Improved Equilibrium Optimizer for Short-Term Traffic Flow Prediction

• Published in: Journal of database management Vol. 34; no. 1; pp. 1 - 20
• Main Authors: Pan, Jeng-Shyang, Hu, Pei, Pan, Tien-Szu, Chu, Shu-Chuan
• Format: Journal Article
• Language: English
• Published: Hershey IGI Global 01.01.2023
• Subjects: Algorithms; Back propagation; Data base management; Heuristic methods; Highway intersections; Machine learning; Microbalances; Parameters; Prediction models; Traffic flow; China
• ISSN: 1063-8016; 1533-8010; 1533-8010
• DOI: 10.4018/JDM.321758

Meta-heuristic algorithms have been widely used in deep learning. A hybrid algorithm EO-GWO is proposed to train the parameters of long short-term memory (LSTM), which greatly balances the abilities of exploration and exploitation. It utilizes the grey wolf optimizer (GWO) to further search the optimal solutions acquired by equilibrium optimizer (EO) and does not add extra evaluation of objective function. The short-term prediction of traffic flow has the characteristics of high non-linearity and uncertainty and has a strong correlation with time. This paper adopts the structure of LSTM and EO-GWO to implement the prediction, and the hyper parameters of the LSTM are optimized by EO-GWO to transcend the problems of backpropagation. Experiments show that the algorithm has achieved wonderful results in the accuracy and computation time of the three prediction models in the highway intersection.