Predicting Software Maintainability Using AI Models
TL;DR
Experiments demonstrate the effectiveness of AI-based predictions in improving software quality assessment, providing actionable insights, and supporting proactive maintenance strategies in improving software quality assessment and reducing maintenance effort.
Abstract
Software maintainability is a critical quality attribute that directly impacts the long-term cost, reliability, and evolution of software systems. Predicting maintainability early in the development lifecycle enables developers and managers to make informed design decisions, allocate resources efficiently, and reduce technical debt. This paper investigates the use of artificial intelligence (AI) models for predicting software maintainability based on code metrics, historical project data, and architectural characteristics. We explore supervised learning techniques, including regression models, decision trees, and neural networks, as well as ensemble and hybrid approaches, to estimate maintainability scores and identify key factors influencing maintainability. Experiments on open-source and industrial datasets demonstrate the effectiveness of AI-based predictions in improving software quality assessment, providing actionable insights, and supporting proactive maintenance strategies. The study highlights the potential of AI-driven methods to enhance software engineering practices and reduce maintenance effort.