Software defect prediction analysis using machine learning algorithms

Software Quality is the most important aspect of a software. Software Defect Prediction can directly affect quality and has achieved significant popularity in last few years. Defective software modules have a massive impact over software's quality leading to cost overruns, delayed timelines and much higher maintenance costs. In this paper we have analyzed the most popular and widely used Machine Learning algorithms — ANN (Artificial Neural Network), PSO(P article Swarm Optimization), DT (Decision Trees), NB(Naive Bayes) and LC (Linear classifier). The five algorithms were analyzed using KEEL tool and validated using k-fold cross validation technique. Datasets used in this research were obtained from open source NASA Promise dataset repository. Seven datasets were selected for defect prediction analysis. Classification was performed on these 7 datasets and validated using 10 fold cross validation. The results demonstrated the dominance of Linear Classifier over other algorithms in terms of defect prediction accuracy.