Measuring Process Similarity For Probabilistic Workflows

With the rapid development of the business environment,workflow technology gets more and more attention.Business process management is the focus of modern enterprise management,the word business in the business process connects the activity sequence with the satisfaction of customer value and growth of the value of company directly.Hence business process becomes the main mean of many large enterprises to normalize their business logic.It can help us to understand the specific process of a business and find the errors and flaws of the business process through analysis of business process,thus optimizing the business process.In order to make the company run efficiently,modern enterprises integrate many business processes.In a rapidly changing business environment,the needs of customers and enterprises continues to change,the corresponding business logic continues to change.Therefore,enterprises need to continuously improve the existing business processes or to develop new business processes.The consequences of this behavior lead to a growing number of business processes.In order to facilitate the management of these business processes,enterprises begin to build their own business process model library,which is a valuable resource for enterprises to solve business process management issues.These business process management libraries are being perfected and its number is being growing.One important problem in the management of process model repository is how to retrieve the relevant business processes.This retrieval capability directly determines the level of business process management.Because the core of business retrieval is to find the most similar relevant workflow,so the key is the calculation of similarity or distance between business processes.Based on measuring the structure similarity of workflows in traditional manner,we focus on computing the process similarity in probabilistic workflows.The workflow system is first formulated in probabilistic structure tree and we define the similarity measurement based tree editing distance.We then translate the tree into the probabilistic acyclic directed graph and the proof of trance equivalence.The corresponding algorithm is presented and its time complexity is also analyzed.Finally,we conduct extensive experiments on the method and evaluate the features by comparing with other methods.