Metaheuristic Approaches for Estimating In-Kind Food Donations Availability and Scheduling Food Bank Vehicles

Food banks provide services that allow households facing food insecurity to receive nutritious food items. Food banks, however, experience operational challenges as a result of constrained and uncertain supply and complex routing challenges. The goal of this research is to explore opportunities to enhance food bank operations through metaheuristic forecasting and scheduling practices. Knowledge discovery methods and supervised machine learning are used to forecast food availability at supermarkets. In particular, a quasi-greedy algorithm which selects multi-layer perceptron models to represent food availability is introduced. In addition, a new classification of the vehicle routing problem is proposed to manage the distribution and collection of food items. In particular, variants of the periodic vehicle routing problem backhauls are introduced. In addition to discussing model formulations for the routing problems, a hybrid genetic algorithm is introduced which finds good solutions for larger problem instances in a reasonable computation time.