Skripsi

ANALISIS KOMPARASI KINERJA INDOBERT FINE-TUNED DAN HYBRID SUPPORT VECTOR MACHINE (SVM) UNTUK ANALISIS SENTIMEN ULASAN APLIKASI BEONE APPS

---

The low rating of the BeOne HRIS application on the Google Play Store indicates user dissatisfaction that requires strategic evaluation through automated sentiment analysis. This study aims to compare the effectiveness of a transformer-based Deep Learning model, namely IndoBERT Fine-tuned, against a hybrid approach that combines IndoBERT as a feature extractor with a Support Vector Machine (SVM), while also testing the significant impact of data pre-processing stages on classification performance. Empirical experimental results prove that the pure IndoBERT Fine-tuned architecture without conventional pre-processing intervention (raw text) is the most optimal approach, recording an accuracy and global F1-Score of 92%, which significantly outperforms the hybrid model. This study concludes that the transformer model's ability to understand the semantic context of the Indonesian language is far superior when the data is maintained as authentic compared to through aggressive text normalization, where the model is proven to be highly sensitive (95% recall) in detecting crucial technical complaints related to the biometric attendance feature, although it still faces challenges in classifying minority classes due to imbalanced data (imbalanced dataset).

---

Availability

Detail Information

Series Title

-

Call Number

T1894522025

Publisher

Indralaya : Prodi Sistem Informasi, Fakultas Ilmu Komputer Universitas Sriwijaya., 2025

Collation

xvi, 109 hlm.; ilus.; tab.; 29 cm.

Language

Indonesia

ISBN/ISSN

-

Classification

006.380 7

Content Type

Text

Media Type

-

Carrier Type

-

Edition

-

Subject(s)

Prodi Sistem Informasi
Analisis Sintimen--Komparasi Kenerja

Specific Detail Info

-

Statement of Responsibility

MI

Other version/related

File Attachment

Comments

---

You must be logged in to post a comment