Process Mining Technology Based On Dependency Analysis

At present,most business processes are implemented through information systems.Compared with the method of artificial process,process mining generates the process model based on the event log in information system,and the generated process model is objective.Therefore,process mining plays a very important role in the realization of business process optimization and intelligent management.One of the important aspects of process model quality is the consistency between event logs and the models.The process models mined by traditional mining algorithms are not able to maintain completely consistent with the log,when the information in logs is limited.Therefore,it becomes a hot topic that proposing a process mining method when the information in logs is limited.State-of-the-art approaches employ activity orders in traces to complete process discovery and they require higher completeness notions of event logs.Thus,they may fail to extract appropriate processes when event logs cannot meet the completeness criteria.To cope with this problem,we propose a novel technique which leverages activity dependencies(including control dependencies,data dependencies)in traces in this thesis.Based on the fact that activities with no dependencies can be executed in parallel,our technique is able to discover processes with concurrencies even if the logs fail to meet the completeness criteria of existing approaches.That is,our technique calls for a weaker notion of completeness.We evaluate our technique through experiments on both real-world and synthetic event logs,and the conformance checking results prove the effectiveness and advantages of our technique compared with state-of-the-art approaches.The main work of this thesis is as follows:1.A process mining method based on event dependency in event logs is proposed.Our event logs are not only collections of event sequences,and they also requires input and output information in the event.By parsing the event log,the log trace information is transformed into dynamic dependence graph(DDG)based on the analysis of control flow and data flow.Then the DDG is converted to process model in Petri net format through a series of transformation rules.2.The concept of dependence completeness is put forward,and the method of calculating the dependence completeness of the log is provided without the process model.Finally,the experimental results prove that our method has a weaker requirement for log completeness.3.A prototype of process mining tool called ProM-D is designed and implemented.The input of the tool is an event log in XES format,and the output is a process model presented by Petri net.The Petri net file is stored with PNML language,which can be displayed visually on the tool graphically.4.We design experiments on real logs and synthetic logs,comparing our method with several existing state-of-the-art methods.By analyzing the conformance checking results of a series of conformance metrics,we can verify the feasibility and effectiveness of our technique.