Abstract: Current workflow mining algorithm using local strategy couldn't ensure that a globally optimal process mode　was mined. The algorithm was also sensitive to noise. To solve the problems,a hybrid adaptive genetic algorithm (HA-GA) was proposed. Firstly, Elementary Workflow net (EW-net) was defined. The enabling and firing rules of EW-net were given, and the process model was described. Secondly, a converting algorithm proposed was used to convert the process model to EW-net, and an evaluating function of the individual fitness was presented in order to measure the　compliance between event log and mined process model. Lastly, hybrid adaptive crossover and mutation rates were do　signed according to evolution stage and parents' similarity. I}he simulation testing results demonstrate that the new algorithm has noise immunity and is more robust than a algorithm,and it can find better solution and converge faster thar the simple genetic algorithm (SGA) employing general genetic strategy.

Key words: Workflow mining, Process mining, Hybrid adaptive genetic algorithm, EW-net, Causal matrix