DeepFarm: AI-Driven Management of Farm Production using Explainable Causality
Yingjie Wang, Jaganmohan Chandrasekaran, Flora Haberkorn, Yan Dong, Munisamy Gopinath, Feras A. Batarseh · 2022
American agriculture has been afflicted by numerous outlier events in the past decade, such as several natural disasters, cyber-attacks, trade wars, and a global pandemic. Such unprecedented black-swans have created outcome uncertainties throughout the food supply chain, starting at the farm level for agricultural producers and aggregating at the consumption level for households and international trade flows. The primary drivers behind the shocks in agricultural productivity include strong weather-related events, transitory transportation disruptions, shipping delays, and policy shifts. This paper presents DeepFarm, an Artificial Intelligence (AI)-enabled framework to measure and manage uncertainties while evaluating multiple cause-effect scenarios in agricultural farm production. We deploy Deep Learning (DL) models to predict the impact of crop yield during outlier events such as extreme weather events and cyber-attacks. Additionally, we use a causal inference-based approach to quantity the impact of such events affecting the critical phases of farm production. Models are developed; experiments are performed; the results are recorded, evaluated, and discussed. Our results suggest that DeepFarm can effectively forecast and quantity the impact of outlier events on crop yield across different regions in the US.