摘要
Testing and debugging software to fix bugs is considered one of the most important stages of the softwarelife cycle. Many studies have investigated ways to predict bugs in software artifacts using machine learningtechniques. It is important to consider the explanatory aspects of such models for reliable prediction. In thispaper, we show how feature transformation can significantly improve prediction accuracy and provide insightinto the inner workings of bug prediction models. We propose a new approach for bug prediction that firstextracts the features, then finds a weighted transformation of these features using a genetic algorithm that bestseparates bugs from non-bugs when plotted in a low-dimensional space, and finally, trains predictive modelsusing the transformed dataset. In our experiment using the proposed feature transformation, the traditionalmachine learning and deep learning classifiers achieved an average improvement of 4.25% and 9.6% in recallvalues for bug classification over 8 software systems compared to the models built on original data. We alsoexamined the generalizability of our concept for multiclass classification tasks such as commit classification insoftware systems and found modest improvements in F1-scores (sometimes up to 3%) for traditional machinelearning models and 4% with deep learning models.
学科与关键词
- Dynamic Selection of Classifiers in Bug Prediction: An Adaptive MethodDi Nucci, Dario*; Palomba, Fabio; Oliveto, Rocco; De Lucia, AndreaIEEE TRANSACTIONS ON EMERGING TOPICS IN COMPUTATIONAL INTELLIGENCE, 2017, 1(3): 202-212.被引 56
- A guided oversampling technique to improve the prediction of software fault-proneness for imbalanced dataRaed Shatnawi; Ziad Al SharifInternational Journal of Knowledge Engineering and Data Mining, 2012, 2(2/3): 200.
- Software fault prediction using data mining, machine learning and deep learning techniques: A systematic literature reviewBatool, Iqra; Khan, Tamim Ahmed*COMPUTERS & ELECTRICAL ENGINEERING, 2022, 100: 107886.被引 36
