An Anti Noise Process Mining Algorithm Based On Information Entropy And Trace Clustering

With the development of industrial intelligence,automatic production is becoming more and more imminent.Many enterprises begin to use business process management system in production.With the operation of business process management system,more and more business process logs and other data have been accumulated.How to effectively utilize and analyze these event logs is a problem that people need to solve urgently.Process mining technology is a branch of business process management.It can extract process knowledge from event logs and build process models,which helps to detect and improve business processes.Through extensive investigation and research,we have learned the current research status of process mining at home and abroad,combed its development process,and conscientiously studied the representative process mining algorithm.We find that most existing process mining algorithms are inadequate in dealing with log noise.The traditional process mining algorithm ignores the impact of noise,which is unrealistic in real-life logs.Most of the process mining algorithms that can handle noise also have a lack of reasonable denoising thresholds.In order to solve this problem,this paper presents a method based on maximum entropy principle to determine the threshold,which is used to remove the influence of noise in event log on process mining results.However,the above method based on information entropy denoising is only suitable for dense logs.Therefore,a denoising method based on trace clustering is proposed to identify noise traces in sparse logs.In addition,based on the improved frequency matrix,the loop structure,selective structure,parallel structure and non-free choice structure are identified according to the activity frequency relationship in logs.Finally,an algorithm framework for adaptively removing the noise and constructing process model is given.This paper presents a new attempt to use noise entropy to remove noise in the field of process mining.The obtained denoising threshold does not need to set the parameters in advance,but can be determined according to the event log.Therefore,the given denoising algorithm has a certain degree of universality.Experimental results show that the algorithm proposed in this paper has a higher accuracy,recall,fitness and behavioral appropriateness on noisy logs.