Specific learning disorder (SLD) is a common neurodevelopmental disorder in children that significantly affects academic performance and quality of life. At present, diagnosis mainly relies on standardized tests and professional evaluations, a process that is complex and time-consuming. Multiple studies have shown that machine learning can analyze diverse data, including test scores, handwriting samples, eye movement data, neuroimaging data, and genetic data, to automatically learn the relationships between input features and output labels and achieve efficient prediction. It shows great potential for early screening, auxiliary diagnosis, and research on underlying mechanisms in SLD. This article reviews the applications of machine learning in the auxiliary diagnosis of SLD and discusses its performance when handling different data types.