Title: Research on demand prediction of cross-border e-commerce supply chain based on stationary characteristic and improved LSTM

Authors: Bingquan Wu; Xueyu Xu; Yatian Yang

Addresses: School of Economics and Management, Quanzhou University of Information Engineering, Quanzhou, 362000, China ' College of Economics and Management, Quanzhou University of Information Engineering, Quanzhou, 362000, China ' School of Management, University of Sanya, Sanya, 572000, China

Abstract: Aiming at the problems of low accuracy and efficiency of traditional cross-border e-commerce supply chain demand prediction, a demand prediction method for cross-border e-commerce supply chain based on stationary characteristic and improved long short-term memory neural network (LSTM) is proposed. Firstly, CNN-LSTM series network is constructed. Then, ECANet attention mechanism is introduced into convolutional neural network (CNN) network, and Sparrow Search Algorithm (SSA) is adopted. Finally, combination model based on CNN-ECANet-SSA-LSTM is obtained. Experiment results show that under Amazon dataset, MAPE, RMSE and R² values of the proposed model are 0.1347, 0.2608 and 0.1591, respectively, and its prediction duration is only 17.58sl. It indicates that it is superior to the traditional prediction models. In other test sets, its prediction accuracy and prediction efficiency are both superior to those of other models. Therefore, the proposed method has higher prediction precision, and shorter prediction, which meets the real-time and accuracy requirements of demand prediction.

Keywords: cross-border e-commerce; stationary feature; supply chain demand prediction; LSTM; long short-term memory neural network; attention mechanism.

DOI: 10.1504/IJCSM.2025.151198

International Journal of Computing Science and Mathematics, 2025 Vol.22 No.3, pp.266 - 283

Received: 12 Feb 2025
Accepted: 05 Jul 2025
Published online: 16 Jan 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article