Using Citation Influence to Predict Software Defects

TitleUsing Citation Influence to Predict Software Defects
Publication TypeConference Proceedings
Year of Publication2013
AuthorsHu, Wei, and Wong Kenny
Refereed DesignationRefereed
Secondary Title10th Working Conference on Mining Software Repositories
Date Published05/2013
Keywordseclipse, netbeans, social network
Abstract

The software dependency network reflects structure and the developer contribution network reflects process. Previous studies have used social network properties over these networks to predict whether a software component is defect-prone. However, these studies do not consider the strengths of the dependencies in the networks. In our approach, we use a citation influence topic model to determine dependency strengths among components and developers, analyze weak and strong dependencies separately, and apply social network properties to predict defect-prone components. In experiments on Eclipse and NetBeans, our approach has higher accuracy than prior work.