Maintenance hemodialysis (MHD) is the primary treatment for end-stage renal disease (ESRD). MHD patients often experience multiple complications with a higher mortality rate. Clinically, there is a need for enhanced prediction of MHD-related complication risks and comprehensive management of hemodialysis. Machine learning (ML), leveraging its data mining and model-building capabilities, has gradually been explored for application in MHD field. This review focuses on the application of ML models in MHD, highlighting the current research and practical use of these models in patient monitoring, complication prediction, and comprehensive management. It aims to provide a reference for further promoting and optimizing such models in clinical practice of MHD.