Maize yield prediction using machine learning: a systematic literature review.

Area of Science:

• Agronomy
• Data Science
• Agricultural Economics

Background:

• Maize is a staple crop critical for food security, especially in sub-Saharan Africa.
• Accurate yield prediction is vital for effective planning and resource management.
• Machine learning (ML) offers potential for improving maize yield estimation.

Purpose of the Study:

• To systematically review the application of ML techniques in maize yield estimation.
• To analyze methodologies, predictor variables, and outcomes in peer-reviewed studies.
• To identify trends and challenges in ML-based maize yield prediction.

Main Methods:

• Systematic literature review following PRISMA 2021 guidelines.
• Synthesis of 81 peer-reviewed studies (2014-2025).
• Analysis of ML algorithms, predictor variables, and evaluation metrics.

Main Results:

• A surge in ML publications for agronomic decision-support post-2021.
• Random Forest (49.4%) and XGBoost (16.1%) were dominant ML algorithms.
• Hybrid deep learning demonstrated superior performance; environmental, remote-sensing, and soil data were key predictors.

Conclusions:

• ML shows increasing utility in maize yield prediction, with hybrid deep learning outperforming others.
• Challenges include data scarcity, limited interpretability, and geographical bias, particularly from Africa.
• Future work requires open-access data, explainable AI, and capacity building in computational agronomy.