Text–emoji-based tourist perception analysis for sustainable tourismdevelopment on Gili Iyang using SVM and LDA

Safinatus Zahroh (Universitas Teknologi Yogyakarta, Yogyakarta, Indonesia)
Enny Itje Sela (Universitas Teknologi Yogyakarta, Yogyakarta, Indonesia)
Bambang Moertono Setiawan (Universitas Teknologi Yogyakarta, Yogyakarta, Indonesia)
Luther A Latumakulita (Universitas Sam Ratulangi, Manado, Indonesia)

MATRIK : Jurnal Manajemen, Teknik Informatika, dan Rekayasa Komputer
Vol. 25 No. 3 (2026)

Gili Iyang Island is widely recognized as an “Oxygen Island” with strong potential for sustainable tourism development. However, challenges related to accessibility, facilities, and service quality continue to affect tourist experiences and may hinder destination development. Previous tourism sentiment studies have mainly relied on textual data, while emotional cues conveyed through emojis have received limited attention despite their ability to enrich sentiment interpretation. In addition, sentiment classification and topic extraction are commonly performed separately, restricting a comprehensive understanding of tourist perceptions. This study aims to analyze tourist perceptions of Gili Iyang Island by integrating sentiment classification and topic extraction using text and emoji data from Instagram comments. A text mining framework was applied, including data preprocessing, rule-based sentiment labeling, TF-IDF and n-gram feature extraction, SVM-based sentiment classification, and LDA for topic extraction. The dataset consisted of 425 Indonesian-language comments collected between January and December 2025. The SVM model achieved a validation accuracy of 81.18%, indicating satisfactory performance in classifying text- and emoji-based sentiments. Topic modeling identified five dominant discussion themes: environmental conditions, tourist attractions, accessibility, facilities, and social interactions. Mapping these themes to sustainable tourism dimensions revealed that environmental and social aspects attracted greater public attention than economic aspects. Positive sentiment was primarily associated with environmental quality and tourist attractions, while accessibilityand facilities generated relatively more negative perceptions. These findings demonstrate that integrating text–emoji-based sentiment analysis with topic extraction provides a more comprehensive understanding of tourist perceptions and offers valuable data-driven insights to support sustainable tourism planning, destination management, and policy development.