Abstract
With the rise of “cross-border-e-commerce”, the third-party-forwarding-logistics (3PFL) service becomes increasingly popular. Different from the traditional third-party-logistics (3PL) service, the 3PFL company provides forwarding services cost-effectively by consolidating orders from different e-tailers/platforms. The random arrivals of orders create a big challenge. Different from most of the existing studies, a deep learning based one-step integration optimal decision making approach S2SCL(Seq2Seq based CNN-LSTM) is proposed in this paper which intelligently integrates inventory optimization and demand-forecasting process. The Seq2Seq based forecasting architecture, which integrates CNN and LSTM network, is able to model the system dynamics and dependency-relations in varying demand for logistics services. Besides generating the point forecasting results, the proposed approach can quantify demand uncertainty via a dynamic distribution and make optimal decision on logistics service capacity allocation. Through a case-study analysis with real data obtained from a 3PFL company in China’s Great Bay Area, we compare the proposed S2SCL with two benchmark models, including a one-step statistics based integration approach ARIMA and a two-step optimization based approach PSO-ELM, for two tasks: (1) point forecasting and (2) optimal logistic service capacity (LSC) allocation. Experimental results show that S2SCL outperforms the two benchmark models in both tasks significantly.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call