Defect Prediction Framework Using Adaptive Neuro-Fuzzy Inference System (ANFIS) for Software Enhancement Projects

Abstract

Software Defect Prediction is the process of forecasting the defect count during various phases of software development life cycle. Defect prediction is vital to successful software project execution since the output is used to proactively plan defect prevention activities. During initial phases of software development life cycle, prediction is quite challenging due to the presence of uncertainty in input parameters, which constitute major component of estimated effort. Multiple attempts have been made by researchers in past to design an appropriate defect prediction model but so far none has found widespread adoption in software industry. In this communication, Adaptive Neuro-fuzzy Inference system (ANFIS) approach has been proposed for designing a defect prediction model. In order to achieve complexity reduction and to increase model adoption, an easy-to-use graphical user interface is designed. The proposed ANFIS based model makes use of organization’s historical projects’ data for building the model. The model provides a defect range (minimum, maximum) as a prediction output. The effectiveness and superiority of proposed ANFIS model is demonstrated through analysis of results achieved.