Short-term Traffic Flow Uncertainty Forecasting Using Modified Bootstrap Method

Short-term traffic flow prediction is one of the key technologies in the intelligent transportation system and is the basis of proactive traffic management and control.Effective short-term traffic flow uncertainty prediction can provide decision support for traffic management and control.Based on the research of scholars at home and abroad,this thesis proposes a modified Bootstrap method for short-term traffic flow uncertainty prediction,and evaluates it using actual traffic data.The specific research contents of this thesis are as follows:First,the research progress of short-term traffic flow level and uncertainty forecast,as well as Bootstrap theory and its application at home and abroad are summarized.The shortcomings of the method of using Bootstrap for short-term traffic flow prediction in the current research are analyzed,and the research ideas of this paper are determined based on this.Second,a modified Bootstrap method is proposed.Based on the classic Bootstrap theory,a modified Bootstrap method which named as Transform Bootstrap(TB),is proposed for the seasonal traffic flow data.During the resampling stage,the seasonal sample data is transformed in this method to eliminate its seasonality,and then the residual data at the same time point in different period is used as the original sample for Bootstrap sampling,statistical estimates such as sample mean and variance and confidence intervals can be then further carried out.Then,the uncertainty prediction model is developed.The modeling process of the time series prediction model is described.Based on the seasonal characteristics of the short-term traffic flow data,the Box-Jenkins model is used to establish the SARIMA model for the shortterm traffic flow data,and the maximum likelihood method is used to estimate the model parameters.On this basis,the TB method and the time series method are combined to construct the SARIMA+TB short-term traffic flow uncertainty prediction model.Finally,the effectiveness and improvement of the method is demonstrated.Using the traffic flow data collected from the British highway system and the Minnesota metropolitan highway system as experimental data,the above combined forecasting model is verified.Mean Absolute Error(MAE),Mean Absolute Percentage Error(MAPE),Root Mean Square Error(RMSE)and Root-Mean-Square Error of Prediction(RMSEP)are used to evaluate the performance of the mean prediction;Kickoff Percentage(KP)and Average Interval Width(Width)are used to measure the effectiveness of the proposed method.In addition,TB,OB(Original Bootstrap),BB(Block Bootstrap)and GARCH model are compared in three aspects:continuous traffic flow throughout the year,working days/non-working days,and different time periods within a day.The results show that the method proposed in this thesis can obtain effective interval prediction results.