Multi Aspect Based Sentiment Analysis pada Ulasan Play Store untuk Benchmark Aplikasi Investasi Reksa Dana Menggunakan IndoBERT

Rakasiwi, Adam Idhofi (2026) Multi Aspect Based Sentiment Analysis pada Ulasan Play Store untuk Benchmark Aplikasi Investasi Reksa Dana Menggunakan IndoBERT. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 22082010026-cover.pdf Download (665kB)
	Text (Bab 1) 22082010026.-bab1.pdf Download (517kB)
	Text (Bab 2) 22082010026.-bab2.pdf Restricted to Repository staff only until 14 July 2029. Download (484kB)
	Text (Bab 3) 22082010026.-bab3.pdf Restricted to Repository staff only until 14 July 2029. Download (381kB)
	Text (Bab 4) 22082010026.-bab4.pdf Restricted to Repository staff only until 14 July 2029. Download (2MB)
	Text (Bab 5) 22082010026.-bab5.pdf Download (227kB)
	Text (Daftar Pustaka) 22082010026.-daftarpustaka.pdf Download (194kB)
	Text (Lampiran) 22082010026.-lampiran.pdf Restricted to Repository staff only until 14 July 2029. Download (454kB)

Abstract

The growth of mutual fund investment apps has increased the need to evaluate service quality based on user experience. Reviews on the Google Play Store can be an important source of information, but the unstructured review texts and general ratings aren't enough to explain user experiences on specific aspects. This study aims to analyze aspect-based sentiment in Bibit and Ajaib app reviews using a Multi Aspect Based Sentiment Analysis approach with the IndoBERT model, and also use it as a basis for app benchmarking. The research was conducted using the CRISP-DM framework, from business understanding to deployment. Data was collected through web scraping, then processed through initial filtering, labeling, text pre-processing, and splitting into 80% train+validation and 20% test sets. The final dataset used consisted of 4,996 reviews. The aspects analyzed include Feature & Functionality, App Interface, and User Satisfaction. The models compared were IndoBERT-Base and IndoBERT-Lite, with fine-tuning using 5-Fold Stratified Cross Validation and class weighting to handle class imbalance. Evaluation results show that IndoBERT-Base performs better than IndoBERT-Lite across all aspects, with a test set F1-score of 0.7904 for Feature & Functionality, 0.6778 for App Interface, and 0.8287 for User Satisfaction. Benchmarking results show that Ajaib has a higher proportion of positive sentiment compared to Bibit across all three aspects, while Bibit has the lowest proportion of positive sentiment in Feature & Functionality at 29.0% and the highest proportion of negative sentiment in the same aspect at 61.1%. The research results were implemented on the ReksaBench website to support easier visualization.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Wahyuni, Eka Dyar	NIDN0001128406	ekawahyuni.si@upnjatim.ac.id
Thesis advisor	Sugata, Tri Luhur Indayanti	NIDN8948770671230422	tri.luhur.fasilkom@upnjatim.ac.id

Subjects:

Q Science > Q Science (General)
Q Science > QA Mathematics > QA76.6 Computer Programming
T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Information Systems

Depositing User:

Adam Adam Rakasiwi

Date Deposited:

15 Jul 2026 02:53

Last Modified:

15 Jul 2026 03:27

URI:

https://repository.upnjatim.ac.id/id/eprint/55493