Detecting Software Defects With Hierarchical Multilabel Classification: Insights From an Industrial Case Study

Abstract

Managing software defects effectively is a major advantage for companies that rely on service-based solutions, as it reduces risks and improves the way issues are tracked and resolved. Numerous methods have been proposed to enhance the identification, localization, and classification of software defects. When it comes to practice, we have found that defects are inherently organized in hierarchies based on class inclusion. Building on this idea, we report in this article our experience of deploying a hierarchical multilabel defects classification approach, within a development team in a banking and finance software company. Overall, we gathered over 2000 defect reports, coming from their agile reporter calculation engine. Key findings reveal that our approach not only provide better interpretability of overlapping categories but also results in significantly better performance results than traditional flat feedforward neural networks and transformers-based large language models.