Anomaly Detection For Cross-Organizational Business Process Model Based On Entropy

With the advent of the Internet era and the development of economic globalization,the degree of enterprise informatization is also getting higher,there are more and more users ' requirements must be met through multi-enterprise and multi-organization cooperation,so that cross-organizational business process management came into being.Cross-organizational business processes typically involve multiple organizations with complex structures in different geographic locations and need to collaboratively integrate the loosely coupled business processes of these organizations into new business processes.Therefore,the design and modeling of cross-organizational business process is a difficult problem in the field of business process management.Meanwhile,business processes are often subject to execution failures due to external factors,workflow system exceptions,or activity execution exceptions during execution.It's better to identify and handle abnormal events in business process as early as possible,which would help improve the business service levels and the quality of service to reduce costs and unnecessary losses.Based on the event log of cross-organizational business process,this thesis discusses the modeling method of cross-organization business process and the research of abnormal detection of business,the main description of work and innovation include:1.This thesis proposes a bottom-up multi-view modeling method for cross-organizational business process.The method mainly obtains the private process model,the common process model and the collaborative process model of the cross-organizational business process from the internal perspective,the public perspective and the collaborative perspective.Taking the event log of the cross-organizational business process as input,the mining algorithms are proposed to obtain the private process model of each organization.Then the simplification method of the model is defined to get the public view of the private process model,finally the cross-organizational business cooperation process model is obtained.2.In the field of data mining,an anomaly detection algorithm based on information entropy is proposed.Different from the traditional K-means clustering algorithm,the improved algorithm uses the concept of information entropy to weight the clustering object when selecting the initial clustering centers,so that the weighting function value is used to determine the initial clustering center with better quality.And the abnormal data would be judged in the process of continuous iterations of the algorithm.Experiments show that the proposed algorithm can effectively avoid the situation that traditional K-means algorithm being vulnerable to the initial clustering center and falling into the local optimal easily.Therefore,the algorithm can be used to obtain the abnormal event log at the data level.3.After getting the abnormal event log,it can be used to check with the existing process model for process compliance,after that,the diagnostic information map between the model and the log is obtained.Then it's easy to determine whether there are business anomalies and exception types in the actual process,so that it can help business process managers to discover and locate anomalies efficiently.