An empirical study of software architectures’ effect on product quality

TitleAn empirical study of software architectures’ effect on product quality
Publication TypeJournal Article
Year of Publication2011
AuthorsHansen, Klaus Marius, Jonasson Kristjan, and Neukirchen Helmut
Secondary TitleJournal of Systems and Software
Volume84
Issue7
Pagination1233 - 1243
Date Published7/2011
ISSN Number01641212
Keywordsflossmole cited, java
Abstract

Software architecture is concerned with the structure of software systems and is generally agreed to influence software quality. Even so, little empirical research has been performed on the relationship between software architecture and software quality. Basedon 1,141 open source Java projects, we calculate three software architecture metrics (measuring classes per package, normalizeddistance, and a new metric introduced by us concerning the excess of coupling degree) and analyze to which extent these metricsare related to product metrics (defect ratio, download rate, methods per class, and method complexity). We conclude that there area number of significant relationships between product metrics and architecture metrics. In particular, the number of open defectsdepends significantly on all our architecture measures.

Notes

"We first collected meta-data on the 21,094 most highly ranked Java projects on 2009-03-17 from SourceForge for which it was possible to get such data. Here “Java projects” were defined as projects belonging to “trove” 198 at SourceForge and “rank” was the SourceForge ranking of projects. The data consisted of characteristics such as number of bugs, time of latest file upload, number of developers, number of open bugs, and SourceForge “rank”." They did not use FLOSSmole, "A more complete analysis of the project status, could be performed by mining the FLOSSMole database (Howison et al., 2006)."

URLhttp://www.sciencedirect.com/science/article/pii/S0164121211000574
DOI10.1016/j.jss.2011.02.037
Short TitleJournal of Systems and Software