Cancer has a high incidence and mortality and is a major threat to human health. Cancer prognosis analysis can effectively avoid excessive treatment and waste of medical resources and provide a scientific basis for medical staff and their families to make medical decisions. It has become a necessary condition for cancer research. With the rapid development of artificial intelligence technology and medical informatization in recent years, smart medicine has received widespread attention. It has become possible to analyze the prognosis of cancer patients automatically. As an essential part of smart medicine, cancer patients need to conduct effective intelligent prognostic analysis. This article reviews the existing machine learning-based cancer prognosis methods. Firstly, it provides an overview of machine learning and cancer prognosis, introduces cancer prognosis and related machine learning methods, and analyzes the application of machine learning in cancer prognosis. Then, it summarizes cancer prognosis methods based on machine learning, including cancer susceptibility prediction, cancer recurrence prediction, and cancer survival prediction, and sorts out their research status, cancer types and data sets involved, and machine learning methods used and prediction performance. Finally, the cancer prognosis methods are summarized and prospected, and the aspects that should be explored and improved are proposed: (1) include other high-fatal cancers in the prognostic analysis; (2) comprehensive analysis of cancer expression data and image data to improve prognostic performance; (3) optimize the prognostic model to improve prognostic performance.