Association Rule Mining for Re-planning Decisions of Software Product Releases

Well-planned projects might end unsatisfactorily if changes in software release plan are not controlled and managed. While process control detects potentially risky changes, re-planning adjusts a current plan to changing reality. Association rule mining can be used to discover interesting relations like associations, correlations etc. among different attributes of large databases. We investigate and evaluate how association rule mining can be used to influence re-planning decisions of software product releases utilizing extracted information (i.e. rules and process factors) from legacy data. Proposed approach Dyna-H2W*, encompasses association rule mining based methods to find multiple process factors and associated rules responsible for project success or failure. Inspired by statistical process control, these rules are continuously monitored throughout the project. Any rule violation triggers a warning followed by an expert-based investigation for root cause analysis behind the warning. Investigation might suggest initiating re-planning as a reactive action. This paper exclusively focuses on empirical analysis of the proposed approach. Two controlled experiments along with three case study designs are presented in this paper. Preliminary experiments and comparative analysis against existing re-planning techniques suggest Dyna-H2W* is capable of generating product of higher value and quality. Additionally proposed approach increases stakeholders’ level of trust on the re-planning method.