Joint pricing and inventory management in a competitive market using reinforcement learning: a combination of the agent-based and simulation-optimization approaches

Abstract

ABSTRACT One of the most important tools is an appropriate pricing mechanism to attract more customers and increase profits. The retailers’ main question is how to set the prices and inventory policies to maximize profit in a competitive heterogeneous market in presence of non-zero lead time and lost sales. A reinforcement learning algorithm is proposed to create appropriate decision-making mechanisms for pricing. A coordinated inventory policy in a competitive environment reduces logistic costs and leads to a higher profit. We use a reinforcement learning algorithm to investigate the performance of a retailer in a competitive environment. An agent-based modeling experimental environment combined with a simulation-optimization method in which a virtual market has been reproduced is used. The market is not homogeneous with respect to customer behavior. It is assumed that the retailer uses (R, Q) policy where the lead time is a fixed amount (L), and the shortage is permissible. The quality, distance, service level, and price are factors that influence customers’ choices. The simulation results for some randomly generated examples show that the algorithm in the competitive environment can make more profit than other available methods and the combined utilization of simulation-optimization methods has been able to find better solutions for the hybrid model of pricing and inventory management considering customer behavior. The results of simulation for three different categories of customers (more sensitive to price, equally sensitive to price, quality and service level, and more sensitive to quality (indicate that the average profit for the proposed algorithm is higher than that of other examined algorithms.