Forecast and analysis of food donations using support vector regression

Food banks collect and distribute food donations to partnering agencies to help fight food insecurity. Donations come from several retail, manufacturing, and community donors and can vary significantly over time with respect to frequency, amount, and quality. Support Vector Regression has shown to have tremendous advantages over other forecasting methods but has not been previously applied to the food donation supply problem. Using data obtained from a local food bank, this study describes the following: (i) the prediction accuracy for food donations when using support vector regression; and (ii) the relationship between prediction accuracy, historical data length, and food donation data variability. Results from our study show that support vector regression has comparable results to other food donation forecasting techniques previously explored in the literature.