A new hybrid-heuristic for large-scale combinatorial optimization: A case of quadratic assignment problem

Abstract

Real world combinatorial optimization problems arise in very-large-scale. Solving such problems to optimality is impractical. In the past several decades, different heuristics and meta-heuristics have been proposed as solution procedure. In particular, to tackle the large-scale real applications, researchers have intelligently combined key components of competing methodologies to create superior solution procedures. In this study, we design a hybrid-heuristic (HH) using high-level solution algorithm framework that blends key components of three well-known meta-heuristics, scatter search, critical even tabu search, and genetic algorithm random keys, and applies to large-scale quadratic assignment problem (QAP). The HH has the potential to go beyond the goals of usual hybrid-heuristics, it potentially provides a framework for a self-adapted heuristic that can be applied to a wide range of problems with minimal inclusion of problem specific information. Extensive computational experiments applied to large-scale QAP instances from database of benchmark problems e.g., QAPLIB and taixxeyy collections, show that the proposed hybrid-heuristic (HH) provides an efficient approach to the large-scale QAP. The algorithm provides new best solutions for 61 out of 70 taixxeyy instances of large-scale QAP and finds the same best known solutions for QAPLIB and the 9 remaining instances from taixxeyy collections.