Paper: Predicting Software Defect Density

June 21, 2007 by Artem Marchenko

Predicting Software Defect Density: 
A Case Study on Automated Static Code Analysis

 Artem Marchenko¹ and Pekka Abrahamsson² 

 ¹ Nokia, Hatanpäänkatu 1, FIN-33100 Tampere, Finland 
² VTT Technical Research Centre of Finland, 
P.O.Box 1100, FIN-90571 Oulu, Finland

Presented on XP2007.

Abstract. The number of defects is an important indicator of software quality. Agile software development methods put an explicit requirement on automation and permanently low defect rates. Code analysis tools are seen as a prominent way to facilitate the defect prediction. There are only few studies addressing the feasibility of predicting a defect rate with the help of static code analysis tools in the area of embedded software. This study addresses the usefulness of two selected tools in the Symbian C++ environment. Five projects and 137 KLOC of the source code have been processed and compared to the actual defect rate. As a result a strong positive correlation with one of the tools was found. It confirms the usefulness of a static code analysis tool as a way for estimating the amount of defects left in the product.

Keywords: agile software development, static code analysis, automation, defect
estimation, quality, embedded software, case study.

Permalink