Mixed Integer Goal Programming Proposed for Personalized Meal Optimization Addressing Serving Granularity and Feasibility

Traditional diet optimization models in operations research often struggle with two major practical hurdles. Most current formulations use continuous variables that result in unrealistic fractional servings, such as 1.7 eggs, and rely on hard nutrient constraints that frequently cause system failures when multiple nutritional targets conflict with one another. This new research addresses these issues by combining integer programming with goal programming to ensure that meal plans are both implementable and mathematically feasible. The proposed Mixed Integer Goal Programming model utilizes integer variables to represent natural serving units like whole eggs or tablespoons while employing goal programming to handle soft nutritional targets. This approach incorporates inverse-target normalization to balance multi-nutrient optimization across different scales. A key discovery in this study is the deviation absorption property, which shows that goal programming deviation variables can buffer the mathematical costs of requiring integer servings, making the results more efficient than standard integer programming. Evaluation across 810 USDA food instances revealed that this method consistently finds better solutions than traditional goal programming followed by manual rounding. It maintained 100% feasibility in scenarios where hard-constraint models failed, demonstrating that the integer solution matched the continuous optimum in every benchmark for meals involving fifteen or more food items. For software engineers developing health and nutrition applications, this framework provides a robust alternative to heuristic-based meal generation and manual post-processing.