Estimation Method of Order Quantity of Manufacturer Based on POS Data by Using LSTM Model

Abstract

Point-of-Sales (POS) data measures how much product the end-users purchase. Using POS data in a retail manufacture setting helps the manufacturer to understand the demand of its product and to improve its demand forecast. On the other hand, the traditional Recurrent Neural Network (RNN) is one of the recursive neural network approaches that can be applied for modeling of sequential data. The key feature of RNN is the network delay recursion, which enables it to describe the dynamic performance of systems. A Long Short-Term Memory (LSTM) model approximate the long-term information with significant delays by expanding the conventional RNN algorithm. <IFEOL> An estimation method of the order quantity of the manufacturer is proposed based on the POS data of the retailer by using the LSTM model in this research. Supply chain simulation model have been constructed to generate the POS data of the retailer and the demand quantity of the manufacturer. Some case studies have been carried out to verify the effectiveness of the proposed LSTM model by using the POS data and the demand quantity obtained by the simulation model.