Analytical Methods for Effective Operation of Hunger-Relief Logistics

Rahul Srinivas Sucharitha, Purdue University

Abstract

Food banks and other non-profit organizations play an essential part in alleviating poverty and improving food security in many countries worldwide. These groups help those in need by providing food and resources. Food banks rely on infrequent food and cash contributions to help them achieve their goals. Due to the finite availability of resources and the dynamic structure, managing food supply and demand in a profoundly uncertain situation like this is difficult. This study tackles these issues by presenting and analyzing various statistical and quantitative models to help food insecure people get food in sustainable and meaningful ways. The aim and objective of this chapter is to develop and implement data-driven models and analytical techniques, as well as decision support frameworks, to help food bank administrators better understand the dynamics of food donation supply and demand and to improve the accuracy of food supply and demand behavior prediction at various planning levels to ensure equitable and efficient distribution of food. First, a systematic review was done to research the evolving literature in food bank logistics. A perusal of the literature shows that research in food bank logistics is evolving, and issues about fairness, sustainability, cost reduction, food quality and nutrition, data uncertainty, and food waste study have not been reviewed in great detail. This study attempts to fill this existing gap utilizing a literature review on these issues and outline future research directions based on research gap analysis. Forty-eight published articles were selected, categorized, analyzed, and literature gaps were identified to suggest future research opportunities. The review will provide its usefulness for academicians, researchers, and experts to better understand food bank logistics and guidance for future research. Second, a unique framework of a hybrid model combining ARIMA and neural network autoregressive (NNAR) model to capture both linearity and nonlinearity in the univariate analysis of the food donation supply is proposed. We introduce an iterative cross-validation method called walk-forward cross-validation to the hybrid methodology. Each model’s parameters can be varied and tested again on an iterative basis to obtain optimized tuning parameters specific to the algorithm. The proposed hybrid approach and methodology are applied to the food supply datasets and give better forecasting accuracy than the state-of-the-art. Additionally, the food supply behavior study is further extended for a multi-variate analysis by leveraging statistical and machine learning algorithms to identify the key predictors of the food supply behavior using the same historical food supply data of a regional food bank. Based on the numerical study, Random Forest (RF) method best captures the complex structure of the data compared to the other developed predictive models. Furthermore, we provide a useful framework for implementing these models for their effectiveness in a non-profit organization such as the food-aid distribution system and implementing the proposed framework for several use case studies based on different levels of planning to noteworthy ease and comfort intended for the respective planning team. Thirdly, understanding the dynamics of the demand that food banks get from food insecurity has significant importance in optimizing operational costs and equitable distribution of food, especially when demand is uncertain.

Degree

Ph.D.

Advisors

Lee, Purdue University.

Subject Area

Agronomy|Artificial intelligence|Banking|Economics|Finance|Food Science|Information science|Operations research|Organizational behavior|Public administration

COinS