Traffic flow prediction for highway vehicle detectors through decomposition and machine learning
Wanlian Lu, Yao Hu, Wangyong Chen, Yutao Qin, Chuliang Wu, Xinyi He
Abstract
Traffic flow prediction is of significant importance in traffic planning. Currently, traffic flow data are primarily collected through loop detectors. However, factors such as road conditions can affect the accuracy of these data. To address this issue, this paper proposes a traffic flow prediction method based on decomposition and machine learning. The improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) method decomposes the sequence into multiple intrinsic mode functions (IMFs). The complexity of each IMF is calculated using the sample entropy (SE), and then the IMFs are reconstructed. Parameters of the variational mode decomposition (VMD) are optimized using the whale optimization algorithm (WOA) for the secondary decomposition, and predictions are made using gated recurrent units (GRU). Finally, the prediction results are reconstructed to obtain the final prediction values. In the case study section, experiments are conducted using datasets from three detectors to explore different decomposition forms and methods.