Test Suite Optimization Using Machine Learning Techniques: A Comprehensive Study

Abstract

Software testing is an essential yet costly phase of the software development lifecycle. While machine learning-based test suite optimization techniques have shown promise in reducing testing costs and improving fault detection, a comprehensive evaluation of their effectiveness across different environments is still lacking. This paper reviews 43 studies published between 2018 and 2023, covering various test case selection, prioritization, and reduction techniques using machine learning. The findings reveal that conventional machine learning techniques, particularly supervised learning methods, have been widely adopted for test case prioritization and selection. Recent advancements, such as deep learning and hybrid models, show potential in improving fault detection rates and scalability, though challenges remain in adapting these techniques to large-scale and dynamic environments. Additionally, Generative AI and large language models (LLMs) are emerging as promising tools for automating aspects of test case generation and prioritization, offering new avenues for future research in enhancing test suite optimization. The study identifies recent trends, challenges, and opportunities for further research, with a focus on both conventional and emerging methods, including deep learning, hybrid approaches, and Generative AI models. By systematically analyzing these techniques, this work contributes to the understanding of how machine learning and Generative AI can enhance test suite optimization and highlights future directions for improving the scalability and real-world applicability of these methods.

Keywords

• Software quality
• software testing
• test suite optimization (TSO)
• machine learning in software testing
• test case selection
• evaluation metrics for test suite optimization