Comparative Analysis of Selection Hyper-Heuristics for Real-World Multi-Objective Optimization Problems

As exact algorithms are unfeasible to solve real optimization problems, due their computational complexity, meta-heuristics usually used them. However, choosing a meta-heuristic particular problem is non-trivial task, and often requires time-consuming trial error process. Hyper-heuristics, which heuristics choose heuristics, have been proposed as means both simplify improve algorithm selection or configuration for problems. This paper novel presents cross-domain evaluation multi-objective optimization: we investigate how four state-of-the-art online hyper-heuristics with different characteristics perform in order find solutions eighteen real-world These were designed previous studies tackle the from perspectives: Election-Based, based on Reinforcement Learning mathematical function. All studied control set of five Multi-Objective Evolutionary Algorithms (MOEAs) Low-Level (meta-)Heuristics (LLHs) while finding problem. To our knowledge, this work first deal conjointly following issues: (i) instead simple operators (ii) focus (iii) experiments world problems not just function benchmarks. In experiments, computed, each execution, Hypervolume IGD+ compared results considering Kruskal–Wallis statistical test. Furthermore, ranked all tested three Friedman Rankings summarize analysis. Our showed that better performance than single meta-heuristics, makes them excellent candidates solving new