A Situation Deduction Method for Rail Transit Network Based on Dynamic Bayesian Network with Gaussian Mixture Model

Zhengyang Zhang,Wei Dong,Xinya Sun,Yindong Ji

DOI: https://doi.org/10.1109/icceic60201.2023.10426690

2023-01-01

Abstract:In the past work, we proposed a transport capacity risk prediction method based on linear Gaussian Bayesian network. Aiming at the problem that the global capacity risk index is single and the existing methods are difficult to deal with dynamic and nonlinear data, this paper proposes a method of transportation safety situation deduction for rail transit network based on dynamic Bayesian network and Gaussian mixture model, combined with the situation awareness theory. The hierarchical transport capacity risk is used as the core to construct the network safety situation index system. According to the network topology and the interaction between the transport capacity risk and the single point risk, the dynamic Bayesian network is constructed. The EM algorithm is used to estimate the parameters of the Gaussian mixture model to realize the Bayesian network parameter learning. This method is verified by the example of Chongqing rail transit network, and the prediction performance outperforms other comparison methods.