ARFT-Transformer: Modeling Metric Dependencies for Cross-Project Aging-Related Bug Prediction

2026-01-01

Journal of Systems and Software

AI Summary

This study addresses the challenges of cross-project aging-related defect prediction, where distribution discrepancies between source and target domains and severe class imbalance hinder performance, and existing approaches often overlook inter-metric dependencies. To tackle these issues, this work proposes the ARFT-Transformer framework, which introduces a metric-level multi-head attention mechanism—the first of its kind in this task—to explicitly model dependencies among software metrics. Additionally, Focal Loss is integrated to enhance learning from hard-to-classify samples. Extensive experiments on three large-scale open-source projects demonstrate that ARFT-Transformer achieves significant improvements over state-of-the-art methods, with Balance scores increasing by 29.54% in single-source and 19.92% in multi-source scenarios, respectively.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Multiagent Systems: Adversarial Agents
• Search and Optimization: Metareasoning and Metaheuristics

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Problem

• cross-project
• Aging-Related Bugs
• class imbalance
• domain adaptation
• metric dependencies

Innovation

• Transformer
• metric dependencies
• cross-project prediction
• Focal Loss
• software aging