Date of Award
Doctor of Philosophy (PhD)
College of Computing and Engineering
Francisco J. Mitropoulos
Michael J. Laszlo
machine learning, mutation testing, parallel processing, reinforcement learning, software engineering, software testing
Mutation testing is a type of software testing proposed in the 1970s where program statements are deliberately changed to introduce simple errors so that test cases can be validated to determine if they can detect the errors. The goal of mutation testing was to reduce complex program errors by preventing the related simple errors. Test cases are executed against the mutant code to determine if one fails, detects the error and ensures the program is correct. One major issue with this type of testing was it became intensive computationally to generate and test all possible mutations for complex programs.
This dissertation used machine learning for the selection of mutation operators that reduced the computational cost of testing and improved test suite effectiveness. The goals were to produce mutations that were more resistant to test cases, improve test case evaluation, validate then improve the test suite’s effectiveness, realize cost reductions by generating fewer mutations for testing and improving software reliability by detecting more errors. To accomplish these goals, experiments were conducted using sample programs to determine how well the reinforcement learning based algorithm performed with one live mutation, multiple live mutations and no live mutations. The experiments, measured by mutation score, were used to update the algorithm and improved accuracy for predictions. The performance was then evaluated on multiple processor computers.
One key result from this research was the development of a reinforcement algorithm to identify mutation operator combinations that resulted in live mutants. During experimentation, the reinforcement learning algorithm identified the optimal mutation operator selections for various programs and test suite scenarios, as well as determined that by using parallel processing and multiple cores the reinforcement learning process for mutation operator selection was practical. With reinforcement learning the mutation operators utilized were reduced by 50 – 100%.In conclusion, these improvements created a ‘live’ mutation testing process that evaluated various mutation operators and generated mutants to perform real-time mutation testing while dynamically prioritizing mutation operator recommendations. This has enhanced the software developer’s ability to improve testing processes. The contributions of this paper’s research supported the shift-left testing approach, where testing is performed earlier in the software development cycle when error resolution is less costly.
Michael Allen Stewart. 2021. Increasing Software Reliability using Mutation Testing and Machine Learning. Doctoral dissertation. Nova Southeastern University. Retrieved from NSUWorks, College of Computing and Engineering. (1156)