Using hybrid algorithm for Pareto efficient multi-objective test suite minimisation

  • Authors:
  • Shin Yoo;Mark Harman

  • Affiliations:
  • King's College London, Strand, London WC2R 2LS, UK;King's College London, Strand, London WC2R 2LS, UK

  • Venue:
  • Journal of Systems and Software
  • Year:
  • 2010

Quantified Score

Hi-index 0.01

Visualization

Abstract

Test suite minimisation techniques seek to reduce the effort required for regression testing by selecting a subset of test suites. In previous work, the problem has been considered as a single-objective optimisation problem. However, real world regression testing can be a complex process in which multiple testing criteria and constraints are involved. This paper presents the concept of Pareto efficiency for the test suite minimisation problem. The Pareto-efficient approach is inherently capable of dealing with multiple objectives, providing the decision maker with a group of solutions that are not dominated by each other. The paper illustrates the benefits of Pareto efficient multi-objective test suite minimisation with empirical studies of two and three objective formulations, in which multiple objectives such as coverage and past fault-detection history are considered. The paper utilises a hybrid, multi-objective genetic algorithm that combines the efficient approximation of the greedy approach with the capability of population based genetic algorithm to produce higher-quality Pareto fronts.