On Evolutionary Algorithms for Multiple Criteria Decision Support in Bayesian Belief Networks Models of Dependable Software Development Processes.

Recent research in software development process assessment and modelling has led to an increase demand for formalisms capable of providing reasoning under uncertainty. Such methods are used for providing decision support and build expert consensus when there is a huge degree of subjectivity. Researchers have argued that Bayesian belief networks (BBNs) is one of the most suitable formalism for this task.

However, Bayesian belief networks have typically been used to allow the user to
identify the most suitable software development process in light of one objective only; this is usually product quality or number of latent faults in the product. In fact, the current BBN formalism does not allow the user to identify the optimal process with respect to many objectives. In this paper we argue that multiple objective genetic algorithms (MOGAs) embedded with the BBN model of the software development process can tackle this limitation. The proposed Decision support system (DSS) searches for those solutions that maximize the confidence in the product integrity whilst minimizing the costs and the time taken to develop the product.