Journal Article

CodeSense: A Static Analysis Tool for Detecting Code Smells

Abu-Alsaad, Hiba A., Alabbasi, Hesham Adnan, Ghanim, Mohanad Ridha

2026

Overview

Code smells are indicators of the design problems that affect the quality and maintainability of the software. Traditional tools that rely on static analysis can detect rule-based violations, but not complex ones. Alternative sophisticated tools that integrate machine learning (ML) can be computationally intensive and need to develop a model for a specific code smell. This paper proposes a new static analysis tool, called CodeSense, to address these challenges, describing the design, concept, implementation and evaluation of this tool. CodeSense integrates program analysis techniques with a unified detection approach, with a goal of achieving context-aware detection of code smells. It optimizes the current tools by improving the detection of design problems in the early stages, enhancing the software quality and decreasing the requirement for an extensive refactoring process. The empirical results for CodeSense show competitive results in accurately detecting code smells; it achieves higher F1-scores (0.87) for all the code smells compared to the baseline tools: SonarQube, Programming Mistake Detector (PMD) and Checkstyle, which are 0.75, 0.71 and 0.66. Our research demonstrates the need for more intelligent and integrated static analysis tools to meet the demands of today's software development.

• Publisher: International Information and Engineering Technology Association (IIETA)
• Subject: Accuracy / Algorithms / Artificial intelligence / Coding standards / Computer science / Large language models / Machine learning / Maintainability / Neural networks / Software development / Software engineering / Software quality / Source code / Static code analysis