SmellyBot : An AI ‐Powered Software Bot for Code Smell Detection
Amal Alazba, Hamoud Aljamaan, Mohammad Alshayeb ABSTRACT
Context
Automating code smell detection through a software bot offers significant benefits in terms of efficiency, code quality, and developer productivity. However, careful consideration of accuracy, integration, and user acceptance is also required to realize these benefits fully. While many previous studies have proposed deep learning models for this purpose, they often lack in automating their methodologies.
Objective
In this paper, we propose the design, development, and deployment of SmellyBot, an AI‐powered bot for code smell detection.
Method
We seamlessly integrated SmellyBot within the GitHub framework to detect four code smells, incorporating automated reporting to enhance code smell detection.
Results
Our evaluation involved 43 developers to assess user perception and feature recommendations, alongside an analysis of SmellyBot's performance on six real‐world projects to examine its efficiency and effectiveness.
Conclusion
The results indicate that developers perceive SmellyBot as highly useful and easy to use, and it has demonstrated notable efficiency and effectiveness in detecting code smells.