Comprehensive model for software fault prediction

Abstract

Software Fault prediction (SFP) is an important task in the fields of software engineering to develop a cost effective software. Most of the software fault prediction is performed on same project date i.e., training and testing with same projects fault data. In case of unavailability of fault training data which is possible for the new project, data from the similar types/category of other projects can be used to train the model for the prediction. The software projects has been categorized into three categories by Boehm. The project within a certain group will be having good similarities with other projects within the group. So it is more suitable to train using the projects from same group. In this work we proposed to develop a model with similar category of data to predict the fault of another project belongs to same category. On basis of KLOC we have taken five organic software projects and performed various cross project and within project experiments. To generate a comprehensive generalized model for organic software's fault prediction, we have modeled various rule based to learner. Various rule-based learners used for comparison are JRip, CART, Conjunctive Rule, C4.5, NNge, OneR, Ridor, PART, and decision table-Naive Bayes hybrid classifier (DTNB).