A search-based software engineering for defect prediction in ubuntu ecosystem

Tarumanagara International Conference on the Applications of Technology and Engineering

工业技术（总论）

Made Murwantara, I.^1 ; Sutrisno, S.^1 ; Joseph, J.^1

Informatics Dept, Faculty of Computer Science, Universitas Pelita Harapan, Indonesia^1

关键词: Defect identification;    Defect prediction;    Inter-dependencies;    Machine learning methods;    Potential defects;    Process management;    Search based software engineering (SBSE);    Search-based software engineering;   

Others  :  https://iopscience.iop.org/article/10.1088/1757-899X/508/1/012140/pdf
DOI  :  10.1088/1757-899X/508/1/012140

学科分类：工业工程学

来源: IOP

【 摘 要 】

A software defect may cost a lot for certain critical deployment such as communication and inter-process management that affects the CPU processes. A lot of time and efforts can be saved if we can predict the potential defect. In this study, we propose an approach to identify and predict the software defects that make use of Search-Based Software Engineering (SBSE) via a network model that collects its dataset from Ubuntu package repositories. We create a network model of package dependency that generated from the repository via a graph utility, namely debtree. In order to explicate the conflict between packages, we analyze the package inter-dependencies as related graph clusters. Since the massive number of dependencies, we use several Machine Learning methods to analyze and predict the potential software defect. Our investigation result shows that the SBSE technique has efficiently improved the defect identification process.