George Washington University
Software systems require the validation of design features through regression testing. Two primary challenges in system validation are ensuring that test suites reflect actual system usage, and managing the test suite size to keep testing costs low while keeping testing results meaningful. To create a test environment that is close to actual system usage, the authors propose using Markov chains to create system behavioral models from available system usage data. Knowing that certain factors are not captured in system usage, they will use the Markov Modulated Markov Process to model hidden processes. The models are used to create test plans that employ a unique, likelihood-based, test prioritization scheme.