A Systematic Review and Meta-Analysis of Machine Learning and Deep Learning Approaches for Crop Yield Prediction: Toward Scalable and Hybrid Modeling Frameworks

The accurate prediction of crop yields is critical for the management of natural resources sustainably, food security, and agricultural planning. The advancements in machine learning and deep learning have significantly enhanced the capabilities of crop yield forecasting. However, the field is fragmented due to the variety of methodologies employed in the existing literature. In this article, a comprehensive review and meta-analysis of deep learning and machine learning techniques for the prediction of crop yield are presented. Through searching important scientific databases for pertinent research in the area, investigations into the different aspects of the studies were performed. The outcomes of the research reveal that deep learning and hybrid models outperform conventional machine learning models in terms of the accuracy of their predictions of crop yields. However, the complex nature and the reliance upon remote sensing data of most of the high-performing machine learning and deep learning models limit their scalability. Furthermore, there is a lack of research concerning the integration of ensemble stacking and multi-task learning approaches within the context of agricultural datasets. The findings of this investigation highlight the need for scalable, machine learning models with high accuracy and reliability in the field of crop yield prediction. Furthermore, future research aiming to develop integrated methods that combine ensemble learning and deep learning models for improved prediction of agricultural production will build upon the findings of this investigation.