Prediction of Crop Yields via Graph Signal Analysis

Predicting crop yield is important for farmers to minimize uncertainty and plan seeding for the next crop cycle. In this project, we address the crop yield prediction problem from a graph signal processing (GSP) perspective. Working with agronomics experts in Growers Edge in Iowa, we first identify relevant features per field, capturing important factors such as precipitation, soil composition, and sunlight condition. We then build a similarity graph to connect fields with similar characteristics as graph nodes based on their corresponding features. Given the constructed graph, we pose the crop yield prediction problem as a graph signal interpolation problem. A PhD student from York University with extensive research experience in graph learning will study this problem as her graduate work and will intern at Growers Edge during the project.

Saghar Bagheri
Faculty Supervisor: 
Gene Cheung
Partner University: