JIT-CF: Integrating contrastive learning with feature fusion for enhanced just-in-time defect prediction

Xiaolin Ju, Yi Cao, Xiang Chen, Lina Gong, Vaskar Chakma, Xin Zhou

Abstract: Highlights•JIT-CF enhances defect prediction using contrastive learning and CodeBERT.•A two-layer network with ReLU activation boosts model performance and robustness.•JIT-CF achieves substantial improvements over existing state-of-the-art methods on JIT-Defects4J.