An Evolutionary Algorithm and a Clustering Technique to Select Good Subsets of Test for Finite State Machines

Abstract

Testing is the technique most widely used to validate the correct behaviour of systems. Essentially, a test consists of applying an input to the system and decide whether it returns the expected output. Unfortunately, budget and temporal constraints limit the amount of testing that can be applied to the system. Therefore, a good selection of tests will reduce the resources devoted to testing while keeping an effective validation process. In this paper, we tackle this problem by using mutation testing, which effectively simulates the possible faults that the system under test may have and suggest which tests are best in finding potential faults. In order to perform test selection, we use a multi-objective genetic algorithm that focuses on two targets: minimising the number of inputs the test suite has to perform and maximising the mutation score. We have performed several experiments and exhaustively compared our proposal with a Machine Learning method, specifically clustering, which groups the tests into classes, from which we select the most suitable test to be applied.