Abstract: Rockburst, as a complex and catastrophic geological hazard, has attracted great attention in research since it was recorded in the 18th century.Currently, although many scholars have done a lots of research on rockburst prediction and early warning, it is still a worldwide problem.In this paper, the existing methods of rockburst prediction are systematically reviewed.Taking time as the main axis, the development and evolution of different rockburst prediction and early warning methods are analyzed.According to their characteristics, the methods of rockburst prediction are divided into four categories:index criterion method, numerical index method, applied mathematics method and in-situ monitoring method.Then, the advantages, disadvantages and application scope of different types of prediction and early warning methods are summarized and evaluated.At the same time, the relationship and correlation among different types of rockburst prediction and early warning methods are analyzed.Finally, the development trend of rockburst prediction and early warning is analyzed, and some suggestions are put forward for the future research of rockburst prediction.The index criterion methods are divided into single index criterions and comprehensive index criterions.It is highlighted that the single index criterion is the necessary condition for rockburst, including stress condition (stress criterion), rock property condition (energy criterion, brittleness criterion) and rock mass quality condition.The comprehensive index criterion mainly synthesizes the information of in-situ stress (stress criterion), rockburst tendency of rock (energy criterion, brittleness criterion) and surrounding rock geological conditions.The numerical index of rockburst develops from statics to dynamics, from single application of rockburst in mine to multi-purpose and multi-function, from elastic theory which can’t characterize rock mass failure to elastic (brittle) plastic theory which can better reflect the deterioration process of mechanical properties of rock mass after failure.The existing mathematical methods for rockburst prediction can be divided into two sub-categories according to whether they depend on the rockburst cases and the index criteria methods.The in-situ monitoring method uses the monitored data of response of excavated rock mass such as stress, deformation and damage of rock mass to assess the stability of rock mass and to infer rockburst occurrence and intensity.With the development of micro-seismic monitoring technique, artificial intelligence, large data and deep learning, rockburst prediction will be more refined and intelligent.