FRAMEWORK FOR AUTOMATED SOFTWARE TESTING USING MACHINE LEARNING AND ARTIFICIAL INTELLIGENCE

Abstract

The increasing nature of contemporary software systems has posed a considerable challenge in the quality, dependability and performance of such systems. In this paper, it is proposed, as a hypothesis, that a framework can be developed, where it will be proposed, based on Machine Learning (ML) and Artificial Intelligence (AI), that by using features of predictive analytics/pattern recognition/intelligent decision-making it is possible to increase the quality of test cases generated, defect detection, and optimization of the execution conditions. The framework combines supervised learning architectures such as predicting fault-prone modules, reinforcement learning to support the prioritization of test cases and natural language processing (NLP) to support automated requirement-to-test mapping. Experimental research and simulations have shown that testing using AI technology saves test execution time, fault detecting accuracy and regression testing performance is more effective. However, there are also practical constraints such as the initial cost of model training is high, it requires large high-quality datasets and model interpretability is challenging. Further studies are suggested to consider explainable AI (XAI) in their transparency testing decisions, a human-AI testing partnership, and transfer learning, in an attempt to lessen depending on a database, a more realistic AI-driven testing approach and become easier to implement or scale up to various software settings.