Machine Learning-Based Analysis of Code Smells at Class Level and Method Level

Abstract

In modern software development, maintaining high-quality code is essential for ensuring system robustness, maintainability, and long-term viability. Code smells, which are indicative of below standard alternatives of software design or potential vulnerabilities, can negatively impact software reliability and developer productivity. This paper confers a comprehensive analysis of code smells in both the aspect of class and methods. So, this study focuses on creating a dataset of code smells at different levels of object-oriented paradigm such as at method level and class level in fourteen Java applications and applies several algorithms based on machine learning, specifically J48, JRip, random forest, and Naive Bayes to examine them. The results show that there are several significant code smells which needs to be explored for better software development. The outcomes in this study are quite promising and will pave the way for researchers working in this domain. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.