The fact that task dependencies may change with case attribute values and the existence of invisible tasks makes it difficult to discover a correct process model from event logs.To address these obstacles

a novel process discovery technique based on case clusters and synchronization cores is presented.The task dependencies among case clusters are first mined

and the sub-process model corresponding to case clusters is obtained by synchronization cores.Then

return cores is proposed to model loop structures and the discovery algorithm of return cores from composite cases is provided.Finally

the synchronization cores and the return cores are integrated

and the integrated synchronization cores are transformed to WF-nets to achieve the business process reconstruction.The proposed technique can solve both the mining problem of invisible tasks and the problem that task dependencies change with attribute values.