Abstract:

To classify the defects, a defect hierarchical prediction model is proposed. From analyzing the defects attributes, a defect attribute prediction method is built. By using this method we can make the similar attributes defects together. From the analysis of the relationships among the similar attributes defect clusters, the defect relationship prediction theory is proposed, which can group the defects with strong relationships. The results indicate that the system architecture defect hierarchical prediction is better than k-means especially on the cluster density, numbers and objectivity.

Key words: system architecture, defect prediction, hierarchy