A Hybrid Genetic Algorithm for the One-Dimensional Minimax Bin-Packing Problem with Assignment Constraints

In this paper, the one-dimensional minimax bin-packing problem with assignment constraints is studied. Among other applications, this problem is used in test-splitting, which consists in assigning several sets of questions into different questionnaires so that every one of these questionnaires contains one question from each one of the original sets. Questions have a weight associated, which typically corresponds to a measure of their difficulty, and the objective is to split the questions among the questionnaires in such a way that the weights are distributed as evenly as possible. We propose a hybrid genetic algorithm for solving this problem, which is then tested on a benchmark set of practically-sized instances. The results show its efficiency in solving large size instances from the literature.