A Similarity Measurement Based On Structure Of Business Process

Today with quickly developing of the business environment, the business processes have been an important and indispensable part of the modern enterprise. A business process consists of a set of tasks, and their order of invocation and information flow, according to controlling flow dependencies. The processes are executed to achieve a particular business goal. And the business processes not only describe the important processes in the organization,it also provides lots of guidance and a series of specifications to optimize resource and avoid loss. These specifications promote the management standardization and enterprise benefit maximization. The essence of similarity measure is analyzing the different processes in a similar degree and quantifying it with a specific value, according to different business needs.The business processes similarity measure is significant in process management. It can be used in many ways such as process model search-indexing, facilitating reuse, etc. The existing researches focuses on solving the specific problems of application scenario at an upper angle, and they are mostly mapping the elements of business process models based on the syntactic or semantic of the activities' text labels. Those researches ignore the relationship of the other components in models, and can not be efficiently applied for heterogeneity of processes.This paper proposes an efficient algorithm for similarity measurement based on the internal structure of the processes. This measure uses Petri-net to make a process model, and the elements(transition and place) in models are both mapped by an iterative mapping strategy respectively to identify the correspondence of every node in different models. And finally after getting a stable mapping which trend to the best, we compute the similarity modulus. This method reduces the dependence on external characteristics such as text labels.The main work of this paper described as follows:1. In the term of establishing mapping function between models, an approach based on the internal structure with Petri-net is proposed. This measurement builds the best mapping function without the text labels of activities, and it is combined with the relative position ofthe various elements of the process and the specific context. This method makes up the drawback of the existing process methods which can not match accurately and calculate effectively in the cross-organizational or cross-disciplinary scene.2. In terms of defining the similarity modulus, the modulus of similarity between elements and models by the structure of processes is proposed. It presents the similarity with a specific value without the use of the syntax and semantics of text labels, and it is calculated by the logical structure of process such as concurrent, asynchronous, choice.3. In terms of similarity calculation, this paper establishes a coarse-grained algorithm to obtain a modulus between 0 and 1 to analyze the total similarity of the different processes. It reduces the dependence on external characteristics such as text labels of activities4. In the experiment, we verify and compare the algorithm in this paper with other similarity measures such as the methods based on the text labels and graph edit distance. We also analyze the accuracy and time efficiency of the algorithm on the real data sets.