Genetic Algorithm for Supply Planning Optimization under Uncertain Demand

Abstract

Supply planning optimization is one of the most important issues for manufacturers and distributors. Supply is planned to meet the future demand. Under the uncertainty involved in demand forecasting, profit is maximized and risk is minimized. In order to simulate the uncertainty and evaluate the profit and risk, we introduced Monte Carlo simulation. The fitness function of GA used the statistics of the simulation. The supply planning problems are multi-objective, thus there are several Pareto optimal solutions from high-risk and high-profit to low-risk and low-profit. Those solutions are very helpful as alternatives for decision-makers. For the purpose of providing such alternatives, a multi-objective genetic algorithm was employed. In practice, it is important to obtain good enough solutions in an acceptable time. So as to search the solutions in a short time, we propose Boundary Initialization which initializes population on the boundary of constrained space. The initialization makes the search efficient. The approach was tested on the supply planning data of an electric appliances manufacturer, and has achieved a remarkable result.