Saputra, Bima Cahya (2026) Perbandingan Metode Naïve Bayes Dan Svm Terhadap Analisis Sentimen Ulasan Pengguna Aplikasi Brainly. Undergraduate thesis, Universitas Hayam Wuruk Perbanas.
![[img]](http://eprints.perbanas.ac.id/style/images/fileicons/text.png) |
Text
artikel ilmiah.pdf Restricted to Registered users only Download (967kB) |
|
|
Text
cover.pdf Restricted to Registered users only Download (655kB) |
|
|
Text
bab i.pdf Download (239kB) | Preview |
|
|
Text
bab ii.pdf Download (550kB) | Preview |
|
|
Text
bab iii.pdf Restricted to Registered users only Download (593kB) |
|
|
Text
bab iv.pdf Restricted to Registered users only Download (1MB) |
|
|
Text
bab v.pdf Download (252kB) | Preview |
|
|
Text
lampiran.pdf Restricted to Registered users only Download (1MB) |
Abstract
Sentiment analysis is a text-mining technique used to identify users’ opinions and emotional tendencies toward an application. This study aims to compare the performance of the Naïve Bayes and Support Vector Machine algorithms in classifying user reviews of the Brainly application. The dataset consists of 1,000 reviews collected through web scraping from the Google Play Store. Sentiment labeling was performed automatically using the IndoBERT model, so the labels generated are pseudo-labels rather than manual annotations. The reviews were categorized into three classes: positive, negative, and neutral. Feature extraction was conducted using the TF-IDF method, and the data were split into training and testing sets with a 70:30 ratio. Results showed that Naïve Bayes achieved an accuracy of 74%, but performed poorly in recognizing negative (recall 0.04) and neutral (recall 0.00) classes, making it biased toward the positive class. Meanwhile, SVM achieved an accuracy of 79% with more balanced performance across classes, including improved detection of minority classes. The evaluation of both algorithms is based on pseudo-labels produced by IndoBERT, which is efficient for large datasets but limited by the model’s accuracy. This study provides both practical and academic insights into the effectiveness of text classification methods for Indonesian-language reviews. In conclusion, SVM is more effective than Naïve Bayes in handling sentiment classification on Indonesian-language review datasets, with evaluations relying on IndoBERT pseudo-labels.
| Item Type: | Thesis (Undergraduate) |
|---|---|
| Subjects: | 000 - COMPUTER SCIENCE, INFORMATION, GENERAL WORKS > 000 - 009 COMPUTER SCIENCE, INFORMATION, GENERAL WORKS > 000 - COMPUTER SCIENCE, INFORMATION, GENERAL WORKS |
| Divisions: | Bachelor of Informatics |
| Depositing User: | BIMA CAHYA SAPUTRA |
| Date Deposited: | 16 Apr 2026 04:45 |
| Last Modified: | 16 Apr 2026 04:45 |
| URI: | http://eprints.perbanas.ac.id/id/eprint/14004 |
Actions (login required)
 |
View Item |