ACMO-Diet: Accelerated Constrained Multi-Objective Optimization for Diet Recommendation

Diet recommendation systems face significant challenges in balancing nutritional adequacy, taste preferences, and dietary diversity within high-dimensional sparse food spaces. Conventional multi-objective optimization (MOO) algorithms often struggle with inefficient genetic operators that ignore sparsity, leading to wasted computational effort and invalid meal plans. To address these limitations, we propose ACMO-Diet, an accelerated constrained multi-objective optimization framework tailored for diet recommendation. ACMO-Diet integrates the U- NSGA-III algorithm with novel Sparse Mask Crossover (SMX) and Sparse Mask Mutation (SMM) operators, specifically designed to handle ultra-sparse binary representations (e.g., <0.1% density). The framework simultaneously optimizes four objectives: caloric adherence, macronutrient balance, taste satisfaction, and dietary variety. Constrained-domination principles enforce WHO- aligned nutritional constraints while preserving valid meal structures. Experimental validation on AllRecipe.com dataset demonstrates ACMO-Diet’s superiority over seven state-of-the-art MOO algorithms. It achieves the highest hypervolume and lowest generational distance, indicating superior convergence and Pareto front coverage. The sparse operators reduce runtime by 18.5% (33.3s vs. 51.2s for SBX+PM) while maintaining 100% constraint compliance. ACMO-Diet establishes a new paradigm for precision nutrition, bridging clinical guidelines with practical dietary customization through efficient exploration of high-dimensional sparse food spaces.