| Multi-objective test case prioritization (MOTCP) is one of the important research dimensions in software testing where test case execution order is identified based on multi-objective. All non-dominated solutions consist of optimal solution set that is called Pareto Front. Ideally, a non-dominated solution generated by multi-objective evolutionary algorithms is gradually close to, and ultimately reaches Pareto Front. But it is not realistic that the multi-objective evolutionary algorithm could always find the Pareto Front. Ant Colony Optimization(ACO) is a kind of swarm intelligence heuristic optimization algorithm, which often has low convergence and easily fall into local optimum when applied in MOTCP. In biology, Epistatic-domain effect is an interaction phenomenon among genes in genetics, which means one gene could suppress another's expression. MOTCP, a complicated problem, also has epistatic-domain effect, i.e., test optimization goals in regression testing are dependent on different test cases in a sequence. The front expression in a sequence could reduce or even prevent the latter for testing goals. Recent research reveals that fitness values are usually determined by the test cases segment in front of the sequence, which is named as epistatic-damain test case segment (ETS). Thus, the relationship between test cases is called epistatic-domain effect in test case prioritization (TCP).We propose a multi-objective ant colony optimization algorithm which combines with epistatic-domain effect. We firstly construct a pheromone update strategy, and then optimize the ant colony algorithm. In the scheme, ETS existed in the test case sequence is selected as a pheromone updating scope, which is able to determine fitness value. In addition, analyzing fitness value increments between test cases and execution time of test cases in ETS could be used to update pheromone on trails. In order to further improving efficiency of ACO and reducing time cost when ants visit test cases one by one, the end of ants'visiting is reset by estimating the length of ETS. The experimental results show that compared with the original ACO and NSGA-?, the optimized MOACO has faster convergence and obtains better Pareto front in MOTCP. |
PDF Full Text Request