A Framework for Execution-based Model Profiling

In Model-Driven Engineering (MDE) models are put in the center and used throughout the software development process in prescriptive ways. Although these prescriptive models are important during system implementation, descriptive models derived from runtime data offer valuable information in later phases of the system life cycle. Unfortunately, such descriptive models are only marginally explored in the field of MDE. Current MDE approaches mostly neglect the possibility to describe an existing and operating system using the information upstream from operations to design. To create a link between prescriptive and descriptive models, we propose a unifying framework for a combined but loosely-coupled usage of MDE approaches and process mining (PM) techniques. This framework embodies the execution-based model profiling as a continuous process to improve prescriptive models at design-time through runtime information. We provide an evaluation case study in order to demonstrate the feasibility and benefits of the introduced approach. In this case study we implement a prototype of our framework to register logs from a running system. The implemented prototype transforms the registered logs into XES-format for further processing and analysis via PM algorithms. We prove that the resulting model profiles are sufficient enough for runtime verification. Furthermore, we demonstrate the possibility to maintain model profiles for multiple concerns, such as functionality, performance and components interrelations, through the unifying framework.