Klasifikasi Aspect-Based Sentiment Analysis Menggunakan SVM pada Umpan Balik Pengguna Aplikasi Tanda Tangan Digital dan E-Meterai

Narendra, Efriza Cahya (2025) Klasifikasi Aspect-Based Sentiment Analysis Menggunakan SVM pada Umpan Balik Pengguna Aplikasi Tanda Tangan Digital dan E-Meterai. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 21082010059.-cover.pdf Download (1MB)
	Text (Bab 1) 21082010059.-bab1.pdf Download (509kB)
	Text (Bab 2) 21082010059.-bab2.pdf Restricted to Repository staff only until 13 June 2027. Download (464kB)
	Text (Bab 3) 21082010059.-bab3.pdf Restricted to Repository staff only until 13 June 2027. Download (295kB)
	Text (Bab 4) 21082010059.-bab4.pdf Restricted to Repository staff only until 13 June 2027. Download (6MB)
	Text (Bab 5) 21082010059.-bab5.pdf Download (207kB)
	Text (Daftar pustaka) 21082010059.-daftarpustaka.pdf Download (205kB)

Abstract

In the digital era, digital signatures and e-Meterai have become essential solutions for digital identity. Several applications such as Privy, Vida, and Xignature are widely used and have thousands of user reviews. However, understanding user needs and complaints from these numerous reviews remains a challenge. This study aims to transform user reviews into useful information by developing an aspect-based sentiment analysis model implemented on a website platform. Review data from the Google Play Store and App Store are processed using the Latent Dirichlet Allocation (LDA) technique to identify discussed aspects. After multilabel aspect and sentiment labeling, term weighting is performed using Term Frequency-Inverse Document Frequency (TF-IDF). To address data imbalance, resampling is conducted using three methods, there are Multilabel Random Oversampling (MLROS), Multilabel Synthetic Minority Oversampling Technique (MLSMOTE), and REsampling MultilabEl datasets by Decoupling highly ImbAlanced Labels (REMEDIAL). Modeling is done using the Support Vector Machine (SVM) algorithm with various scenario combinations, including data splits, resampling methods, parameter C values, and either normal SVM or Classifier Chain strategies. Performance evaluation using Hamming Loss shows the best results in the scenario with a 70:30 data split, MLROS resampling method, C value of 1, and normal SVM, achieving a Hamming Loss of 0.0559 or 94% accuracy. This model is then implemented on a website using the Flask framework, allowing users to easily predict sentiment from text or file inputs.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Arifiyanti, Amalia Anjani	NIDN0712089201	amalia_anjani.fik@upnjatim.ac.id
Thesis advisor	Sugata, Tri Luhur Indayanti	NIP199206162024062001	tri.luhur.fasilkom@upnjatim.ac.id

Subjects:

T Technology > T Technology (General)

Divisions:

Faculty of Computer Science > Departemen of Information Systems

Depositing User:

Ms Efriza Cahya Narendra

Date Deposited:

13 Jun 2025 08:50

Last Modified:

13 Jun 2025 08:50

URI:

https://repository.upnjatim.ac.id/id/eprint/37663