On the Quantitative Assessment of Class Model Compositions: An Exploratory Study

Author(s): Kleinner Oliveira, Alessandro Garcia, Jon Whittle
Venue: Empirical Studies of Model-Driven Engineering (ESMDE '08)
Date: unknown

Type of Experiement: Case Study

Quality
3

Summary:
This paper is a study to explore a set of metrics used for assesing model compositions. These metrics assess the impact inflicted on the input models properties after a model composition. To accomplish this the authors set up to two case studies for experimentation.

The first case study deals with real-world models for an Automated Highway Toll System. In this case study there are three packages that will be merged into a composite model. The first package is a UML diagram which specifies functionality for: creating user account, adding funds, ect. The second package specifies: synchronization of account, processing credit cards, ect. And package three represents functionality to add transporters.

the NOPC metrics have a higher measurement following merge strategy than override strategy
According to the measures concerning the number of associations between classes (NASC), the number of abstract syntax conflicts (NAbSC), and the number of subclasses (NSUBC) and superclasses of a class (NSUPC), no significant difference was detected in favor of a specific composition strategy when applied to the two case studies.

The second case study deals with a literature based calculator which has two packages. The first package is a UML diagram which specifies the the functionality of add and subtract while package two specifies the functionality for multiply and divide.

Although not showing differences between each other regarding the NASC, NSUBC, and NSUPC metrics, the output models present significant differences. The Package AB produced by the merge strategy has higher values for some metric measures such as NATC, NOPC, CBC, NSCC, NCME, and NBFC. On the other hand, Package AB produced by the override strategy provided higher results in one only measure, NUM

0