Research On Noise-tolerant Process Mining Algorithm Based On Hybrid Log

With the booming of Internet and cloud computing,process mining technology plays an important role in information systems.Most of the existing process model mining methods are only applicable to event logs consisting of point events.In practice,many event logs are hybrid,namely,they involve both point events and interval events representing the start and end of an activity.To this end,this paper proposes a novel process model mining approach,which can effectively and efficiently mine high-quality process models as well as time constraints of activities from hybrid event logs.Since real-world event logs inevitably involve noise,we also adapt our approach with heuristic rules to make it noise-tolerant without the need of user input thresholds.The main work of this paper is as follows:(1)A process mining method aim to hybrid event log is proposed.First,the low-level event relations are mined,then the high-level event relations are derived,and finally the process model is generated.This method can not only mine from hybrid event logs containing point events and interval events,but also apply to event logs containing only point events or only interval events.(2)We improve our approach to make it noise-tolerant.Specifically,we devise effective heuristic rules to determine high-level event relations and the final process model.The improved method does not need any user defined thresholds.(3)A method aim to mine the time constraints corresponding to each activity by analyzing event logs based on the high-level event relations is proposed.The algorithm can be directly applied to event logs with or without noise.(4)A prototype tool and a Pro M plug-in were implemented,and we employ the plug-in to perform experiments on both synthetic and real-life event logs,the results of which demonstrate the effectiveness and efficiency of our approach.