Workflow technology is widely used in business process management. However, there are still many problems during the execution of business process because of the imperfect workflow model. Process mining is the most useful tool of workflow modeling, which can obtain objective and valuable information from event logs and build process model. Nevertheless, the existing process mining algorithms still have some problems, such as low accuracy, long operation time and overfitting, which will decreace the accuracy of the workflow model. This paper proposed a new process mining algorithm based on statistical α-algorithm, which can not only ensure the accuracy and suitable fitness, but also decrease the operation time. First, cognominal activity identification rules were proposed to be the pre-treated process of process mining, which could improve the accuracy of algorithm. Second, statistical α-algorithm was proposed as the core algorithm of process mining to eliminate the influence of noise in event logs. Moreover, a new algorithm was proposed to identify non-free-choice constructs, which improved the robustness and accuracy of the algorithm. The accuracy and efficiency of the algorithm are verified by simulation and real case.