Combining the requirement information for software defect estimation in design time

• We examine the software defect predictive ability for the requirement and design metrics.
• We evaluate the ability of machine learning methods to predict software defects through widely used evaluation metrics.
• We cannot tell which one is more important than the other one.
• The predictors built on the combination of the requirement and design metrics are very effective.

This paper analyzes the ability of requirement metrics for software defect prediction. Statistical significance tests are used to compare six machine learning algorithms on the requirement metrics, design metrics, and combination of both metrics in our analysis. The experimental results show the effectiveness of the predictor built on the combination of the requirement and design metrics in the early phase of the software development process.