Short-term Traffic Flow Prediction Based On Multidimensional Spatio-temporal NPCA-PSR-IGM Combination Model

Real-time access to urban road traffic flow,the effective analysis of urban road space and time correlation,accurately predict future urban road traffic flow,is to realize the urban road traffic control and guidance of the premise and key,to ease the urban road traffic congestion has important theoretical significance,but also has a very broad application value.In this thesis,on the basis of the existing work,based on the spacetime characteristics of short-time traffic flow has strong,chaos and nonlinear characteristics,by using the nonlinear principal component analysis,phase space reconstruction theory and the improved grey model prediction algorithm,to achieve the real-time accurate prediction of traffic flow,specific work includes the following aspects:(1)In view of the interaction and influence of traffic flow data in time and space,a nonlinear principal component analysis algorithm based on data correlation is proposed.In the traffic network,the short-term traffic flow prediction is not only related to the single point traffic flow of predicted sections,but also affected by the traffic flow of nearby sections.Considering the nonlinearity and complexity of urban traffic,the correlation coefficient of sequence data was used to remove the traffic flow series with low correlation coefficient with the predicted road sections.In the time dimension,nonlinear principal component analysis is used to eliminate the traffic flow data with low correlation and retain the traffic flow data with high correlation.At the same time,the error caused by nonlinear data in the sequence to short-term traffic flow prediction is eliminated,which improves the accuracy of prediction.(2)According to the nonlinear and chaotic characteristics of the traffic flow series,based on the phase space reconstruction of the multi-dimensional spatial and temporal traffic flow series,an improved grey model prediction algorithm based on the background value is proposed.Combined with single variable time series phase space reconstruction to fully amplify internal microscopic features of sequence,but the reconstruction of the single sequence is not very accurate to describe the evolution law of traffic system state variables,multi-dimensional space-time traffic sequence is used to reconstruct phase space,make it on the basis of the original system to establish a more rich and complete information system,improve the quality of prediction.Then,the prediction accuracy of the conventional grey model is improved by the improvement of the background value,considering that the grey model prediction algorithm is suitable for non-linearity and requires less data.(3)The traffic flow data near the intersection of Wuhu Road and Huizhou Avenue in Hefei were collected as samples to establish an improved grey(NPCA-PSR-IGM(1,1))combined prediction model based on multi-dimensional spatio-temporal nonlinear principal component analysis and phase space reconstruction,and the actual traffic flow in the collected sections was taken as model samples.The experimental results show that the average relative error of the NPCA-PSR-IGM(1,1)combined prediction model is 3.21% lower than that of the NPCA-PSR-GM(1,1)combined prediction model.The standard deviation relative to the PCA-PSR-IGM(1,1)combined prediction model decreased from 15.7091 to 2.0589.At the same time,compared with some existing forecasting models,the combined forecasting model has achieved better expected results in improving the accuracy of prediction.