The thesis explores an approach to automated graphical user interface testing through the application of neural networks. The primary objective is to show that we can develop a system capable of generating test cases based on sequences of user actions, which will imitate the behavior of real users, and can automatically execute them later on. Machine learning techniques for modeling user behavior patterns to generate graphical user interface test cases offer the potential to eliminate the need for manual test creation while enhancing test coverage. By adopting such approaches, the stability of the tested software in production environments can be improved, and human resources allocated to this task can be significantly reduced. The research focuses on recurrent neural networks due to their suitability for our domain, as they are specifically designed to model interdependent sequential data.
|
