Applying DPPI: A defect causal analysis approach using Bayesian networks

Abstract

Defect causal analysis (DCA) provides a means for product-focused software process improvement. A DCA approach, called DPPI (Defect Prevention-based Process Improvement), was assembled based on DCA guidance obtained from systematic reviews and on feedback gathered from experts in the field. According to the systematic reviews, and to our knowledge, DPPI represents the only approach that integrates cause-effect learning mechanisms (by using Bayesian networks) into DCA meetings. In this paper we extend the knowledge regarding the feasibility of using DPPI by the software industry, by describing the experience of applying it end-to-end to a real Web-based software project and providing additional industrial usage considerations. Building and using Bayesian networks in the context of DCA showed promising preliminary results and revealed interesting possibilities. © 2010 Springer-Verlag.