Improved Understanding on the Searching Behavior of NSGA-II Operators Using Run-Time Measure Metrics with Application to Water Distribution System Design Problems
In recent years, multi-objective evolutionary algorithms (MOEAs) have been widely used to handle various water resources problems. One challenge within MOEAs’ applications is a lack of understanding on how various operators alter a MOEA’s behavior to achieve its final performance (i.e., MOEAs are black-boxes to practitioners), and hence it is difficult to select the most appropriate operators to ensure the MOEA’s best performance for a given real-world problem. To address this issue, this study proposes the use of the run-time measure metrics to reveal the underlying searching behavior of the MOEA’s operators. The proposed methodology is demonstrated by the non-dominated sorting genetic algorithm II (NSGA-II, a widely used MOEA in water resources) with five commonly used crossover operators applied to six water distribution system design problems. Results show that the simulated binary crossover (SBX) and the simplex crossover (SPX) operators possess great ability in extending the front and finding Pareto-front solutions, respectively, while the naive crossover (NVX) strategy exhibits the overall worst performance in identifying optimal fronts. The obtained understanding on the operators’ searching behavior not only offers guidance for selecting appropriate operators for real-world water resources problems, but also builds fundamental knowledge for developing more advanced MOEAs in future.
KeywordsSearching behavior Multi-objective optimization NSGA-II Operators Water distribution system
- Deb K, Agrawal RB (2000) Simulated binary crossover for continuous search space. Comput Syst 9(3):1–15Google Scholar
- Higuchi, T., Tsutsui, S. and Yamamura, M. (2000). Theoretical analysis of simplex crossover for real-coded genetic algorithms. Proceedings of the 6th International Conference on Parallel Problem Solving from Nature, pp. 365–374Google Scholar
- Ono, I. and Kobayashi, S. (1997). A real coded genetic algorithm for function optimization using unimodal normal distributed crossover. Proceedings of the 7th International Conference on Genetic Algorithms, East Lansing, MI, USA, July 19–23, 1997, pp. 246–253Google Scholar