An Attribute-Based Explainable Recommendation System for Culinary Tourism

Abstract

The rapid growth of digital tourism platforms has increased the availability of culinary destination information, making it challenging for users to identify destinations that best match their preferences. Recommendation systems have become an effective solution for providing personalized recommendations; however, most conventional approaches only present recommendation results without explaining the underlying reasons, thereby reducing user trust and transparency. This study proposes an Explainable Content-Based Recommendation System for culinary tourism by integrating TF-IDF feature representation with the Cosine Similarity algorithm. Culinary destination data were collected and preprocessed through text normalization and TF-IDF vectorization to represent the characteristics of each culinary destination. Cosine Similarity was then applied to measure the similarity between user preferences and culinary destinations, while an explanation module was developed to provide understandable reasons for each recommendation based on shared culinary attributes, including category, main ingredients, flavor characteristics, and price range. The proposed system was implemented as a web-based application using Python, Flask, and MySQL. Experimental results show that the system achieved a Precision@5 of 0.92, a Recall@5 of 0.88, and an average response time of 0.73 seconds, indicating that the proposed approach is capable of generating relevant recommendations with efficient computational performance. Furthermore, the explanation module enhances recommendation transparency by enabling users to understand the factors contributing to each recommendation. These findings demonstrate that the proposed system provides a practical, lightweight, and explainable solution for culinary tourism recommendation and has the potential to improve user experience in selecting culinary destinations.