Operational Policies for Ensuring Top-N Recommendation under Incomplete Nutrient Data: Uncertainty Quantification and Feasibility Ranking in Nutrient-Constrained Recipes

Key Points

• The study aims to address the challenges of recommending recipes with incomplete nutrient data while ensuring compliance with dietary constraints.
• Categorized recommendation candidates into three states: 'satisfied', 'violated', and 'indeterminate'.
• Proposed serving policies to prioritize clearly satisfied candidates followed by uncertain ones.
• Developed an extension to display candidates with a higher likelihood of meeting constraints among the indeterminate options.
• The proposed policy consistently provided a stable number of recommendations despite strong constraints or additional limitations.
• Increased uncertainty was interpretable, enhancing operational transparency in the recommendation process.

Abstract