A Unified Multi-Label Code Smell Dataset for Code Smell Detection at Different Granularities
TL;DR
This article introduces a unified multi-label dataset for code smell detection, addressing the limitation of single-label approaches by combining four existing datasets at method and class levels to capture co-occurring smells.
A Unified Multi-Label Code Smell Dataset for Code Smell Detection at Different Granularities
Haneen M. Alhadeaf; Mubarak Alrashoud
https://doi.org/10.1109/ACCESS.2025.3648907
Volume 14
Code smell detection is critical for maintaining software quality and enabling effective refactoring, yet much prior work identifies only one smell at a time. This single-label framing misses the real-world complexity where a code element can exhibit multiple co-occurring smells. We address this gap by creating a unified multi-label dataset that combines four existing datasets at two levels—method...