A data-driven model assisted hybrid genetic algorithm for a two-dimensional shelf space allocation problem

This paper investigates a two-dimensional shelf space allocation problem (2DSSAP) in the retail field. A data-driven model assisted hybrid genetic algorithm (DMA-HGA) is proposed to address the considered problem effectively. The proposed DMA-HGA applies an improved genetic algorithm (GA) as the optimization method, capable of modifying infeasible solutions while generating new solutions to satisfy model constraints. In addition, a two-stage search assistance module is implemented to facilitate a more efficient search process. In the first stage, a data-driven model is developed and used as a surrogate model for rapid fitness measurements and filtering out inferior solutions. With the generation of new solutions, the data-driven model will gradually lose its accuracy, and the second stage thus begins, using a taboo list to facilitate an in-depth search. To validate the performance of the proposed DMA-HGA, experiments on twenty-five simulation instances from five scenarios and two real-world cases are conducted. Experimental results show that the proposed DMA-HGA yields a better solution and higher accuracy compared to considered benchmarking methods. Finally, management insights for the 2DSSAP are provided based on the extended discussion of parameters.