Predictive Modeling of Food Choices for Client Choice Food Pantries

摘要

Greensboro Urban Ministry (GUM) distributes approximately 1.3M lbs. of food annually to an average of 150 households per day through its food pantry. Traditional food pantries like GUM have several disadvantages including overstocking of certain foods, increased food waste, and negative psychological effects on the clients. As a result, management wishes to convert to a client choice food pantry. Client choice food pantries, like grocery stores, allow clients to choose which foods they will take home, reducing food waste, and providing a psychological boost to the clients. This study will (1) determine client food preferences, (2) identify the gap between client food preferences and food donations, and (3) propose an approach to close the gap between client food preferences and food donations. Logistic regression is used to develop a predictive model of client food preferences using the developed results of a survey. Independent variables include such predictors as SNAP recipient and size of the household. Dependent variables are the various foods available at the food pantry. Results show a significant difference exists between client food preferences and food donations. Results also indicate that logistic regression shows promise in the development of a predictive model of client food preferences.