<p>The powerful data mining capabilities of machine learning (ML) facilitate the capture of complex and dynamic correlations among parameters and the underlying patterns within the drying process, particularly excelling in addressing nonlinear relationships. As a result, ML models are increasingly applied in agricultural product drying, particularly for quality prediction, process optimization, and control, and have become a key focal of recent research. This review provides a comprehensive overview of current research on ML models in agricultural product drying. The essential stages in developing ML-based models are introduced, encompassing data collection during drying, data pre-processing, model tuning, and model evaluation. Furthermore, the review examines the principles and specialized optimization strategies for representative supervised learning algorithms, with a focus on ML models with time series features such as recurrent neural networks and long short-term memory networks. This paper synthesizes advancements in the applications of ML models over the past 8 years, focusing on the prediction, optimization, and control in drying dynamics, physical transformations, chemical changes, and energy consumption in agricultural product drying. Finally, the challenges of current technological development are discussed, along with potential research approaches to address these issues.</p> Graphical Abstract <p></p>