A Practical Failure Prediction Model based on Code Smells and Software Development Metrics

Making errors during software development is unavoidable. Developers inevitably make errors that take additional time to fix later. Consequently, efforts for bug fixing compete with implementing new features. Typically, the later bugs are found, the higher the cost for remediation. To address this concern, software testing should start as early as possible in software development lifecycle. For this purpose, static analysis is proposed, but typically shows too many findings and hence do not support development teams appropriately. So, it would be a benefit to premature detect those findings in static analysis that will result in failures to reduce subsequent efforts notably. The purpose of the paper is to analyze failure data from issue tracking systems that are correlated to findings from static analysis. Thereupon an artificial intelligence-based approach is used to train practicable models for business environment that enables effective prediction of software faults. The results from static analysis show that predefined complexity measures encompassed the most defects. While there are commonalities in relevant defect findings in static analysis reports, meaningful prediction models cannot be expected based solely on this data. In addition to the findings of the static analysis, metrics like code changes in a time period or number of authors involved in code changes were considered for building the prediction models. Two of the developed prediction models have a high accuracy and excellent utility rate. These resulting prediction models are currently used at Raiffeisen Software GmbH for a long-term study on failure prediction based on code smells.