Application of Tree-Based and Gaussian Process Models in Predicting Wheat Yields

The article investigates the use of machine learning (ML) models for predicting the yield of wheat crop with special reference to seed priming techniques, such as hydropriming, and the environmental factors including soils, climate and water regimes. With rapid population growth, wheat as a staple food in most countries is in very high demand; therefore, improving the crop yield is necessary. The paper highlights boundaries of conventional regression models that are used in agricultural data, and how agri-ML models like Random Forest, Support Vector Machines (SVM) and Gaussian Process Regression (GPR) performs better in accuracy and in forecasting. Data were collected from various wheat traits, including the length of spikes, number of fertile spikes, grains per spike, and overall grain yield, using different seed priming treatments. The study shows that tree-based models and GPR deliver the best performance in yield prediction, with the highest R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> and lowest RMSE values. Neural networks, however, underperform due to the dataset's limited size, reinforcing the need for larger datasets for training deep learning models. The findings underscore the potential of ML to revolutionize crop management, with future research needed to refine these models and incorporate IoT-enabled real-time data.