Software fault prediction using object-oriented metrics

Fault-prediction techniques aim to predict the fault prone software modules in order to streamline the effort to be applied in the later phases of software development. Many fault-prediction techniques have been proposed and evaluated for their performance using various performance criteria. However, due to the lack of compiling their performances in proper perspective, one significant issue about the viability of these techniques has not been adequately addressed. In this study, an adaptive cost evaluation framework is proposed that incorporates cost drivers for various fault removal phases, and performs a cost-benefit analysis for the misclassification of faults. Accordingly, our study focuses on investigating two important and related research questions regarding the viability of fault prediction. First, for a given software product, whether performing fault prediction analysis is economically effective or not?. In case of an positive affirmation, then emphasis is provided on how to choose a fault prediction technique for an overall improved performance in terms of cost-effectiveness. In this study, Object-Oriented software metrics have been considered to provide requisite input data to design a classifier using statistical, machine learning and hybrid methods of soft computing. This work, also extends the study on finding the effectiveness of feature reduction techniques. From the obtained results, it is observed that performing fault prediction is quite desirable for those software systems, when the percentage of faulty modules are below the range of certain threshold value