Evidential Diagnosis Of Inconsistencies In Object-Oriented Designs

This paper presents a diagnostic framework for assessing the significance of inconsistencies (i.e., violations of consistency rules) in software design models expressed in the Unified Modeling Language (UML). The assessment is based on significance criteria that software designers can specify and associate with specific consistency rules. These criteria define characteristics that the model elements involved in the violation of a rule should have for the inconsistency to be significant, and they are specified in a formal language derived from the Object Constraint Language (OCL). The satisfiability of the criteria by individual model elements is measured by belief functions defined by the framework. The measures generated by these functions are used to rank the inconsistencies caused by different model elements. The presented framework has been evaluated through a set of experiments. The results of these experiments show that criteria definable in the framework can be used to produce relatively elaborate significance-rankings of inconsistencies.