Recent Trends in Machine Learning, Deep Learning, Ensemble Learning, and Explainable Artificial Intelligence Techniques for Evaluating Crop Yields Under Abnormal Climate Conditions

Abstract

Crop yield prediction (CYP) has become increasingly critical in addressing the adverse effects of abnormal climate and enhancing agricultural productivity. This review investigates the application of advanced Artificial Intelligence (AI) techniques including Machine Learning (ML), Deep Learning (DL), Ensemble Learning, and Explainable AI (XAI) to CYP. It also explores the use of remote sensing and imaging technologies, identifies key environmental factors, and analyzes the primary causes of yield reduction. A wide diversity of input features was observed across studies, largely influenced by data availability and specific research goals. Stepwise feature selection was found to be more effective than increasing feature volume in improving model accuracy. Frequently used algorithms include Random Forest (RF) and Support Vector Machines (SVM) for ML, Artificial Neural Networks (ANNs) and Convolutional Neural Networks (CNNs) for DL, as well as stacking-based ensemble methods. Although XAI remains in the early stages of adoption, it shows strong potential for interpreting complex, multi-dimensional CYP models. Hyperspectral imaging (HSI) and multispectral imaging (MSI), often collected via drones, were the most commonly used sensing techniques. Major factors contributing to yield reduction included atmospheric and soil-related conditions under abnormal climate, as well as pest outbreaks, declining soil fertility, and economic constraints. Providing a comprehensive overview of AI-driven CYP frameworks, this review offers insights that support the advancement of precision agriculture and the development of data-informed agricultural policies.