Inventory Cost Control Model for Fresh Product Retailers Based on DQN

Aiming at the cost control problem of fresh product retailers in e-commerce, this paper fully considers the retailer’s inventory limitation factors and the realistic factors such as corruption rate, overdue cost and shortage cost cannot be ignored in cost control. By designing reinforcement learning quaternion (state, action, state transition, reward), this paper constructed a cost control model for fresh product inventory based on reinforcement learning. In the proposed model, Q-learning algorithm and DQN algorithm are used to train the ordering strategy and optimize the parameters. The simulation experiments show that learned by the proposed can not only effectively reduce the fresh product spoilage rate and total inventory cost when the demand distribution, product life cycle and lead time are known, but also solve the dimensionality disaster problem that the Q-learning ordering strategy model cannot solve. Therefore, the model has strong application value and wider applicability.