Analisis Pemodelan Topik Ulasan Wisatawan Digital: Pendekatan Latent Dirichlet Allocation pada Destinasi Wisata

Abstract

The exponential growth of user-generated tourist reviews on digital platforms holds significant potential for understanding traveler experiences and perceptions of destinations. However, the vast and unstructured nature of this textual data requires reliable methods for systematic theme extraction. This study aims to map the dominant topics embedded in digital tourist reviews using the Latent Dirichlet Allocation (LDA) approach. Employing an exploratory descriptive research design, a total of 2,000 reviews from Indonesian tourism destinations were collected and processed through tokenization, normalization, and stemming, before being modeled using Gensim’s LDA. The model was evaluated using the coherence score (c_v), resulting in an optimal model with ten topics (K=10). The identified topics reflect various tourism aspects, including natural beauty, historical heritage, family activities, culinary experiences, and popular landmarks. Topic distribution analysis revealed the spread and proportion of themes across the entire corpus, offering rich contextual insights into tourist narratives. These findings contribute to data-driven destination management and form a foundation for the development of smart tourism applications, while also providing new insights into tourist preferences in Indonesia. The LDA-based thematic mapping demonstrates strong potential in supporting intelligent destination management and enhancing personalized tourism recommendation systems.