Predicting Sweetpotato Parent Compatibility Using Machine Learning

Sweetpotato breeding programs rely on resource-intensive, trial-and-error crossing with high failure rates to create new varieties. To address this, we propose a novel machine learning approach using Graph Neural Networks (GNN) and ensemble methods to predict parental compatibility. This predictive tool aims to optimize breeding efficiency, reducing costs and accelerating the development of improved sweetpotato varieties for both farmers and consumers.