Heuristic Process Log Repair Technology

In the era of big data,business process supported by the information system produce a mass of event logs.Those event logs play an increasingly important role in enterprise.Based on those event logs,the enterprise can carry out processing application and analysis,such as process mining,provenance analysis and complex event processing,to extract valuable information.However,in the open,dynamic and changeable internet environment,scenarios such as system failures and hybrid storage of executions of different processes can inevitably introduce the "noise" in event sequences in the event logs.In these nonconforming event sequences,there exists missing events,redundant events,dislocated events and their mixtures.The results of applications and the analysis which based on those low quality event logs are not credible.Therefore,recovering these low quality event log and improving the quality of log receives more attention.Recovering an nonconforming event sequence based on a prior business process is the current mainstream technology,but it has proven to be NP-hard.The existing recovery method either focus on event logs with missing events or enumerating all equivalent recoveries and lack of scalability.Thus,in this paper,we proposes a heuristic recovery method which can handle all nonconforming situations.We carry out a series of experiments to demonstrate the validity and scalability of our method.The main works is as follows:1.Develop a general heuristic event log recovery approach based on the alignment between event logs and process models which can handle event log with missing,redundant and dislocated events.It can not only be used to seek the alignment between the log and the model,but also provide diagnostic information for the deviations of the process instance from the process definition during alignment.2.Utilize the behavior and structural characteristics of business process combined with some heuristic rules to reduce the searching space for seeking the best alignment reference trace.More specifically,for business process with no alternative routings or iterative routings,we present a trace replaying to seek the reference trace.For this situation,we have proved that can guarantee the minimum repair properly;for business process involves alternative routings,process decomposition algorithm is used to decompose business process into unrelated sub-processes,then we use heuristics to choose the reference sub-process,at last the trace replaying is used to get the reference trace;For business process involve iterative routings,we firstly use heuristics to determine the right occurrence number of iterative routings,and then utilize trace replaying and the right occurrence number of iterative routings to get the reference trace.3.Design and implement a Prototype tool Effa to support the theoretical content and methodology of the heuristic log recovery algorithm.4.Utilize a number of real business processes and synthetic event logs,and conduct a series of experiments compared with the de facto the state-of-the-art recovery method.Verify the effectiveness and time efficiency to illustrates the advantages of our method.