LARGE LANGUAGE MODEL-BASED AUTOMATED TESTSCRIPT GENERATION FOR CONTINUOUS SOFTWAREQUALITY ASSURANCE

Abstract

Continuous Software Quality Assurance (CSQA) has become an essential practice in modern software engineering due to the rapid adoption of Agile, DevOps, and Continuous Integration/Continuous Deployment (CI/CD) methodologies. The increasing complexity of software applications and the demand for rapid release cycles have exposed the limitations of conventional manual and rule-based automated testing approaches. Developing and maintaining test scripts manually is often labor-intensive, time-consuming, error-prone, and difficult to scale as software requirements evolve. Recent advancements in Large Language Models (LLMs) have introduced new opportunities for automating software testing by generating intelligent, context-aware, and executable test scripts directly from software requirements, user stories, source code, application programming interfaces, and functional specifications. This research proposes a Large Language Modelbased automated test script generation framework for continuous software quality assurance that integrates natural language processing, software repository analysis, prompt engineering, machine learning, DevOps pipelines, and continuous feedback mechanisms within a unified intelligent testing architecture. The proposed framework automatically analyzes software artifacts, extracts testing requirements, generates executable test scripts for multiple testing frameworks, validates generated scripts through static and dynamic analysis, and continuously improves script quality using developer feedback and execution results. Advanced LLMs are combined with retrieval-augmented generation, code analysis, and quality evaluation techniques to produce accurate, maintainable, and reusable test cases for unit testing, integration testing, regression testing, API testing, and user interface testing. Experimental evaluation demonstrates that the proposed framework achieves test script generation accuracy exceeding 97.8%, reduces script development effort by approximately 72%, decreases testing time by 60%, improves defect detection capability by 38%, and increases overall software quality assurance efficiency by nearly 45% compared with conventional automated testing approaches. The proposed framework provides a scalable, intelligent, and adaptive solution for autonomous software testing in modern DevOps and continuous delivery environments