Frequency Distribution Based Hyper-Heuristic for the Bin-Packing Problem

In the paper, we investigate the pair frequency of low-level heuristics for the bin packing problem and propose a Frequency Distribution based Hyper-Heuristic (FDHH). FDHH generates the heuristic sequences based on a pair of low-level heuristics rather than an individual low-level heuristic. An existing Simulated Annealing Hyper-Heuristic (SAHH) is employed to form the pair frequencies and is extended to guide the further selection of low-level heuristics. To represent the frequency distribution, a frequency matrix is built to collect the pair frequencies while a reverse-frequency matrix is generated to avoid getting trapped into the local optima. The experimental results on the bin-packing problems show that FDHH can obtain optimal solutions on more instances than the original hyper-heuristic.