MiTS: A New Approach of Tabu Search for Constructing Mixed Covering Arrays

Abstract

Software systems have been increasingly used by our society, so a failure in them can lead to large losses. To reduce the failures of a software it is necessary to carry out the testing process appropriately. The combinatorial testing helps in the testing process by providing structures with a test set of small size, like Mixed Covering Arrays (MCAs). However, the problem of constructing an optimal test set is an NP-complete problem leading to the development of non exhaustive approaches to solve it. This paper proposes a new approach of Tabu Search (TS) called MiTS (that stands for Mixed Tabu Search) which focuses on constructing MCAs. The approach is based on the use of a mixture of neighborhood functions and a fine tuning process to improve the performance of the TS. Experimental evidence shows a poor performance when a single neighborhood function is used. In the other side, the TS (using a mixture of neighborhood functions) is competitive in the construction of MCAs over a known benchmark reported in the literature.