Software Defect Prediction based on Machine Learning and Deep Learning

Abstract

Software system quality can be enhanced by reducing the possible software defects in the system, this can be achieved by regular monitoring of the system for any defect alerts. Identifying the defects with the features is difficult and this system is researched less in the literature. The automated tool will be useful for maintaining high quality of the system. The defects may create a huge loss to enterprises, thus detecting early and accurately is mandatory for the system quality and helps in Software development life cycle (SDLC). The proposed work introduced machine learning algorithm Naïve Bayes and deep learning algorithm Long Short Term Memory(LSTM) and Deep Neural Network (DNN). The dataset considered for the proposed study is PROMISE dataset as a binary prediction. As the software defect prediction is binary, the classification model is opted for this study. Thus NB on ML model is used and DNN, LSTM are compared with their accuracy. Experimental study showed that DNN algorithm outperform in accurate detection of software defects.