Komparasi Kinerja Algoritma SVM dan Random Forest dengan Particle Swarm Optimization pada Aspect-Based Sentiment Analysis Ulasan Aplikasi Qpon

Rahma, Dea Ananda Refiza (2026) Komparasi Kinerja Algoritma SVM dan Random Forest dengan Particle Swarm Optimization pada Aspect-Based Sentiment Analysis Ulasan Aplikasi Qpon. Undergraduate thesis, UPN Veteran Jawa Timur.

This is the latest version of this item.

	Text (Cover) 22082010144-cover.pdf Download (3MB)
	Text (Bab 1) 22082010144-bab 1.pdf Restricted to Repository staff only until 20 July 2029. Download (150kB)
	Text (Bab 2) 22082010144-bab 2.pdf Restricted to Repository staff only until 20 July 2029. Download (414kB)
	Text (Bab 3) 22082010144-bab 3.pdf Restricted to Repository staff only until 20 July 2029. Download (4MB)
	Text (Bab 4) 22082010144-bab 4.pdf Restricted to Repository staff only until 20 July 2029. Download (4MB)
	Text (Bab 5) 22082010144-bab 5.pdf Download (11kB)
	Text (Daftar Pustaka) 22082010144-daftarpustaka.pdf Download (88kB)
	Text (Lampiran) 22082010144-lampiran.pdf Restricted to Repository staff only until 20 July 2029. Download (385kB)

Abstract

QPon is a digital voucher application that provides real-time promotional services across various categories, such as food, beverages, entertainment, local transportation, and lifestyle. User reviews of QPon contain opinions related to several service aspects, making general sentiment analysis insufficient to describe user experience in detail. This study aims to compare the performance of machine learning algorithms in Aspect-Based Sentiment Analysis (ABSA) of QPon application reviews by applying hyperparameter optimization using Particle Swarm Optimization (PSO). The data were collected by scraping QPon user reviews from the Google Play Store, resulting in 4,101 reviews. After filtering and preprocessing, the final dataset consisted of 3,453 reviews. The labeling process was conducted by three raters, consisting of two human annotators and one supporting annotator based on a Large Language Model, namely ChatGPT. The reliability of the labeling results was evaluated using Fleiss’ Kappa, with values of 0.9573 for sentiment labels and 0.7311 for aspect labels, indicating a high level of agreement. The final labels were determined using majority voting. Feature weighting was performed using TF-IDF with a maximum of 5,000 features. The models used in this study were Support Vector Machine (SVM) and Random Forest, evaluated through six testing scenarios, namely baseline, PSO, and SMOTE + PSO. The results show that the best model varied across aspects based on macro F1-score as the main evaluation metric. For the customer service aspect, SVM + PSO and SVM + SMOTE + PSO achieved the highest F1-score of 95.09%. For the functionality aspect, SVM + SMOTE + PSO achieved the highest F1-score of 71.94%. Meanwhile, for the transactions and payments aspect, SVM baseline and SVM + PSO achieved the highest F1-score of 65.10%. These results indicate that SVM tends to outperform Random Forest on the QPon review dataset. However, the effectiveness of PSO and SMOTE is conditional, as it is influenced by data characteristics, class distribution, and the presence of minority classes in each aspect.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Najaf, Abdul Rezha Efrat	NIDN0029099403	rezha.efrat.sifo@upnjatim.ac.id
Thesis advisor	Permatasari, Reisa	NIDN0014059203	reisa.permatasari.sifo@upnjatim.ac.id

Subjects:

T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Information Systems

Depositing User:

Dea Ananda Refiza Rahma

Date Deposited:

21 Jul 2026 01:11

Last Modified:

21 Jul 2026 02:04

URI:

https://repository.upnjatim.ac.id/id/eprint/56972