In recent years, the use of surface-enhanced Raman spectroscopy (SERS) technology to detect biological samples has become a hot topic. Significantly, the application of SERS technology combined with machine learning methods in clinical sample diagnosis has become increasingly mature. Machine learning methods based on unsupervised and supervised algorithms to solve complex samples and large, high-dimensional data, have received high attention. This review describes the relevant applications of SERS technology combined with machine learning methods, especially in the biomedical field. SERS can detect fingerprint information of biological samples using label-free strategies, or indirect SERS detections for tracking biomarkers such as proteins. This review summarizes SERS technology combined with machine learning for disease diagnosis in clinical samples such as blood, urine, and biotissues. In addition, we also summarize their applications on many cellular samples and other complex samples. An overview of the latest advances in this field is provided and this study offers a reference that can be followed by researchers working in SERS bioanalysis.