Purba, Viviana (2026) ANALISIS EMOSI BERBASIS ASPEK PADA ULASAN APLIKASI IDENTITAS KEPENDUDUKAN DIGITAL MENGGUNAKAN INDOBERT. Undergraduate thesis, UPN Veteran Jawa Timur.
|
Text (Cover)
22082010152_Cover.pdf Download (1MB) | Preview |
|
|
Text (Bab 1)
22082010152_Bab 1.pdf Download (174kB) | Preview |
|
![[img]](https://repository.upnjatim.ac.id/style/images/fileicons/text.png) |
Text (Bab 2)
22082010152_Bab 2.pdf Restricted to Repository staff only until 26 May 2029. Download (404kB) |
|
|
Text (Bab 3)
22082010152_Bab 3.pdf Restricted to Repository staff only until 26 May 2029. Download (627kB) |
|
|
Text (Bab 4)
22082010152_Bab 4.pdf Restricted to Repository staff only until 26 May 2029. Download (1MB) |
|
|
Text (Bab 5)
22082010152_Bab 5.pdf Download (139kB) | Preview |
|
|
Text (Daftar Pustaka)
22082010152_Daftar Pustaka.pdf Download (154kB) | Preview |
|
|
Text (Lampiran)
22082010152_Lampiran.pdf Restricted to Repository staff only until 26 May 2029. Download (882kB) |
Abstract
The Identitas Kependudukan Digital (IKD) application launched by Ditjen Dukcapil Kemendagri has been downloaded by more than 10 million users on Google Play Store with a rating of 3.2 from approximately 70,700 reviews. User reviews serve as a significant indicator reflecting service quality and influencing public trust and acceptance of digital government services, while also serving as an essential resource for developers in identifying improvement priorities. However, general sentiment analysis is considered inadequate as it is unable to specifically identify the aspects and sources of issues discussed in reviews. This study applies Aspect-Based Emotion Analysis (ABEA) with multilabel classification using an end-to-end IndoBERT model on Indonesian-language reviews collected from Google Play Store covering the period of June 2024 to November 2025. To identify the dominant service aspects, two topic modeling methods were compared, namely Latent Dirichlet Allocation (LDA) and BERTopic. Evaluation results show that BERTopic outperformed LDA with a coherence score of 0.6196 and diversity of 0.92, compared to LDA which yielded a coherence score of 0.5083 and diversity of 0.9333. BERTopic successfully identified five more specific and granular aspects: Login & Akses Akun, Scan Barcode ke Dukcapil, Verifikasi Foto Wajah, Dokumen & Layanan Digital, and Kompatibilitas Perangkat Android. These aspects were subsequently used in manual data labeling by three human annotators and two AI annotators (Claude and ChatGPT), with inter-annotator agreement measured using Krippendorff's Alpha ranging from 0.71 to 0.73. The classification model was developed through eight experimental scenarios that progressively combined preprocessing, loss functions (standard BCE, Weighted BCE with pos_weight, and Focal Loss), threshold tuning, and variations in data split ratios. The best-performing model used a combination of preprocessing, Focal Loss, threshold of 0.4, and a 60:20:20 split ratio, achieving an F1 Score Macro of 0.3916 (a 24.1% improvement over baseline), F1 Score Micro of 0.9134, Recall of 0.9423, and Hamming Loss of 0.0308. As a final output, an aspect-based emotion classifier website was developed using the Flask framework to present complex classification results into informative and more easily understood information.
| Item Type: | Thesis (Undergraduate) | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Contributors: |
|
||||||||||||
| Subjects: | Q Science > Q Science (General) Q Science > QA Mathematics > QA76.6 Computer Programming Q Science > QA Mathematics > QA76.87 Neural computers |
||||||||||||
| Divisions: | Faculty of Computer Science > Departemen of Information Systems | ||||||||||||
| Depositing User: | Viviana Purba | ||||||||||||
| Date Deposited: | 26 May 2026 03:47 | ||||||||||||
| Last Modified: | 26 May 2026 03:47 | ||||||||||||
| URI: | https://repository.upnjatim.ac.id/id/eprint/52722 |
Actions (login required)
 |
View Item |