VS-TACO: A Tuned Version of Ant Colony Optimization for Generating Variable Strength Interaction in T-Way Testing Strategy

Abstract

Ever since, software technologies have been through a rapid evolution. In a real application, the interaction between input variables may vary, thus the exhaustive testing is no longer practical since it is time-consuming and lead to combinatorial explosion. One of the strategies that able to cater fault due to the interaction is Ant Colony Optimization (ACO) algorithm. Typically, amount of ants in the ACO algorithm is fixed at certain number while the search space technique (i.e. to explore or exploit new possible solutions) is randomized for each iteration in the entire algorithm, are potentially affect the optimization's efficiency. Thus this paper proposes a new variant of ACO algorithm called as a tuned version of ACO for generating variable strength interaction in t-way testing strategy (VS-TACO). VS-TACO applied a Mamdani fuzzy logic in order to dynamically choose the number of ant and decide which search space technique to be used. Experiments that have been conducted on VS-TACO and benchmarked with other strategies, shows VS-TACO produce a competitive result in term of test suite size.