To fill that gap, Madeyski and Majchrzak \cite{Madeyski2014} proposed using a new tool called DePress. DePress (Defect Prediction for Software Systems), based on the KNIME framework \cite{KNIMESite}, allows the creation of graphical workflows and uses an intuitive, user-friendly interface (Figure \ref{fig:workflow}). Possible applications of DePress vary from software measurement, product and process improvement, quality assurance processes, integrating data from various tools and other sources for the sake of data analyses as well as its main task: KNIME’s data mining capabilities for software defect prediction purposes. Being intuitive and highly customizable, DePress makes itself a perfect tool which can be conveniently utilized (thanks to its user-friendly interface) in different commercial software development projects for defect prediction introduction and is freely available for any purposes under its GNU General Public License \cite{DePressSite}.
Potential benefits from usage of DePress framework for commercial software development projects have not been investigated \cite{Madeyski2014}. To fill that gap, our study aimed to answer the following research questions:
RQ1: What is the highest level of prediction recall achievable by DePress tool in basic configuration mode, using an industrial project’s data?
The possible benefit varies, depending on the potential prediction effectiveness. This implies the need of first verifying what is the highest recall value of the defect prediction process handled entirely by the DePress tool. As DePress can be highly customizable thanks to its plugin-based architecture, as well as being open source can make possible adjustments complex and expensive. Therefore, we decided to restrict the DePress usage only to its basic set-up.
RQ2: Will usage of the DePress framework pay off for an industrial project?
To answer this question, we had to compare the costs of introducing and using the DePress based defect prediction to the potential benefits generated by its introduction. To achieve this, we used values such as return on investment (ROI) and benefit-cost ratio (BCR).
RQ3: How cost effective is the DePress framework for defect prediction for an industrial software development project and what is the highest possible prediction recall achievable with that tool?
If the answer to the first research question is positive, the next step is to verify what will be the profit from the highest prediction recall achievable using DePress’ basic defect prediction set-up.