Project Plan

The work plan is divided into five work packages with dependencies and expected interactions among them, following the usual stages needed to reach the project’s objectives.

WP1 Determine a taxonomy of defect types and prepare case studies for software defect prediction

• T1.1 – Conduct a literature review to identify current defect type taxonomies and approaches for their analysis
• T1.2 – Determine  a taxonomy of defect types using unsupervised learning applied on open-source data sets/software systems
• T1.3 – Perform software quality  analysis on the systems identified at T1.2
• T1.4 – Prepare appropriate case studies that will be used for defect prediction, according  to the established bug taxonomy

WP2 Establish machine learning based methods for feature engineering in software defect prediction

• T2.1 – Literature review on existing methods for automatic feature learning, coupling and cohesion-based features for software defect prediction
• T2.2 – Developing deep learning methods for learning semantic  features from software artifacts
• T2.3 – Experimental  evaluation and analysis of the learned features
• T2.4 – Defining coupling and cohesion-based software metrics for software defect prediction
• T2.5 – Experimental  evaluation and analysis of the coupling and cohesion-based software metrics   

WP3 Develop new deep learning algorithms software defect prediction

• T3.1 – Conduct literature review to identify current machine learning results in software defect prediction
• T3.2 – Develop one-class classification approaches for defect prediction
• T3.3 – Develop one-shot learning approaches for defect prediction
• T3.4 – Experimentally evaluate deep learning techniques and compare them with existing approaches

WP4 Development of QuaDeeP

• T4.1 – Design the QuaDeeP system using an incremental development process
• T4.2 – Development of defect prediction functionality
• T4.3 – Testing and final validation of the QuaDeeP prototype

WP5 Project management