Software Quality Modeling as a Reliability Tool
Published Online: 15 SEP 2008
Copyright © 2007 by John Wiley & Sons, Inc.
Wiley Encyclopedia of Computer Science and Engineering
How to Cite
Seliya, N. and Khoshgoftaar, T. M. 2008. Software Quality Modeling as a Reliability Tool. Wiley Encyclopedia of Computer Science and Engineering. 1–8.
- Published Online: 15 SEP 2008
Software quality and reliability activities are important parts of software project development, and advances in related tools aim to deliver a high-quality and dependable software product. This article focuses on an effective tool in software quality engineering, namely software quality estimation models. Among existing software quality classification models, tree-based models are particularly attractive because of their model simplicity and ease in model interpretation. We perform an empirical comparative study of five tree-based software quality models. Software measurement and defect data of a very large real-world software system is used in our case study. The tree-based models are compared for model complexity, model structure, and classification accuracy. In addition, statistical analysis using ANOVA and multiple-pairwise hypothesis testing is used to evaluate the respective models based on their expected costs of misclassifica-tion. It is shown that at lower cost ratios the Sprint-Sliq model performs the best, whereas at higher cost ratios the CART model performs the best. Considering ease in model interpretation and model complexity, CART and S-PLUS are good choices. For a more robust model with good classification accuracy, Sprint-Sliq is a better tree-based model.
- software quality;
- classification trees;
- software metrics;
- software reliability;
- statistical analysis