New CPDP Framework Uses Two-Stage Classifier Selection to Improve Defect Prediction

This paper introduces a novel framework for Cross-Project Defect Prediction (CPDP) designed to mitigate performance degradation caused by distribution shifts between training and target software projects. The proposed system employs a two-stage multiple classifier system (MCS) selection process. The first stage identifies an MCS configuration that generalizes well across various training projects, aiming for diverse classifier specialization. The second stage, operating at test time, selects the most competent classifiers on a module-by-module basis within the target project, enhancing robustness to distribution changes. Experimental results on 82 projects across four benchmark datasets indicate that this module-level selection approach outperforms existing state-of-the-art CPDP methods in many scenarios. AI

IMPACT Improves accuracy in software defect prediction by addressing distribution shifts between training and target projects.

RANK_REASON The cluster contains a research paper detailing a new method for defect prediction in software engineering. [lever_c_demoted from research: ic=1 ai=0.7]

Read on arXiv cs.LG →

• alphaXiv
• arXiv
• CatalyzeX
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• Multiple Classifier Systems
• Multi-stage Dynamic Selection
• ScienceCast
• software engineering

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →