ProPRED: A probabilistic model for the prediction of residual defects

In this paper, we propose ProPRED, a probabilistic model for predicting residual defects based on Bayesian Networks (BN) in the software development lifecycle. With the chain rule for BN, ProPRED can be used to take the evidence of the influential factors to the activities (Analyze and Design, Development, Maintain, and Review and Test) that bring about the defects introduction and removal to reason and predict the probable residual defects. We refine and classify the influential factors to the four basic activities, and construct the ProPRED. Giving a case study, we conclude that the ProPRED improve its performance in reasoning under uncertainty and convenience in decision-making and quality control.