An Approach to Approximating Contributions Received From Supermarkets by Food Banks

摘要

Because supermarkets do not share how much food is available for collection, food banks are limited in how confidently they can assess their ability to provide an uninterrupted supply of different commodities to the agencies that they serve. This research addresses this limitation by proposing a feed forward artificial neural network to approximate the volume received in isolated collections at supermarkets. A methodology is discussed which identifies the most appropriate neural network configuration to evaluate past collections from a set of supermarkets within the Food Bank of Central and Eastern North Carolina (FBCENC) service area. The selected configuration is then compared with multiple linear regression in terms of its ability to approximate future donation volumes.