Recovering Behavioral Design Models from Execution Traces

Recovering behavioral design models from execution traces is not an easy task due to the sheer size of typical traces. In this paper, we describe a novel technique for achieving this. Our approach is based on filtering traces by distinguishing the utility components from the ones that implement high-level concepts. In the paper, we first define the concept of utilities; then we present an algorithm based on fan-in analysis that can be used for the detection of utilities. To represent the high-level behavioral models, we explore the Use Case Map (UCM) notation, which is a language used to describe and understand emergent behavior of complex and dynamic systems. Finally, we test the validity of our approach on an object-oriented system called TConfig.