A Location-Aware Culinary Recommendation System Using Content-Based Filtering and Session-Based User Profiling
The growing number of culinary establishments and digital information sources can make restaurant selection difficult, particularly for students who must consider food preferences, distance, and affordability. Existing recommendation services often emphasize popularity or general ratings and may not adapt to short-term user interests during a browsing session. This study aimed to develop a location-aware culinary recommendation system that integrates content-based filtering and session-based user profiling without requiring registration or login. The system was developed using the Rapid Application Development approach. Restaurant data were obtained from the Nominatim OpenStreetMap application programming interface. Culinary queries and restaurant descriptions were represented using term frequency–inverse document frequency and compared through cosine similarity. Candidate restaurants were filtered using Haversine-based distance calculation and price categories, while searches, price selections, and restaurant-detail clicks were aggregated into a temporary session preference vector. The final ranking combined content similarity and session relevance using weights of 0.7 and 0.3. System evaluation employed seven black-box test scenarios and User Acceptance Testing involving 21 university students and 13 questionnaire items. All black-box scenarios produced valid outputs, resulting in a 100% functional success rate. From 273 questionnaire responses, 149 were rated Very Good and 84 Good, yielding 85.35% positive responses. The proposed system successfully integrated semantic matching, geographic feasibility, affordability filtering, and short-term personalization in a lightweight web application. The findings demonstrate functional correctness and favorable initial user acceptance, although verified restaurant attributes and ranking-based evaluation are required in future studies.