Design and automation of a COSMIC measurement procedure based on UML models

Context. Many organizations are adopting the COSMIC method to size software products for estimating and controlling their development costs and performances. Using a functional size measurement method requires specialized expertise and can be time-consuming. Objectives. Since UML is the de facto industrial modeling language standard for object-oriented systems, it is very useful to understand how to exploit UML models for measuring software systems and for developing tools that can automatically derive the COSMIC size from them. This paper provides an answer to these needs. Method. We present a measurement procedure to derive the COSMIC functional size from UML software artifacts and a tool, named J-UML COSMIC, for the automation of the procedure. Based on the observation that different development processes are characterized by the use of different UML models, the tool has been designed to work with different UML artifacts (such as use case models, package diagrams, component diagrams, class diagrams, activity diagrams, and sequence diagrams) and to adapt to the specific employed process. To assess the measurement procedure and J-UML COSMIC, we have carried out two case studies and compared the measurement results provided by the tool with the ones obtained by experts applying the standard COSMIC method. Results. Using the proposed measurement procedure the tool is able to identify from UML software models all the COSMIC concepts and data movements identified by the experts. Moreover, the tool allows us to obtain incremental accurate measurements when new models are considered or existing ones are detailed. Conclusions. The designed approach is able to automatically measure the functional size starting from UML artifacts and providing higher accurate results when more data is available.