1605 An Experimental Evaluation of Unexpected Rule Mining toward Causal Analysis on Defect Correction Time

Abstract

This paper proposes an approach for supporting causal analysis of defect correction effort from defect attributes in a bug tracking system. The proposed approach supports analyst by showing relationships between correction efforts and combined attributes as rules. The approach first mines major rules that apply many defects in the bug tracking system. Then the approach mines unexpected rules that are similar to major rules and that have larger correction effort than major rules. This paper shows results of a case study of a bug tracking system in a commercial software development.