Predicting software defects in varying development lifecycles using Bayesian nets

An important decision problem in many software projects is when to stop testing and release software for use. For many software products, time to market is critical and therefore unnecessary testing time must be avoided. However, unreliable software is commercially damaging. Effective decision support tools for this problem have been built using causal models represented by Bayesian Networks (BNs), which incorporate both empirical data and expert judgement. Previously, this has required a custombuilt BN for each software development lifecycle. We describe a more general BN, which, together with the AgenaRisk toolset, allows causal models to be applied to any development lifecycle without the need to build a BN from scratch. The model and toolset have evolved in a number of collaborative projects and hence capture significant commercial input. Extensive validation trials have taken place among partners on the EC funded project MODIST (this includes Philips, Israel Aircraft Industries a...

• Type: Journal
• Year: 2007
• Where: INFSOF
• Authors: Norman E. Fenton, Martin Neil, William Marsh, Peter Hearty, David Marquez, Paul Krause, Rajat Mishra
• Keywords: Causal Models, Development Lifecycle, INFSOF 2007, Software