An interaction regression model for crop yield prediction | JoVE Visualize

Area of Science:

Agricultural Science
Machine Learning
Data Science

Background:

Global food security relies on accurate crop yield prediction, a complex task influenced by genotype, environment, and management interactions.
Existing methods often struggle to integrate these diverse factors effectively, limiting predictive accuracy and actionable insights for agronomists.

Purpose of the Study:

To develop and validate a novel predictive model for crop yield prediction that integrates optimization, machine learning, and agronomic knowledge.
To achieve high prediction accuracy while simultaneously providing explainable insights into yield-determining factors and their interactions.

Main Methods:

Developed an interaction regression model incorporating optimization, machine learning, and agronomic principles.
Trained the algorithm to select spatially and temporally robust features and interactions, balancing prediction accuracy and generalizability.
Validated the model using corn and soybean yield data from Illinois, Indiana, and Iowa.

Main Results:

The interaction regression model achieved a relative root mean square error of 8% or less for both corn and soybean yield prediction in three US Midwest states, outperforming state-of-the-art machine learning algorithms.
Identified approximately a dozen significant environment-by-management interactions for crop yield, with some aligning with established agronomic knowledge.
Quantitatively dissected crop yield into contributions from weather, soil, management, and their interactions, enabling identification of key influencing factors.

Conclusions:

The developed interaction regression model offers a powerful tool for accurate and explainable crop yield prediction.
The model's ability to dissect yield contributions and identify interactions provides valuable insights for agronomists to optimize crop production.
This approach advances the integration of machine learning and agronomic science for enhanced agricultural decision-making and food security.