Using Citation Influence and Social Network Analysis to Predict Software Defects
Wei Hu · ERA: Education and Research Archive (University of Alberta) · 2013
Resource constraints, e.g. lack of time and human resources, is a major issue in software testing practice. In short, testers have limited time to test software systems. Therefore, managers are expected to spend more resources on software components that are likely to contain many defects. To help managers make better decisions of selective testing, it is beneficial to identify defect-prone software components before the actual testing. In this thesis, we propose a model for software defect prediction. The proposed model combines the topological properties of the software dependency network and the textual information in source code to predict defect-prone software components. We evaluate our model on data from Eclipse, Netbeans, and Gnome projects at different levels of granularity. The evaluation results are encouraging, showing that our model achieves higher prediction accuracy than prior work.