Detection of God Class and Data Class code smells based on an automatic machine learning tool

Khleel, Nasraldeen Alnor Adam and Nehéz, Károly (2025) Detection of God Class and Data Class code smells based on an automatic machine learning tool. In: Proceedings of the International Conference on Formal Methods and Foundations of Artificial Intelligence. Eszterházy Károly Katolikus Egyetem Líceum Kiadó, Eger, pp. 115-128. ISBN 9789634963035

Preview

Text fmfai2025_pp115-128.pdf - Published Version Download (744kB) | Preview

Abstract

Code smells are symptoms of poor design or incomplete implementation that can degrade software quality and maintainability. Detecting them is crucial for improving software reliability and guiding refactoring efforts. Traditional detection methods rely on predefined rules or thresholds, which are inflexible and prone to errors, while modern machine learning approaches require significant expertise and large, balanced datasets. To address these challenges, we propose an automated code smell detection method using AutoGluon, an AutoML framework that streamlines model selection, hyperparameter tuning, and handling of imbalanced datasets. To evaluate the effectiveness of the proposed method, experiments were conducted using two code smell datasets: God Class and Data Class. The performance of the method was evaluated using six different metrics: accuracy, precision, recall, F-measure, Matthew’s correlation coefficient (MCC), and the area under the receiver operating characteristic curve (AUC). Additionally, we have also compared our proposed method with stateof- the-art code smell detection methods. Experimental results show that AutoGluon achieves high predictive performance—up to 0.98 accuracy for God Class and 1.00 for Data Class, which often matches or outperforms stateof- the-art methods, demonstrating the potential of AutoGluon for efficient and scalable code smell detection.

Actions (login required)

Edit Item