Process mining can extract knowledge from recorded event logs through modem information systems,which provides a new method for process discovery,monitoring and improvement in various application fields.Nowadays,more and more enterprises use business process management system to support their business process.As a key technology of workflow redesign and analysis,process mining technology can automatically reconstruct the process model,calculate the matching degree between the log records in the event log and the process model,and repair and extend the process model so that it can better reflect the reality of the production process.The starting point of process mining is event log,and the handling of event log plays an important role in the process mining.Real life event logs can contain chaotic activities,and the existence of chaos activities seriously affect the quality of the process model.The current modus operandi for filtering activities from event logs is to simply filter out infrequent activities.We show that frequency-based filtering of activities does not solve the problems that are caused by chaotic activities.In addition,the existing model repair method cannot effectively solve the deviation between the event log and the model for the discovered process model,and the repair process is complicated.For the above two problems,this paper proposes a process discovery method based on chaos activities filtering and model repair method based on deviation reduction.The main contributions of this article are as follows:(1)Process mining technology automatically discovers business processes from the execution data of business processes.The real life business process event data logs usually contain chaotic activities which makes the traditional event data log filtering method not able to effectively filter the chaotic activities in event data logs.This paper proposes a novel chaotic activities filtering method based on bidirectional causal dependence.The method achieves the filtering of chaotic activities in event data logs by analyzing the bidirectional causal dependence between the model and event data logs and taking the precision as a constraint.At the end of this paper,the proposed method is used in the Tianyuan big data platform to verify the effectiveness.By comparison experiments of chaotic activity filtering method based on information entropy is evaluated from the aspects of time complexity.The evaluation shows that the method can discover more precise process models through the analysis of precision between multiple sets of event data logs and the process models generated before and after chaotic activities filtering.(2)Big data transaction is a typical complex process model,which makes the traditional model repair method not able to effectively discover and reduce the deviation between process execution and process rules.This paper proposes an approach of repairing big data transaction model based on deviation reduction.With the help of the reachable marking graph,the approach discovers the deviation between the event log and the process model found,reduces the deviation between the event log and the model,and gets the repaired model based on the effective deviation.At the end of this paper,the proposed approach is used in the Tianyuan big data platform to verify the effectiveness.By comparison experiments of those repair methods based on model alignment and the iteration of the effect of repairing is evaluated from the aspects of fitness,precision,simplicity and time complexity.The evaluation shows that the proposed approach has an advantage over existing methods. |