A Novel Feature Subset Selection Algorithm for Software Defect Prediction

Abstract

Software Defect Prediction has been an area of growing importance. It is always required to maintain high reliability and high quality for any software being developed. A software quality prediction model is built using software metrics and defect data collected from a previously developed system release or similar software project. Upon validation of such a model, it could be used for predicting the fault-proneness of program modules that are currently under development. A low quality or fault prone prediction can justify the application of available quality improvement resources to those programs. In contrast, a non fault prone prediction can justify non- application of the limited resources to these already high- quality programs. And finally high software reliability and quality are maintained with an effective use of the available resources. A feature extraction algorithm based on graph clustering is applied over the historical software data collected for defect prediction purpose and its impact on different data sets are analyzed in this paper.