Automating and Evaluating Probabilistic Cause-Effect Diagrams to Improve Defect Causal Analysis

Defect causal analysis (DCA) has shown itself an efficient means to obtain product-focused software process improvement. A DCA approach, called DPPI, was assembled based on guidance acquired through systematic reviews and feedback from experts in the field. To our knowledge, DPPI represents an innovative approach integrating cause-effect learning mechanisms (Bayesian networks) into DCA meetings, by using probabilistic cause-effect diagrams. The experience of applying DPPI to a real Web-based software project showed its feasibility and provided insights into the requirements for tool support. Moreover, it was possible to observe that DPPI’s Bayesian diagnostic inference predicted the main defect causes efficiently, motivating further investigation. This paper describes (i) the framework built to support the application of DPPI and automate the generation of the probabilistic cause-effect diagrams, and (ii) the results of an experimental study aiming at investigating the benefits of using DPPI’s probabilistic cause-effect diagrams during DCA meetings.