Research On Change Management Of Data-centric Business Processes

With the rapid development and change of the business environment,business processes in enterprises need to be improved constantly.The key to achieving business process improvement quickly and effectively lies in effective business process change management.With the increasing importance of data,people pay more attention to data-centric business processes.The artifact is the key data entity in the data-centric business process,which drives business operations to achieve business targets and evolves as business operations are implemented.Compare with the traditional activity-centric business processes,artifact-centric business process has high flexibility in the aspect of business process implementation and improvement.However,change management of Artifact-centric business process still has its deficiencies to meet an enterprise's need to improve business processes quickly and effectively.At present,the studies on artifact-centric business process change management mainly focus on business process model change.Few researchers have noticed business process instance changes.The existing business process monitoring methods mainly focus on the activity-centric business process,which can not effectively monitor the artifact-centric business process,thus indirectly prolonging the cycle of artifact-centric business process change.These defects lead to the poor adaptability of artifact-centric business processes to dynamic environment.In order to realize the rapid and effective change of artifact-centric business processes,this paper studies the change management of artifact-centric business processes.Artifact-centric business processes change management includes multiple tasks in each stage of the lifecycle of artifact-centric business process change.Artifact-centric business process instance migration,real-time performance monitoring of business processes,correctness prediction of business process instance changes,and compliance prediction of business goals are four key tasks in the artifact-centric business process change management.Aiming at these key tasks,the main work of this paper is as follows:(1)We propose an artifact-centric business process instances migration and its correctness verification method.This method is based on a special business rule,i.e.,adaptation rule and uses the defined adaptation rule to realize the semi-automatic migration of artifact-centric business process instances.The soundness and correctness of the migration rules ensure that the migrated business process instances can achieve users desired final state without deadlock.Here,we provide a verification tool to help process designer to check the soundness and correctness of an adaptation rule.(2)We propose a model-driven real-time performance monitoring framework for monitoring artifact-centric business processes.Real-time performance monitoring is an important method to find the opportunity of improving business process models.The proposed framework obtains the real-time data needed by monitor based on event rules and removes the close coupling between the business process system and the real-time performance monitoring system.Beyond this,this method constructs the monitoring model based on artifacts of the collaborative business process model and transforms the monitoring model into the technical implementation of real-time performance monitoring.It makes the real-time performance monitoring configurable and reduces the sensitivity of real-time performance monitoring to business process changes and monitoring requirements changes.Through technical analysis,the proposed real-time performance monitoring framework is feasible.Through experiments,the real-time data acquisition method based on event rules is feasible and time-effective.(3)We propose a method for predicting the correctness of artifact-centric business process instance change and present two optimizations of the proposed prediction.When the artifact-centric business process instance changes,it is necessary to ensure the correctness of the business process instance changes,i.e.,the changed business process instance can still reach the final state expected by users.a random forest classifier is trained based on business process instances' historical snapshot.Then,the trained classifier is used to predict the correctness of the business process instance changes.The optimizations reduce the negative impact of “conflict samples” on classifiers by removing the “conflict snapshot” in the business process instances' historical snapshots and remove redundant features which are helpless for classification by projecting on the “constraint attributes”.Through experiments,the proposed method has better accuracy and F1-score when compared with the traditional formal verification method and the proposed optimizations are effective.As such,the proposed method can avoid incorrect business process instance changes better.(4)We propose a business goal compliance prediction framework based on business-related data and context data.The proposed framework consists of two parts: offline training and runtime prediction.In the offline training part,the feature extraction method of time series is used to obtain the features of context data.Features of context data and business-related data are combined to train a random forest classifier.In the run-time prediction part,the trained classifier is used to predict the compliance of business goals.Through experimental analysis,when context data has an impact on the outcomes of a business process,the business goal compliance prediction results based on business-related data and context data have better accuracy.Therefore,it can find more incompliances of business goals.As such,we can have more improvement opportunities for individual business process instances.