Smart Prediction for Test Smell Refactorings

Software plays a critical role in modern society, making software testing essential for ensuring quality. Poor testing practices, known as test smells, reduce test code quality and maintainability. Although refactoring is a widely used technique to address these issues, little is known about when developers choose to refactor test smells and whether these actions actually improve test code quality. This study introduces a machine learning approach to guide test smell refactoring. First, we aim to mine refactorings performed by developers to derive a catalog of test-specific refactorings. Findings show testing framework evolution helps address test smells. Second, we aim to understand whether developers target low-quality test codes to perform refactorings and their effects on code quality. Our findings reveal that developers tend to refactor structurally low-quality test code more often. Third, we aim to learn whether developers would perform refactorings and which refactorings they would apply to improve the test code quality. Results show that Support Vector Machine models predicted refactoring decisions with 30–100% accuracy.