Actionable Insights from Developer Behavior: A Practical Approach to Software Defect Prediction

Abstract Software defect prediction using code metrics has been extensively researched over the past five decades. However, prediction using non-software metrics remains under-researched. Given that the root cause of software defectsis often attributed to human error, human factors theory might offer key forecasting metrics for actionable insights. This paper explores automated software defect prediction based on developers' coding habits. First, we proposea framework for deciding which metrics to use for prediction. Next, we comparethe performance of our metrics to the state-of-the-art code metrics. We then analyze the predictive importance of each metric. Our analysis of twenty-one critical infrastructure, large-scale, open-source software projects yielded the following results: 1) We have presented a human-error-based framework with metrics useful for defect prediction at the method level. 2) Models using our proposed metrics achieved a better average prediction performance than state-of-the-art code metrics and history measures. 3) The predictive importanceof all metrics distributes differently, with each of the novel metrics having abetter average importance than code and history metrics. 4) Our novel metrics dramatically enhance the explainability, practicality, and actionability of software defect prediction models, which significantly advances the field.We present a systematic approach to forecasting defect-prone software methodsvia a human error framework. This work empowers practitioners to act on predictions, empirically demonstrating how developer coding habits contribute todefects in software systems.