Multi-objective analysis of the inventory planning problem using particle swarm optimization

Traditional inventory models only involve single objective that relates to several cost items or service requirements. Even in its multi-objective formulation, most models have been solved with traditional optimization techniques by combining several objectives into a single one. The solutions obtained are unsatisfactory because their non-dominance is not guaranteed. This paper incorporates a local search and clustering mechanism into the multi-objective particle swarm optimization (MOPSO) algorithm to solve two bi-objective inventory planning models, both having a cost minimization objective along with the stockout occasions minimization objective (named as N-model) and the number of items stocked out minimization objective (named as B-model), respectively. The way of multi-objective analysis can determine the non-dominated solutions of order size and safety factor simultaneously. We compare the set coverage metric of both non-dominated solution sets by their expected relevant cost and the service level per order cycle. Our results show that even under the service level measure favorable to the N-model, the non-dominated solution set of the B-model are closer to the Pareto-optimal front than that of the N-model.