Advances In Crop Yield Prediction Through Machine Learning And Deep Learning: Insights Into Yield Estimation

Abstract

Crop yield prediction methods are becoming increasingly accurate and scalable because of the growing demand for sustainable agriculture. We identify various Machine Learning (ML) and Deep Learning (DL) techniques in yield prediction through a systematic review, which significantly contributes to state-of-the-art research for modern agriculture. It identifies and examines essential data sources such as remote sensing images, climate variables or meteorological information, soil health indicators, and sensor metrics that are key to improving prediction. In particular, it explores challenges with these methodologies relating to varying data quality, model size scalability, and a lack of interpretability in complex DL models. The future trajectory of these trends is then discussed, highlighting the increased utilization of hybrid models that capture the strengths and informational value of multiple techniques alongside further integration of multimodal learning approaches to take advantage of complementary data sources. These developments show promise to enhance the accuracy of yield predictions while overcoming regional as well as environmental discrepancies. The paper ends with some recommendations for future work, including the need to model the region; the use of explainable AI (XAI) techniques as integration between common and domain/ML approaches; and scalable frameworks to support real-time agricultural prediction. The present study synthesizes recent achievements and points to key challenges that can foster AI-driven agricultural practices, sustainability of agriculture, and meeting global food security demands.