An improved simulated annealing algorithm for process mining

The target of process mining is to automatically extract process models from event logs related to actual business process executions. One of the most important fields concerned is control flow mining, i.e., ordering of activities. However, the presence of complicated constructs, such as duplicate tasks, invisible tasks and non-free-choice structures, hinders us from correctly discovering the relations between activities. Therefore, an improved simulated annealing approach is proposed in this paper to tackle these problems. To verify the performance, experiments are conducted in the process minig framework. The result is expressed in terms of Petri net.