A Software Evaluation Model Using Component Association Views ** This work was funded by IBM Canada Ltd. Laboratory - Center for Advanced Studies (Toronto) and the National Research Council of Canada.

In this paper, we introduce a view-based architectural design evaluation model that allows to quantitatively evaluate and categorize the design of a software system. The model is based on the notion of component association which is a generalization of coupling and cohesion metrics. The component association is defined as a measure of the overall dependency among high-level system components such as files, modules, or subsystems with regard to a collection of criteria. The associations are discovered by applying data mining techniques on a database of data and control flow dependencies extracted from the software system. The proposed association-views and modularity metrics allow the user to evaluate the design quality of a software system.