Prediction of software development faults in PL/SQL files using neural network models

Cost-sensitive boosting neural networks for software defect prediction

“…Lines of code (LOC) is a commonly used size metric for defect prediction (Akiyama, 1971) while McCabe (1976) and Halstead (1977) are the mostly used complexity metrics. Many works have been done to find the correlation of software metrics and defectproneness by building different predictive models including discriminant analysis (Khoshgoftaar, Allen, Kalaichelvan, & Goel, 1996;Munson & Khoshgoftaar, 1992), logistic regression (Basili, Briand, & Melo, 1996;Gyimothy, Ferenc, & Siket, 2005;Zhou & Leung, 2006), factor analysis , fuzzy classification (Ebert, 1996), classification trees (Gokhale & Lyu, 1997;Gyimothy et al, 2005;Koru & Liu, 2005;Menzies et al, 2007), Bayesian network (Pai & Dugan, 2007;Zhou & Leung, 2006), artificial neural networks (ANN) (Gondra, 2008;Gyimothy et al, 2005;Kanmani, Uthariaraj, Sankaranarayanan, & Thambidurai, 2007;Khoshgoftaar, Lanning, & Pandya, 1994;Khoshgoftaar, Allen, Hudepohl, & Aud, 1997;Neumann, 2002;Quah & Thet Thwin, 2004) support vector machines (Gondra, 2008;Xing, Guo, & Lyu, 2005), etc. Since the relationship between software metrics and defect-proneness of software modules are often complicated and nonlinear, machine learning methods such as neural networks have been shown more adequate for the problem than traditional linear models (Khoshgoftaar et al, 1994(Khoshgoftaar et al, , 1997 AdaBoost -an adaptive boosting algorithm (Freund, 1995;Freund & Schapire, 1997), which has shown to be an effective ensemble learning method to significantly improve the performance of neural network classifiers (Schwenk & Bengio, 2000).…”

Regression via Classification applied on software defect estimation

“…Recent studies exploring regression models suggest discriminant analysis (Bellini, Bruno, Nesi, & Rogai, 2005) and neural networks for predicting the number of faults in modules (Lanubile, Lonigro, & Visaggio, 1995;Lanubile & Visaggio, 1997). Neural nets are also used in Quah and Thwin (2004) in order to predict the number of faults in PL/SQL projects. In Khoshgoftaar and Seliya (2002), a case study is presented where various tree based regression models using design metrics are suggested for predicting the number of faults in modules.…”

Prediction approach of software fault-proneness based on hybrid artificial neural network and quantum particle swarm optimization

“…Although software faults have been studied using these techniques, there are still many aspects of faults remaining unclear. We notice that the relationships between software metrics and fault-proneness are often complex and nonlinear [13], the adequacy of traditional linear models is compromised, which results in the development of non-linear models, and is expected to provide superior performance than the linear models.…”

“…Some of the authors have used PCA to reduce dimensionality [13]. However, a disadvantage of PCA is that, in contrast to the original input variables, the derived dimensions may not have an intuitive interpretation.…”