The primary objective of our project is to build a ML model to predict the food trade links and weights. Traditionally, to build this kind of model, we use linear programming method with the objective to minimize the trade cost and constraint by the food production and demand of each study area. We believe that ML is another promising tool. The trade behavior will be impacted by other characteristics of trade area for example: GDP,  food production, food consumption, transportation data employments, and other demography data and US had tons of public dataset that we can use. The core dataset of this project will be the CFS and FAF dataset which include several years trade data between different CFS area pairs. We need to do an EDA later to figure out the features we will use to predict the flow. The loss function will be the sum of difference between predict result and the existing food flow weight.  Building on this model, we want to employ some scenario analysis like how infrastructure change, climate change and infectious disease could impact the food supply chain.
A good prediction of food flow will help in informing how food shocks from extreme conditions like climate change, infrastructure change and infectious disease COVID-19 can impact various regions in the US.