Reduction in Mutation Testing of Java classes

In mutation analysis many simple modification of the original program called “mutants” are created. Test cases which are supposed to identify the introduced program changes are designed. Each mutant must be “killed” by a test case, i.e. the test case should detect the purposely introduced modification. Mutation testing is known to be effective but computationally demanding and time consuming because a large number of mutants has to be tested. Mutation score, which is the fraction of mutants that are killed by a test set, is often used to evaluate the effectiveness of mutation testing. An interesting research question is if the number of mutants can be reduced without significantly decreasing the effectiveness of the test. We were exploring selective reductions of mutants generated for Java programs. The results of several experiments conducted in the Eclipse environment are presented in this paper. These results show that selective reduction in mutants can significantly reduce the cost of testing with acceptable mutation score and code coverage.