A Web-Based Tourism Recommendation System for Boyolali Using Content-Based Filtering and Cosine Similarity

Abstrak

Tourism is a primary economic driver for Boyolali Regency. However, destination information remains fragmented, lacking a personalized approach to meet diverse visitor preferences. To address this issue, this study developed "BoyTure", a web-based tourism application integrated with a recommendation system. The system development followed the ICONIX Process methodology, selected for its Robustness Analysis phase, which validates system logic before code implementation. The recommendation engine uses Content-Based Filtering with the Cosine Similarity algorithm, applied to a curated dataset of 74 verified destinations sourced from the Youth, Sports, and Tourism Office (Disporapar) of Boyolali Regency. Unlike standard approaches, the TF-IDF feature extraction in this system explicitly concatenates four textual attributes, destination name, category, facilities, and description, to mitigate data sparsity and enrich the semantic context. A comparative analysis justifies the selection of Cosine Similarity over Euclidean Distance or Jaccard Similarity because of its robustness in handling variable-length tourism text descriptions. Testing was conducted using the Black-Box method to ensure functional compliance, and a System Usability Scale (SUS) evaluation yielded an average score of 81.5. This SUS score demonstrates that BoyTure successfully abstracts complex algorithms into a user-friendly interface to provide accurate and personalized tourism recommendations.