GMM Clustering for In-depth Food Accessibility Pattern Exploration and Prediction Model of Food Demand Behavior

TL;DR

This paper introduces a GMM clustering-based framework to analyze food insecurity patterns, identify causes, and improve food demand prediction, aiding better inventory and redistribution strategies in food assistance networks.

Contribution

The novel application of GMM clustering to identify food insecurity causes and enhance demand prediction models in food assistance networks.

Findings

GMM clustering effectively identifies food insecurity patterns.

The framework improves prediction accuracy of food demand.

Practical case studies demonstrate ease of implementation.

Abstract

Understanding the dynamics of food banks' demand from food insecurity is essential in optimizing operational costs and equitable distribution of food, especially when demand is uncertain. Hence, Gaussian Mixture Model (GMM) clustering is selected to extract patterns. The novelty is that GMM clustering is applied to identify the possible causes of food insecurity in a given region, understanding the characteristics and structure of the food assistance network in a particular region, and the clustering result is further utilized to explore the patterns of uncertain food demand behavior and its significant importance in inventory management and redistribution of surplus food thereby developing a two-stage hybrid food demand estimation model. Data obtained from a food bank network in Cleveland, Ohio, is used, and the clusters developed are studied and visualized. The results reveal that this proposed framework can make an in-depth identification of food accessibility and assistance patterns and provides better prediction accuracies of the leveraged statistical and machine learning algorithms by utilizing the GMM clustering results. Also, implementing the proposed framework for case studies based on different levels of planning led to practical results with remarkable ease and comfort intended for the respective planning team.