Aprilia, Eka Fahira (2025) Analisis Sentimen Berbasis Aspek Terhadap Persepsi Pengguna Aplikasi Dompet Digital OVO Menggunakan Support Vector Machine (SVM). Undergraduate thesis, UPN Veteran Jawa Timur.
![[img]](https://repository.upnjatim.ac.id/style/images/fileicons/text.png) |
Text (Cover)
21082010218-cover.pdf Download (2MB) |
|
Text (Bab 1)
21082010218-bab1.pdf Download (448kB) |
|
Text (Bab 2)
21082010218-bab2.pdf Restricted to Repository staff only until 5 June 2027. Download (914kB) |
|
Text (Bab 3)
21082010218-bab3.pdf Restricted to Repository staff only until 5 June 2027. Download (1MB) |
|
Text (Bab 4)
21082010218-bab4.pdf Restricted to Repository staff only until 5 June 2027. Download (4MB) |
|
Text (Bab 5)
21082010218-bab5.pdf Download (213kB) |
|
Text (Daftar Pustaka)
21082010218-daftarpustaka.pdf Download (255kB) |
|
Text (Lampiran)
21082010218-lampiran.pdf Restricted to Repository staff only until 5 June 2027. Download (497kB) |
Abstract
The advancement of financial technology in Indonesia has driven the widespread adoption of digital wallets as a practical payment alternative. One of the most popular digital wallet applications in Indonesia is OVO, which, despite having a high number of active users, continues to receive various negative perceptions from its users. This study aims to categorize topics in OVO user reviews using the Latent Dirichlet Allocation (LDA) method to identify key aspects, and develop an aspect-based sentiment analysis model using the Support Vector Machine (SVM) algorithm, implemented within a web-based system. Review data were collected from OVO application versions 3.115 to 3.119 available on the Google Play Store and Apple Store. The LDA modeling results identified four main aspects: Transaction Efficiency, User Experience, Account Access and Registration, and Balance and Charges. However, the accuracy of automatic aspect labeling by LDA compared to manual labeling was only 11.46%, increasing to 40.60% after keyword refinement, indicating LDA’s limitations in understanding review context. The best classification model used an SVM with an RBF kernel without resampling, achieving a macro average F1-Score of 0.715 and a Hamming Loss of 0.099 on the modeling dataset. The model was further evaluated using system validation data, resulting in an F1-Score of 0.794 and a Hamming Loss of 0.102. The application of ML-SMOTE oversampling did not yield significant improvements. The developed web-based system can predict aspects and sentiments of user reviews interactively via text input or CSV file, presenting the prediction results through informative visualizations. Keywords: Aspect-Based Sentiment Analysis, Latent Dirichlet Allocation, Support Vector Machine, OVO, User Reviews.
| Item Type: | Thesis (Undergraduate) | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Contributors: |
|
||||||||||||
| Subjects: | T Technology > T Technology (General) > T58.6-58.62 Management Information Systems | ||||||||||||
| Divisions: | Faculty of Computer Science > Departemen of Information Systems | ||||||||||||
| Depositing User: | Eka Fahira Aprilia | ||||||||||||
| Date Deposited: | 13 Jun 2025 08:45 | ||||||||||||
| Last Modified: | 13 Jun 2025 08:45 | ||||||||||||
| URI: | https://repository.upnjatim.ac.id/id/eprint/37153 |
Actions (login required)
 |
View Item |