Abstract: Source identification of water inrush in tunnel is quite important to water flow predication and prevention and cure on groundwater inundation. In order to offer scientific reference to the departments of construction and management, Support Vector Machines is introduced to build a model to identify source of groundwater inrush in tunnel, based on the chemical information of groundwater inrush in tunnel and its possible sources. Application of this model to the Tongluoshan tunnel in Dianjiang to Linshui expressway reveals that the karst waterbearing system in the Jialingjiang formation and Leikoupo formation is the major supplier rather than the waterbearing system in the factured Xujiahe formation, due to the difference of the water abundance. This conclusion corresponds to the phenomena reflected by dynamic monitoring of surface water, wells and springs, as well as water flow from mines, indicating that SVM is valuable to source identification of water inrush in tunnel.