Parsimonious Classifiers for Software Quality Assessment

—Modeling  to  predict fault­proneness of software modules is an important area  of research in software engineering. Most such models employ a large number of basic and derived metrics as predictors. This paper presents modeling results based on only two  metrics, lines of code and cyclomatic complexity, using  radial basis functions with Gaussian kernels as classifiers. Results from two  NASA systems are presented and analyzed. Index  Terms –  Software quality, Software metrics, Classification, Parsimonious classifiers

• Type: Conference
• Year: 2007
• Where: HASE
• Authors: Miyoung Shin, Sunida Ratanothayanon, Amrit L. Goel, Raymond A. Paul
• Keywords: HASE 2007, Only two Metrics, Software Metrics, —Modeling To Predict