A Systematic Review of Learning-Based Software Defect Prediction

Abstract

This survey paper on Learning-Based Software Defect Prediction reviews recent advancements in applying machine learning (ML) and deep learning (DL) techniques for software defect prediction (SDP). It covers topics including application scenarios, types of ML/DL, datasets, source code representation, prediction granularity, evaluation metrics, validation methods, and challenges with existing solutions. The paper highlights the increased interest in SDP due to the growing complexity of software and presents a detailed analysis of various ML and DL approaches, their capabilities, and the challenges they present in the context of SDP. It also discusses the evolving nature of these technologies in SDP and their impact on developing reliable and maintainable software systems.