Conference paper Element-Based Test Suite Reduction for SARSA-Generated Test Suites
Abstract
The widespread use of Android apps motivates exploration of efficient testing approaches that improve quality while addressing real-world time constraints and budgets for testing. Automated test generation with Reinforcement Learning (RL) algorithms have shown promise, but there is room for improvement as these algorithms often produce test suites with redundant coverage. Fine tuning RL algorithms is one possible solution but time-consuming due to complex characteristics of software under test. In this study, we employ a hybrid methodology to address the problem at a more general level. The hybrid methodology takes test suites generated by reinforcement learning as an input and applies test suite reduction to remove redundancy. The proposed algorithm utilizes a greedy approach to quickly reduce test suites generated by SARSA based on elements. Outcomes show a significant reduction ranging from 25.61% to 65.78% while maintaining a high level of code coverage with a loss of 0.69% at most.