| Organizations continuously modify the e-commerce systems that implement workflows of their daily operations. However, e-commerce systems may be designed without reference to the workflows specified by business analysts, or the original references are lost after revisions of the e-commerce systems. Therefore, it is difficult to get an overview picture of the current state of the workflows encoded in the e-commerce system. In addition, business analysts often lack the technical background to understand the source code. Furthermore, it is a labor intensive, time consuming and error prone process to reconfigure e-commerce systems to accommodate changes in the workflows. In this thesis, we present a technique that automatically extracts hierarchical workflows from the source code of e-commerce systems. To generate a high level abstract representation of workflows, we analyze the user-system interaction and users' navigation flows to identify tasks and the coordination between the tasks. To provide detailed descriptions of workflows, we capture business policies and data flows in the source code to infer the business logic.; To refine the results of the extracted workflows, we incorporate workflows in the documents specified by business analysts. We compare the documented workflows with the extracted workflows and calculate the similarity between two types of workflows based on their structures. From the comparison results, we establish the linkages between the source code and the documented workflows. Finally, we develop a prototype tool as an eclipse plug-in. This prototype tool extracts workflows from three different e-commerce systems to evaluate the effectiveness of our techniques for workflow extraction. |
