A Hybrid Metaheuristic Guided by Frequent Patterns for the One-Dimensional Bin Packing Problem

One-dimensional bin packing (1D-BPP) seeks to pack items into the smallest number of fixed-capacity bins, supporting decisions on efficient resource use. We propose a metaheuristic that extends the Max–Min Ant System proposed by \citet{stutzle1998improvements}, combining frequent patterns extracted from elite solutions, local search, and final intensification via local branching. On 1,587 instances from BPPLIB, the average success rate, root mean squared error, and standard deviation, computed as unweighted averages across the eight sets, were 82\%, 0.20, and 0.02, respectively. In the comparison with the literature, restricted to solution quality, the method achieved the highest reported success rate, either outright or tied, on six sets.