Healthy Food Recommendations for User Communities through Trust and Popularity-Aware Aggregation

In recent years, food recommendation systems (FRS) have gained importance in inspiring healthy eating and supporting dietary decision-making. However, traditional methods focus on distinct users and tend to rank taste over nutrition, leading to unhealthy food suggestions. To overcome this, a Healthy Group Food Recommendation (HGFR) framework is proposed to balance health factors with cooperative user preferences. Initially, user–food interactions, ratings, and nutritional details were gathered from the Food.com dataset, which contains more than 178,000 recipes with comprehensive nutritional information and user reviews. Then, a graph-based model was created, where users were signified as nodes, and their similarities were computed using a time-aware rating mechanism. Next, a deep community detection method was employed for grouping similar users, and a trust network and popularity calculation confirmed that influential members had appropriate weights in group decisions. Furthermore, a health-aware prediction module compared nutrition values against World Health Organization (WHO) ranges, which produces normalized health scores that were combined with preference predictions. Finally, group aggregation was executed using a popularity-weighted average to recommend the top-k healthy foods. Experimental evaluation established that HGFR attained greater results in terms of Group Satisfaction Metric (GSM) with 93.0% when compared with the existing Group Preference Aggregation (GPA) method.