Multiobjective Ingredient Optimization for Recipe Planning

Menu planning in food services is a complex process that involves multiple objectives and a large solution space. Its core purpose is to offer an appealing yet diverse selection of dishes every week, whilst maintaining operational efficiency. Here, the underlying problem of ingredient selection is formulated as a multi-objective knapsack problem, considering historical ingredient scores with regard to customer satisfaction and ingredient frequency. We also evaluate an extension of the model that starts to consider the impact of repeat decisions in this problem and attempts to control diversity of ingredient sets over time. The model is implemented on data supplied by an industry partner, and used to derive approximation fronts for both formulations.