Comparing the Adaptability of a Genetic Algorithm and an LLM-Based Framework for Automated Software Test Data Generation: In the Context of Web Applications

In the fast-paced world of software development, ensuring software quality is paramount. Software Quality Assurance (SQA) plays a vital role, primarily through testing, which can be carried out manually or automatically. Yet, creating comprehensive test data (TD) for web applications can be a formidable task. Manual test data generation (TDG) is time-consuming and error prone. Automation of TDG has become increasingly important in the realm of software quality assurance as it enables efficient and effective testing of software systems. The need for an appropriate framework for automated TDG is critical to achieve comprehensive and reliable test coverage. Automated TDG offers significant advantages, including time and resource savings, improved test coverage, and seamless integration into the software development process. The core aim of this research is to bridge the gap between manual and existing automated methods, resulting in time and cost savings, heightened testing efficiency, and elevated software quality. Research objectives encompass comparing the adaptability of an AGA based automated TDG model and a LLM based automated TDG model to a web application. The results from the LLM model for triangle classification program was found to be potentially acceptable and accurate than the AGA model's results. This research discusses the challenges encountered when implementing and using the AGA-based framework in the web application context and how an LLM model could overcome the challenges. The study highlights the benefits of using the LLM approach, demonstrating its relevance and accuracy in generating test data compared to the Genetic Algorithm-based model. The practical implications for software quality assurance practices are discussed, emphasizing the enhanced efficiency and effectiveness of the LLM model in improving software quality.