Enhancing Test Smell Prediction with Stacking Ensemble Learning

Abstract

Test smells are symptoms of sub-optimal design choices adopted when developing test cases. Previous studies have demonstrated their harmfulness for test code maintainability and effectiveness. As a result, researchers have proposed automated, heuristic-based techniques, and machine learning algorithms to detect them. However, the performance of these detectors is still limited, such as depending on tunable thresholds, low performance. In this study, we propose an ensemble learning model, using Stacking Ensemble algorithm with cross-validation technique to enhance the accuracy of test smell prediction. The proposed model consists of two layers, in which the base-layer (base-models) uses three machine learning algorithms including XGBoosting, Random Forest, Support Vector Machine, while the meta-layer (meta-learner) uses the Logistic Regression algorithm. The experimental results show that our approach outperforms the state-of-the-art techniques in terms of accuracy.