Integrating meteorological and breeding data to predict maize yields using machine learning algorithms.

Area of Science:

• Agricultural Science
• Genetics
• Data Science

Background:

• Accurate crop yield prediction is crucial for global food security, especially with climate change impacts.
• Deep learning models for yield prediction often require extensive data and computational power.

Purpose of the Study:

• To develop a machine learning model for predicting maize hybrid yields.
• To integrate meteorological data with genetic information (breeding values) for improved prediction accuracy.

Main Methods:

• Evaluated four machine learning algorithms: Random Forest (RF), XGBoost, Support Vector Regression (SVR), and Gaussian Process Regression (GPR).
• Optimized models using hyperparameter tuning.
• Integrated meteorological data with Best Linear Unbiased Prediction (BLUP) estimated breeding values.

Main Results:

• The Random Forest (RF) algorithm demonstrated superior performance.
• RF achieved a coefficient of determination (R²) of 0.64.
• Key performance metrics included RMSE of 1010.59 kg/ha and MAPE of 8.3%.

Conclusions:

• The RF-based model provides accurate maize yield predictions for specific cultivars and environments.
• The framework supports farmers in selecting adapted hybrids and aids breeders in identifying high-yielding varieties.
• This promotes efficient breeding strategies and precise cultivation recommendations for enhanced agricultural productivity.