TY - GEN
T1 - An empirical study of eServices product UML sizing metrics
AU - Chen, Yue
AU - Boehm, Barry W.
AU - Madachy, Ray
AU - Valerdi, Ricardo
PY - 2004
Y1 - 2004
N2 - Size is one of the most fundamental measurements of software. For the past two decades, the source line of code (SLOC) and function point (FP) metrics have been dominating software sizing approaches. However both approaches have significant defects. For example, SLOC can only be counted when the software construction is complete, while the FP counting is time consuming, expensive, and subjective. In the late 1990s researchers have been exploring faster, cheaper, and more effective sizing methods, such as Unified Modeling Language (UML) based software sizing. In this paper we present an empirical 14-project-study of three different sizing metrics which cover different software life-cycle activities: requirement metrics (requirement), UML metrics (architecture), and SLOC metrics (implementation). Our results show that the software size in terms of SLOC was moderately well correlated with the number of external use cases and the number of classes. We also demonstrate that the number of sequence diagram steps per external use case is a possible complexity indicator of software size. However, we conclude that at least for this 14-project eServices applications sample, the UML-based metrics were insufficiently well-defined and codified to serve as precise sizing metrics.
AB - Size is one of the most fundamental measurements of software. For the past two decades, the source line of code (SLOC) and function point (FP) metrics have been dominating software sizing approaches. However both approaches have significant defects. For example, SLOC can only be counted when the software construction is complete, while the FP counting is time consuming, expensive, and subjective. In the late 1990s researchers have been exploring faster, cheaper, and more effective sizing methods, such as Unified Modeling Language (UML) based software sizing. In this paper we present an empirical 14-project-study of three different sizing metrics which cover different software life-cycle activities: requirement metrics (requirement), UML metrics (architecture), and SLOC metrics (implementation). Our results show that the software size in terms of SLOC was moderately well correlated with the number of external use cases and the number of classes. We also demonstrate that the number of sequence diagram steps per external use case is a possible complexity indicator of software size. However, we conclude that at least for this 14-project eServices applications sample, the UML-based metrics were insufficiently well-defined and codified to serve as precise sizing metrics.
UR - http://www.scopus.com/inward/record.url?scp=11244296280&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=11244296280&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:11244296280
SN - 0769521657
SN - 9780769521657
T3 - Proceedings - 2004 International Symposium on Empirical Software Engineering, ISESE 2004
SP - 199
EP - 206
BT - Proceedings - 2004 International Symposium on Empirical Software Engineering, ISESE 2004
T2 - Proceedings - 2004 International Symposium on Empirical Software Engineering, ISESE 2004
Y2 - 19 August 2004 through 20 August 2004
ER -