Applying Documentation Metrics in Cross Version Defect Prediction Modeling
Open AccessSoftware documentation such as documented Application Programming Interface (API) and comments embedded with the software code aid in faster debugging of software defects. The existence of this documentation is used as a measurement of software quality. Over the last 40 years, a series of object-oriented metrics-based defect prediction models have been successfully developed. However, documentation metrics in combination with object-oriented metrics in software defect prediction modeling has not been explored in predicting defects. By leveraging a publicly available GitHub dataset that contains both documentation and object-oriented metrics and applying the cross version defect prediction approach, this research determined documentation metrics impact on defects across a project and developed a predictive model using these metric in combination with baseline metrics to improve model performance. As a result, Boosting ensemble method returned improved model performance when combining the documentation metrics with commonly used object-oriented code metrics. Likewise, the Random Forest returned an improved model when using a feature subset. Random Forest using a subset of metrics provided the most promising results with F-Measure performance improvement of 8.9 percent. The results of this research highlight quantitatively the impact documentation metrics have on software defect prediction and that model performance can improve when identifying a subset of metrics. The results also demonstrate the use of data from three previous versions versus solely using the latest version, the models perform within an average of five percentage points of each other. This knowledge can be leveraged by managers to enhance the application of documentation throughout the lifecycle of software.
- All rights reserved
Notice to Authors
If you are the author of this work and you have any questions about the information on this page, please use the Contact form to get in touch with us.