The change in time-varying factors in traffic evolution has a great impact on the data generation condi-tions of road traffic flow.To determine the influence of these factors,this paper establishes the relationship between road traffic control and traffic flow prediction data and proposes a short-term traffic prediction method based on the Bayesian network(BN)with multiple residual compensations.The lane and vehicle information of large-scale multi-intersection trunk roads in the city are extracted to construct multiple parallel BNs,and the Bayesian relation-ship and expectation maximization(EM)algorithm are used for short-term traffic prediction.Then,the prediction accuracy is improved through data autocorrelation residual compensation,vehicle lane changes,and linear residual compensation of multi-intersection connectivity,which solves the problem of insufficient processing ability of tradi-tional research on factors such as errors caused by adjacent intersections and lane change.The simulation results show that the short-term traffic flow in complex traffic evolution scenarios can be predicted more accurately using BN,compensating for the accuracy based on residual vehicle behavior.
维普
万方数据