The invention provides an automatic operation method based on user behavior analysis. The method includes the user classification processing stage and the application layer calling stage, wherein in the user classification processing stage, after being classified and processed, user data is stored in a database for an application layer to call; in the application layer calling stage, operation strategies are automatically executed by means of operation data, business indexes and current period KPI; a user state transfer matrix is calculated in the application layer calling stage, the Markov process theory serves as the basis, and transaction frequency, number and trade volume in the future are predicted; in the application layer calling stage, user multi-dimensional information is used foruser portrait analysis, and a user group report is obtained. The method overcomes the defects that common operation means cannot focus on individuals and cannot conduct personalized operation recommendation according to individual habits, operation means are refined to each user individual, more appropriate operation means, such as excitation and guidance, are provided, and market application prospects are good.