Title: An improved EEMD-PE-BiLSTM model for forecasting traffic flow

Authors: Liang Zhu; Wenlong Zhu; Wanli Xiang; Xuelei Meng; Chunmin Zhang

Addresses: School of Traffic and Transportation, Lanzhou Jiaotong University, Lanzhou 730070, China; Key Laboratory of Railway Industry on Plateau Railway Transportation Intelligent Management and Control, Lanzhou 73000, China ' School of Management, Anhui University, Hefei 230601, China ' School of Traffic and Transportation, Lanzhou Jiaotong University, Lanzhou 730070, China; Key Laboratory of Railway Industry on Plateau Railway Transportation Intelligent Management and Control, Lanzhou 73000, China ' School of Traffic and Transportation, Lanzhou Jiaotong University, Lanzhou 730070, China; Key Laboratory of Railway Industry on Plateau Railway Transportation Intelligent Management and Control, Lanzhou 73000, China ' School of Traffic and Transportation, Lanzhou Jiaotong University, Lanzhou 730070, China; Key Laboratory of Railway Industry on Plateau Railway Transportation Intelligent Management and Control, Lanzhou 73000, China

Abstract: Accurately predicting traffic flow is of paramount importance for enhancing road network traffic efficiency and alleviating urban traffic congestion. However, in the presence of external influences, the raw traffic flow data exhibited noticeable noise, thereby increasing prediction complexity. To address this challenge, we propose the ATMS-EEMD method. Specifically, we address the issue of mode mixing in EEMD by combining signal-energy calculations and introducing an adaptive stopping criterion. Subsequently, we employed the permutation entropy (PE) algorithm to assess the complexity of the decomposed components and reconstruct new subsequences with similar complexity. Finally, we constructed a bidirectional long short-term memory (BiLSTM) neural network for traffic flow prediction. We validated the performance of the proposed model using traffic flow data from three detectors. The experimental results demonstrate that ATMS-EEMD-PE-BiLSTM achieves an average absolute error (MAE), root mean square error (RMSE), and mean absolute percentage error (MAPE) of 1.422, 2.183, and 3.603, respectively. This method exhibits outstanding performance in traffic flow prediction, overcomes challenges posed by external factors, and achieves high-precision forecasting.

Keywords: traffic flow forecasting; ITS; ATMS-EEMD; permutation entropy; BiLSTM.

DOI: 10.1504/IJISTA.2024.143265

International Journal of Intelligent Systems Technologies and Applications, 2024 Vol.22 No.4, pp.445 - 468

Received: 14 Nov 2023
Accepted: 26 Sep 2024
Published online: 11 Dec 2024 *

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