Training Machine Learning Models for Software Defect Prediction in Agile Development

EasyChair Preprint 13186

Abstract

In Agile software development, the rapid pace of iteration demands efficient identification and mitigation of defects to ensure product quality. Machine learning (ML) techniques offer promising avenues for defect prediction, aiding Agile teams in preemptively addressing potential issues. This abstract explores the process of training ML models for software defect prediction within Agile frameworks.

First, it elucidates the significance of defect prediction in Agile environments, where the continuous integration and delivery cycles necessitate proactive defect management. It highlights the challenges posed by the dynamic nature of Agile projects, including frequent code changes and evolving requirements, which underscore the need for adaptable prediction models.

Next, the abstract delves into the foundational principles of ML model training for defect prediction. It discusses the importance of feature selection, emphasizing the relevance of both static code metrics and dynamic project data. It also addresses the pivotal role of dataset preparation, including data cleaning, normalization, and balancing techniques to enhance model performance and generalizability.

Keyphrases: defect, prediction, software

@booklet{EasyChair:13186,
  author    = {Louis Frank and Saleh Mohamed},
  title     = {Training Machine Learning Models for Software Defect Prediction in Agile Development},
  howpublished = {EasyChair Preprint 13186},
  year      = {EasyChair, 2024}}