Existing model metrics and relations to model quality

Abstract

This paper presents quality goals for models and provides a state-of-the-art analysis regarding model metrics. While model-based software development often requires assessing the quality of models at different abstraction and precision levels and developed for multiple purposes, existing work on model metrics do not reflect this need. Model size metrics are descriptive and may be used for comparing models but their relation to model quality is not well-defined. Code metrics are proposed to be applied on models for evaluating design quality while metrics related to other quality goals are few. Models often consist of a significant amount of elements, which allows a large amount of metrics to be defined on them. However, identifying useful model metrics, linking them to model quality goals, providing some baseline for interpretation of data, and combining metrics with other evaluation models such as inspections requires more theoretical and empirical work.