Application of adaptive neuro-fuzzy inference system for predicting software change proneness

Abstract

In this paper, we model the relationship between object-oriented metrics and software change proneness. We use adaptive neuro-fuzzy inference system (ANFIS) to calculate the change proneness for the two commercial open source software systems. The performance of ANFIS is compared with other techniques like bagging, logistic regression and decision trees. We use the area under receiver operating characteristic (ROC) curve to determine the effectiveness of the model. The present analysis shows that of all the techniques investigated, ANFIS gives the best results for both the software systems. We also calculate the sensitivity and specificity for each technique and use it as a measure to evaluate the model effectiveness. The aim of the study is to know the change prone classes in the early phases of software development so as to plan the allocation of testing resources effectively and thus improve software maintainability.