The invention discloses a short-time prediction method of a road traffic flow state, and the method is used for solving the technical problem that an existing road traffic flow state prediction method is poor in precision. According to the technical scheme, at the beginning of prediction, parameters provided by expert knowledge serve as DBN network parameters, an antecedent recursive method is used for inference prediction, and traffic flow state transfer probabilities between time slices of a DSN and transfer probabilities between traffic flow states and traffic flow parameters in the time slices are obtained. When prediction times reach a set value, Dirichlet conjugate prior distribution is used as prior distribution of the probabilities, a moment estimation method is used for estimating super parameters of the prior distribution, then equivalent samples formed by prior distribution sampling and observed values are used for updating the network parameters of the DBN, and the updated network parameters are used as the new network parameters for the next time of prediction. According to the invention, the DBN parameters are adopted for self-adaptive online adjustment, the dominant phenomenon of subjective factors is avoided in the prediction process, and the prediction precision is improved.
Short-time prediction method of road traffic flow state
2016-02-03
Patent
Electronic Resource
English
IPC: | G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS |
Short-time traffic flow prediction method based on state pattern
European Patent Office | 2015
|The Urban Road Short-Term Traffic Flow Prediction Research
Trans Tech Publications | 2013
|