Implications of Ceiling Effects in Defect Predictors
The development of fault prediction models has been a very active research area. The reason for such a significant attention to automated quality predictors lays in their practical importance. Current models are useful, as they allow software project managers to better guide the allocation usually meager quality assurance resources to artifacts which need them the most. Recent results now indicate that this current research paradigm, which relied on relatively straightforward application of machine learning tools, has reached its limits.