Putri, Nabila Rizky Amalia (2024) Analisis Sentimen Pada Ulasan Aplikasi Digital Korlantas Polri Menggunakan Model DistilBERT. Undergraduate thesis, UPN Veteran Jawa Timur.
![[img]](https://repository.upnjatim.ac.id/style/images/fileicons/text.png) |
Text (Cover)
20083010018-Cover.pdf Download (1MB) |
|
Text (Bab 1)
20083010018-Bab 1.pdf Download (203kB) |
|
Text (Bab 2)
20083010018-Bab 2.pdf Restricted to Repository staff only until 29 July 2026. Download (743kB) |
|
Text (Bab 3)
20083010018-Bab 3.pdf Restricted to Repository staff only until 29 July 2026. Download (401kB) |
|
Text (Bab 4)
20083010018-Bab 4.pdf Restricted to Repository staff only until 29 July 2026. Download (1MB) |
|
Text (Bab 5)
20083010018-Bab 5.pdf Download (186kB) |
|
Text (Daftar pustaka)
20083010018-Daftar Pustaka.pdf Download (421kB) |
|
Text (Lampiran)
20083010018-Lampiran.pdf Restricted to Repository staff only Download (419kB) |
Abstract
The application of digitalisation in public services by Korlantas Polri helps speed up administration, expand access, and improve service quality. Launched in April 2021, the Korlantas Polri Digital app has been downloaded more than 5 million times on the Google Play Store with a rating of 3.7 and around 110 thousand reviews. Many criticisms can affect the app's reputation, so sentiment analysis is needed to classify user reviews into positive, negative, or neutral, helping developers recognise the app's shortcomings. This study uses DistilBERT, a deep learning model distilled from BERT, to evaluate the effectiveness of review sentiment analysis. Data was collected from user reviews on the Google Play Store in the time span of 1 September 2023 to 31 May 2024, resulting in 8,752 reviews for the dataset. The model was evaluated into three data ratios: 60:20:20, 70:15:15, and 80:10:10. The results of this study showed the best performance at 80:10:10 ratio, achieving 88% accuracy using hyperparameters (batch size 16, learning rate 2e-5, and epoch 10). Increasing the ratio of training data shows a positive impact on model performance. However, the model still struggles to classify neutral sentiment, indicating the need for further improvement in detecting this class. The results of this study are expected to provide important insights for the development of more sophisticated and reliable sentiment analysis systems in the future.
| Item Type: | Thesis (Undergraduate) | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Contributors: |
|
||||||||||||
| Subjects: | Q Science > QA Mathematics > QA76.6 Computer Programming T Technology > T Technology (General) |
||||||||||||
| Divisions: | Faculty of Computer Science > Departemen of Data Science | ||||||||||||
| Depositing User: | Nabila Rizky Amalia Putri | ||||||||||||
| Date Deposited: | 30 Jul 2024 03:49 | ||||||||||||
| Last Modified: | 30 Jul 2024 03:49 | ||||||||||||
| URI: | https://repository.upnjatim.ac.id/id/eprint/27993 |
Actions (login required)
 |
View Item |