Computational intelligence and safe reduction of test suite

Systems are frequently regression tested during the maintenance phase due to corrective, preventive, adaptive or perfective actions. Regression testing is used to prevent the undesirable effects of these changes on the previously tested version. Due to these changes, new test cases become part of the test suite making it huge and inefficient for `retest all´ strategy. The ultimate solution of this problem is optimization or reduction of the test suite. Computational Intelligence (CI) based approaches like evolutionary computation, fuzzy logic, neural networks and swarm optimization have been used for test suite reduction. Optimization approaches reduce the test suite by compromising its safety. Ideally optimization of test suite must guarantee safe reduction. In this work, we have optimized the test suite using some CI based approaches and then analysed the test suite for `safe reduction´. Safe reduction can be gauged using control flow graphs. Test cases of optimal solutions were traversed on these graphs. We found that these solutions partially cover control flow graph. This showed that optimal solutions returned by CI based approaches except fuzzy logic are not safe and will be inadequate for regression testing.