Defect Prediction

Software defect prediction aims to identify potentially faulty code sections early in the development process, minimizing costly debugging later. Current research focuses on improving prediction accuracy using advanced techniques like hierarchical transformer models and leveraging various data sources, including code content (e.g., words, data types), code context, and even programming style. These efforts utilize both traditional machine learning algorithms (e.g., Naive Bayes, Random Forest) and more recent deep learning approaches, with a growing emphasis on utilizing pre-trained models and semi-supervised learning to address data scarcity. Ultimately, effective defect prediction can significantly reduce software development costs and improve software quality.