Year: 2026
Type: article
Abstract: The growth of Sukabumi's culinary industry has significantly increased dining options, creating constraints for users in finding places that match personal preferences and situational conditions. This study develops a prototype of a culinary recommendation system based on Context-Aware Hybrid Filtering. The model integrates Content-Based Filtering (CBF), Singular Value Decomposition (SVD) Matrix Factorization, the Haversine Formula, and weather adaptation factors. Using the Cross-Industry Standard Process for Data Mining (CRISP-DM), the dataset was collected via web scraping from Google Maps, yielding 715 records with an extreme matrix sparsity of 94.81%. CBF was constructed using TF-IDF keyword extraction and cosine similarity, while SVD collaborative filtering used implicit feedback with a Confidence Alpha (α=40). The final ranking aggregates scores from CBF, SVD, popularity, ratings, distance, and current weather. The computational results demonstrated a consistent HitRate@10 and Recall@10 of 93.0%, proving that this hybrid architecture and dynamic context inclusion effectively generate relevant and personalized recommendations.
Institution: Universitas Muhammadiyah Sukabumi
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7224733741