Enhancing Tourism Industry Insights: LeveragingLDA and KMeans for Topic Extraction in Indonesian OTAs' Reviews
Laurentia Alyssa Castilani, Natasha Hartanti Winata, Nathania Christy Nugraha, Alexander Agung Santoso Gunawan, Karli Eka Setiawan, Lili Ayu Wulandari · 2024
Indonesia's tourism sector, boosted by its captivating landscapes, has seen a rise in the popularity of Online Travel Agencies (OTAs) like Traveloka, Tiket.com, and Agoda. To effectively assist tourists, OTAs are anticipated to comprehend tourists' needs and interests, with insights gathered through scraping data from the Google Play Store. Through topic extraction from customers' reviews, this research aims to optimize the tourist experience and offer actionable insights for the tourism industry's marketing strategies. Furthermore, it endeavors to eliminate the reliance on brute force approaches in determining the candidate number of topics. To achieve the research goal, this research employs topic modeling, specifically Latent Dirichlet Allocation (LDA) enhanced through the K- means elbow method. The evaluation of the optimal topic number utilizes a coherence score, while human judgement serves as a quantitative metric for overall performance, with a remarkable validity rate across three platforms: Agoda (91%), Tiket.com (88%), and Traveloka (94%). This highlights the approach's effectiveness in accurately identifying and categorizing topics. In conclusion, the model accurately discerned essential topics, revealing that the majority of the reviews on these OTAs focus on the transaction and refund systems on each platform. Furthermore, this model offers promising recommendations to enhance the understanding and response to customers' reviews, facilitating faster and more effective responses. These insights are invaluable for continuous development and improvement strategies within the dynamic changes in the tourism industry across these influential platforms.