Inventory control model for intermittent demand: a comparison of metaheuristics

In this era of trying to get more information from data, demand forecasting plays a very important role for companies. Companies implement many of their plans based on these forecasts. However, forecasting demand is not always easy. Demand is divided into many different classes depending on its structure. One of them is the problem of intermittent demand. Estimating intermittent demand is a more difficult problem than estimating series with low variability and no null demand. When working with intermittent demands, either a good demand forecast or a good inventory policy is required. Determining an efficient inventory policy is important in terms of meeting customer demand and customer satisfaction. In this study, an inventory lower bound and upper bound are calculated to balance the inventory cost of intermittent demand and the lost sale cost. For this purpose, 7 different test data with intermittent demand structure were studied. A mathematical model is proposed to calculate the cost with the upper and lower inventories under intermittent demand. A feasible solution could not be obtained with the proposed model, and a fitness function for the relevant model was proposed. This function was run on test data using genetic algorithm (GA) and particle swarm optimization (PSO). The results and solution times of GA and PSO were compared. In this way, the variability of demand and the difference between arrival times of demands were met with minimal cost.