In industrial software development projects, quality assurance (defect fixing) consumes a significant amount of time and resources. By using the defect prediction technique, project members can obtain information on possible defect-prone elements of the software, before defects will occur, to optimally plan their quality assurance process. What is proposed in this paper, is a simple effort allocation strategy which is based on DePress framework-driven defect prediction, Pareto principle and Boehm Law, and which eliminates most of the quality assurance work during after-development project’s phases. Such an approach significantly increase quality assurance costs in development phase, however overall, QA costs will decrease (Figure \ref{fig:cost_comparison}) in comparison to actual, real-life costs observed in the investigated project, as significantly less discoverable defects are left to be fixed in later phases, where, according to Boehm’s Law, bug-fixing costs are considerably higher. At the same time, we need to mention the low investment costs, when a Knime-based DePress framework will be used for defect prediction purposes. Low investment costs and high recall of even simple defect prediction performed by DePress, can result with high \(NetReturn\) of DePress-aided quality assurance planned basing on proposed effort allocation strategy.
\printbibliography