Mutation testing cost reduction by clustering overlapped mutants

• We defined the term conditionally-overlapped (c-overlapped) mutants.
• C-overlapped mutants are expected to produce the same results against a test case.
• Clustering c-overlapped mutants effectively reduces the cost of mutation testing.
• Clustering c-overlapped mutants retains the effectiveness of mutation testing.

Mutation testing is a powerful but computationally expensive testing technique. Several approaches have been developed to reduce the cost of mutation testing by decreasing the number of mutants to be executed; however, most of these approaches are not as effective as mutation testing which uses a full set of mutants. This paper presents a new approach for executing fewer mutants while retaining nearly the same degree of effectiveness as is produced by mutation testing using a full set of mutants. Our approach dynamically clusters expression-level weakly killed mutants that are expected to produce the same result under a test case; only one mutant from each cluster is fully executed under the test case. We implemented this approach and demonstrated that our approach efficiently reduced the cost of mutation testing without loss of effectiveness.