Diabetic kidney disease (DKD), the most common and severe complication of diabetes, represents a leading global cause of chronic kidney disease (CKD) and the foremost etiology of end-stage renal disease (ESRD). This condition imposes substantial disease burdens on patients, families, and healthcare systems worldwide. The development of risk prediction models to identify high-risk individuals for DKD at an early stage, coupled with personalized diagnostic and therapeutic strategies, holds critical importance for reducing the incidence and progression of DKD. Over the past several decades, numerous scholars have constructed predictive models for early DKD occurrence. With the advancement of big data and artificial intelligence, building predictive models through machine learning and deep learning algorithms has become a research hotspot. This article provided an overview of the methods for constructing risk models for the early occurrence of DKD, compared the advantages and disadvantages of traditional prediction methods, machine learning, and deep learning algorithms, and summarized the current research status of risk prediction models for the early occurrence of DKD, aiming to provide a reference for the construction of prediction models and early clinical intervention of DKD.