Extraction of Operational Workflow-based User Behavior Profiles for Software Modernization

Abstract

Static and dynamic analysis are the core parts in the software modernization process. They are required for the architecture reconstruction and the assessment of legacy software systems. One important use case is the extraction of user behavior profiles which can help in improving the system’s frontend layer. In this paper, we present our approach to extract and to visualize operational workflow-based user behavior profiles. Its implementation is based on two Java command line tools. The first tool extracts and anonymizes sessions from the records emitted by the monitoring framework Kieker. Based on these sessions, the second tool extracts a behavior model which is stored in several different graph formats on the file system. We evaluate our tools by instrumenting an industrial workflow-based Java web application for insurers. We show that our approach is able to automatically build and visualize a corresponding hierarchical behavior model. Such a model represents business processes as parents of workflows which in turn contain the visited views. Moreover, we show that this model can help in planning and prioritizing the software modernization process by identifying the most used and the least used views as well as the common screen- and workflow.