Knowledge Graph Integration to Hybrid Recommender Architecture System Based on Recommended Dietary Allowances

Abstract

In Indonesia, making daily nutritional choices is still a big challenge. This is because there are various choice of local foods and no special tools to help people choose the right diet. This research propose a recommendation system that combines Collaborative Filtering (CF) and Content-Based Filtering (CBF) with a Knowledge Graph (KG) integration. The TKPI and USDA datasets are used for food composition, along with the NHANES user profiles dataset, with the Recommended Dietary Allowances mandated by the Ministry of Health (Regulation No. 28 of 2019) used as the nutritional adequacy target. In the proposed system, KG functions as a semantic reasoning layer that enforces strict dietary boundaries while preserving individual user prefere. The results of the evaluation show that adding the KG to the CF-CBF hybrid makes the average calorie deviation much lower, at 38.0 kcal per recommendation. This research shows that adding an ontology-based layer successfully balances personal cravings and nutritional compliance, providing a useful solution for those looking for health-conscious dietary advice.