<![CDATA[The rapid growth of the digital culinary industry increases the need for intelligent menu recommendation systems that can assist customers in making accurate and personalized choices. This study develops a hybrid food recommendation system that integrates three complementary approaches: popularity-based ranking, Term Frequency–Inverse Document Frequency (TF-IDF) with K-Nearest Neighbors (KNN) item similarity, and tag-based cosine matching. The system also incorporates a Content-Based Filtering component that leverages cosine similarity to strengthen similarity modeling across textual and tag-based representations. A total of 77,157 real transaction records from SR Cipali Restaurant, collected between April and December 2024, were used as the primary data source for system development and evaluation. Data preprocessing includes cleaning, category filtering, TF-IDF transformation for product names, One-Hot Encoding for tags, and price normalization to generate structured and comparable feature representations. Experimental results show that the TF-IDF KNN model achieves the best performance with an accuracy of 0.94, recall of 1.00, and F1-score of 0.89. The popularity-based model reaches an accuracy of 0.89 with balanced precision and recall of 0.80, while the tag-based model obtains a precision of 1.00 but lower recall due to tag inconsistency and ranking selectivity. The novelty of this study lies in the use of a hybrid lightweight framework evaluated on real-world restaurant transactions, which is rarely explored in previous research dominated by benchmark datasets. The proposed system demonstrates strong practicality for small and medium-sized restaurants that lack rating data and can be further improved by enhancing tag quality and incorporating more product attributes.]]>
Copyrights © 2026
