PENGEMBANGAN METODE DETEKSI PERGESERAN SENTIMEN DINAMIS MENGGUNAKAN INDOBERT DAN ANALISIS TEMPORAL UNTUK EVALUASI APLIKASI HALODOC

Alwi, Pratama (2026) PENGEMBANGAN METODE DETEKSI PERGESERAN SENTIMEN DINAMIS MENGGUNAKAN INDOBERT DAN ANALISIS TEMPORAL UNTUK EVALUASI APLIKASI HALODOC. S2 thesis, Universitas Mataram.

[img] Text
LAPORAN TESIS_ALWI PRATAMA_I2S32410002.pdf
Restricted to Repository staff only

Download (2MB)

Abstract

The rapid growth of telemedicine in Indonesia makes Google Play Store reviews vital for service evaluation. However, traditional static sentiment analysis faces a specific limitation: it fails to capture the chronological shifts in user perception and ignores the context of when a sentiment change occurs. To address this, this study aims to detect temporal sentiment changes by developing a dynamic sentiment shift detection method. The proposed method integrates IndoBERT with keyword-based similarity and a temporal comparison mechanism (comparing period T against T-1) to identify transitions in user opinions and detect sentiment anomalies. Experimental results show that the IndoBERT model achieved a high classification accuracy of 0.95 and a weighted average F1-score of 0.95. Evaluation through User Acceptance Testing yielded an average score of 4.37 for mapping logic and 4.47 for management utility, with 90% of respondents recommending the system. This research contributes a novel computational framework for longitudinal sentiment tracking, providing a theoretical basis for adaptive NLP-based evaluation in high-velocity digital healthcare data.

Item Type: Thesis (S2)
Keywords (Kata Kunci): Dynamic Sentiment Analysis, Halodoc, IndoBERT, Telemedicine, Temporal Analysis.
Subjects: T Technology > T Technology (General)
Divisions: Fakultas Teknik
Depositing User: Imam Sanjaya
Date Deposited: 03 Aug 2026 06:58
Last Modified: 03 Aug 2026 06:58
URI: http://eprints.unram.ac.id/id/eprint/54607

Actions (login required)

View Item View Item