Abstract
Software defect prediction is a vital part in software reliability field. Here, the software modules are identified as defective or non-defective based on selection of the right set of software metrics. Software quality has become an emerging area for researchers to deal with challenges like defect prediction, defect removal, bug severity etc. The objective of this paper is to apply Machine Learning techniques for feature selection to remove redundant metrics and build software defect prediction model to label the software modules. We use principal component analysis technique as feature selection technique to reduce the redundant features on basis of Eigen value. After identifying the best set of software metrics, we develop software defect prediction models to classify software modules using Artificial Neural Network (ANN) and Decision Tree (DT). We have worked on data sets collected from data repository available on GitHub. The best set of software metrics act as predictors, which are obtained after removing irrelevant software metrics and building model for software defect prediction. We also compare the two classifying techniques on the basis of F-score, AUC, precision and sensitivity and accuracy.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have