Advancing Automated Software Testing : An Analysis of AI-Generated Test Cases

Abstract

This thesis explores the use of artificial intelligence as part of software testing. The study delves into the generation of unit test cases based on verbal requirement descriptions. The development of AI in recent years has opened up new possibilities to make testing more efficient as a process. Testing as an activity consumes a significant part of software development resources. The objective of this thesis is to explore how AI can be used in the software development testing process. There are many steps in software testing, which require a considerable amount of human labour, which is why there is still room for automation. AI opens up new opportunities in software development to reduce human effort and achieve a higher degree of automation. In this thesis, data was collected by generating unit tests and analysing the final results. The tests were generated using the GPT-4 model. In this study, generation was performed in two rounds, the first round selecting the best performing prompt technique and the second round generating the final results with a larger amount of data. The study found that AI demonstrated the ability to understand verbal descriptions and generate unit tests based on them. However, it should be noted that full automation could not be achieved and there are still significant risks associated with the use of the generated tests.