Towards Quality-Driven Model Transformations: A Replication Study

Commonly, there are several ways to transform a source model into a target model. These alternative target models may have the same functionality but can differ in their quality attributes. One of the key challenges of an automated transformation process is to identify the transformations that will produce a target model with the desired quality attributes. In this paper, we present a replica of a controlled experiment to investigate the selection of alternative transformations to obtain UML class models from a Requirements Model. This is a concrete instantiation of a wider domain-independent approach for quality-driven model transformation. Specifically, we focus on a set of transformations related to structural relationships between classes (association, aggregation and association class) and the understandability quality attribute. Although, some results could be foreseen even by a superficial analysis of the alternatives, the goal of this work is to use experimentation to gather empirical evidence about which alternative transformation produces the UML class model that is the easiest to understand. The empirical results support the original results showing that there is a tendency to favor the use of association relationships to drive these transformations when understandability is chosen.