Repository staff only until 22 May 2027. Download (372kB) \| Request a copy
	Text (Bab 3) 21082010094-bab3.pdf Restricted to Repository staff only until 22 May 2027. Download (372kB) \| Request a copy
	Text (Bab 4) 21082010094-bab4.pdf Restricted to Repository staff only until 22 May 2027. Download (4MB) \| Request a copy
	Text (Bab 5) 21082010094-bab5.pdf Download (15kB)
	Text (Daftar Pustaka) 21082010094-daftarpustaka.pdf Download (109kB)

Abstract

The development of digital technology has driven the transformation of banking services through mobile banking, including the launch of the Wondr by BNI application as the latest innovation from Bank Negara Indonesia. Although the application has been downloaded more than 5 million times, as of the end of December 2024, its rating was only 3.8 lower than the previous application's rating of 4.5 indicating a level of user dissatisfaction. This study aims to understand user perceptions of the Wondr by BNI application by implementing Aspect-Based Sentiment Analysis (ABSA) using a stacking ensemble learning method on user reviews, in order to identify the main aspects and their associated sentiments. The data were collected through scraping from Google Play Store and App Store, followed by preprocessing and labeling, before undergoing two stages of classification: aspect identification and sentiment classification for each aspect. Evaluation results show that the stacking ensemble model without resampling yielded the best performance. In aspect classification, the model achieved an F1 score of 99.4% for the Interface aspect, 99.3% for Authentication, and 99% for Transaction. For sentiment classification per aspect, the model obtained F1-scores of 82.2% for Interface, 87.8% for Authentication, and 92.4% for Transaction. The implementation of Local Interpretable Model-Agnostic Explanations (LIME) also successfully enhanced model interpretability by highlighting the words that contributed to both aspect and sentiment classification.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Arifiyanti, Amalia Anjani	NIDN0712089201	amalia_anjani.fik@upnjatim.ac.id
Thesis advisor	Wati, Seftin Fitri Ana	NIDN0020039104	seftin.fitri.si@upnjatim.ac.id

Subjects:

T Technology > T Technology (General) > T58.6-58.62 Management Information Systems

Divisions:

Faculty of Computer Science > Departemen of Information Systems

Depositing User:

Rendi Hardiartama

Date Deposited:

23 May 2025 03:51

Last Modified:

23 May 2025 03:51

URI:

https://repository.upnjatim.ac.id/id/eprint/36431