AbstractMining software repositories (MSRs) such as source control repositories, bug repositories, deployment logs, and code repositories provide useful patterns for practitioners. Instead of using these repositories as record‐keeping ones, we need to transform them into active repositories that can guide the decision processes inside the company. By MSRs with several data mining algorithms, effective software fault prediction models can be built and error‐prone modules can be detected prior to the testing phase. We discuss numerous real‐world challenges in building accurate fault prediction models and present some solutions to these challenges. © 2012 Wiley Periodicals, Inc.This article is categorized under: Application Areas > Science and Technology