Crop Yield Forecasting Using Machine Learning and Deep Learning approaches: A comprehensive review

Abstract

The ability to predict crop yields in precision farming is essential, as it enables the use of data-driven decision-making that can help optimise resource utilisation, improve food security, and facilitate sustainable farming. The advancements in machine learning and deep learning in recent years have made significant strides in predictive modelling, enabling the estimation of yield in a more precise and scalable manner. The paper shows an extensive overview of 20 recent research studies that examine the application of ML and DL to predict crop yield. The review is a systematic analysis of the methods used, traditional ML models, including Random Forest, XGBoost, Support Vector Machine, and Artificial Neural Networks, as well as more sophisticated DL-based architectures, including Convolutional Neural Networks, Long Short-Term Memory networks, and Graph Neural Networks. It also explores the most significant properties that predict primary outcomes, including soil factors (NPK values, pH), climate indicators (temperature, rain, and humidity), and the use of remotely sensed imagery (user plane surveillance, satellite-aided observation). The paper also assesses data, performance indicators (MSE, RMSE, R²), and algorithms used to test the usefulness of the models. The given review provides a comparison of the performance of ML/DL methods and offers insights into better forecasting crop yields, as well as research to continue exploring the under-explored field of AI-driven sustainable agriculture.