Abstract:Frequent pattern mining from defect repositories has started playing an important role in software defect detection and analysis. In this paper, we present a new improved template-based method to generate association rules from defect repositories. We aim to provide manager and tester with more efficient method of generating rules. The improved method optimizes the generation using rule templates firstly, then reduces unnecessary rule generation by the theorems described in Section III-A. The advantages of the improved method are validated by an experiment based on 1860 defect reports. The experimental results show the improved template-based method can significantly reduce the total number of the rules, and shorten the run time. The pruned association rule set makes it easier to find the interesting rules.

An Improved Template-Based Method for Mining Association Rules from Defect Repositories