Fuzzy rule-based variable neighborhood search algorithm for single-machine weighted earliness/tardiness scheduling with common due date

Fuzzy rule-based variable neighborhood search algorithm for single-machine weighted earliness/tardiness scheduling with common due date

연구 분야: Verification

학회: Neural Computing and Applications

We introduce a fuzzy rule-based variable neighborhood search (FVNS) algorithm for a single-machine weighted earliness/tardiness scheduling problem with common due date. The problem’s optimal sequence has some properties. We adapt these properties to our proposed neighborhood structures. Some existing variants that use reinforcement learning, tabu search and roulette wheel selection within variable neighborhood search (VNS) are compared with our proposed FVNS. Using well-known test instances for the problem, our experimental study reveals that our proposed FVNS algorithm produces better solutions in view of solution quality than other considered VNS variants. While FVNS generally presents better solutions for the problem, we explore significant differences in algorithm solution quality by examining solutions in detail. We consider parameters in the experiment one by one to elucidate these variances. We discuss our findings about considered VNS variants in the experimental study.