Simulated annealing: An experimental application on pairwise testing

Abstract

Testing plays a vital role in Software industries. Of all the methods, Combinational Testing is most preferred due to its robust approach in testing each and every possible pair, triple or n-way of a configuration at least once. Most of the faults occurred through these interactions between components can be captured using pairwise Testing. Many of the existing research proposes novel methods which aims to find a test set of configurations to cover all the possible different pairs. The difficulty of solving this problem is demonstrated to be NP-Complete. In this paper we propose to deeply explore the usage of Simulated Annealing, a meta heuristic algorithm to pairwise testing that generates the least number of test configurations with the aim to cover all distinct pairs possible. Furthermore, we proposed a fitness function which is dynamic in nature. This provides us the flexibility of choice between the runtime of this algorithm and the different pairs coverage of the test configurations generated. An ideal scenario would be a combination of both (runtime and coverage), but the results show that they are inversely related.