Abstract: Providing different services for different ticket vending machine(TVM) users according to their features is an effective way to increase the TVMs' ticketing efficiency. Therefore, TVMs should be installed at suitable location according to their type and be configured with suitable human machine interface(HMI). Based on the Internet user behavior analysis technologies, some new user feature analysis methods were proposed for railway auto-ticketing system, including KMeans cluster analysis and TopN analysis. By the applications of these analysis technologies, the TVM users' cluster features and the TVM operation modes could be drawn out to improve the TVMs' location layout and the TVMs' HMI configuration.