Research On Short-term Traffic Flow Prediction Method Based On Weather Factors

With the rapid development of road traffic all over the world,people’s travel demand has increased greatly,and the consequent traffic congestion problem is becoming more and more severe,which has become an important issue that needs to be solved all over the world.Intel igent Transportation Systems(ITS)is an important approach to solve traffic problems in the world.Short-term traffic flow prediction,as one of the key technologies in ITS,aims to use historical data to predict the future shortterm traffic flow state.For the realization of efficient,accurate,real-time traffic management and control provides an important basis.In this paper,the diurnal characteristics and long-term autocorrelation of traffic flow time series were explored by using historical traffic flow data,and the effects of PCA,KPCA,RPCA and Isomap on extracting the diurnal characteristics of traffic flow time series were compared and analyzed.At the same time,R/S analysis method and detrended fluctuation analysis method are used to study the long-term autocorrelat io n of traffic flow series,and the relationship between diurnal cycle characteristics and long-term autocorrelation is discussed through Hurst index.The actual road traffic flow data are used to verify that traffic flow has diurnal cycle characteristics and long-term autocorrelation.This provides theoretical support for the construction of the prediction model based on the periodic characteristics and long-term autocorrelation of traffic flow in the short-term traffic flow prediction method.At the same time,considering the influence of rainfall on traffic flow,this paper analyzes the influence of rainfall weather on highway traffic speed and flow.Then,the characteristics of traffic speed and flow with or without accidents under differe nt rainfall intensities and the changes of traffic flow parameters under different rainfa ll intensities during congestion periods are further analyzed and compared by using the data.Through the comparison and analysis of a large number of data,the relations hip between the above parameters is proved by statistical theory analysis,which provides a theoretical basis for the analysis of traffic flow operation characteristics in adverse weather and a theoretical support for traffic flow prediction in rainy days.The above traffic flow characteristic analysis methods provide an effective tool for in-depth understanding of the dynamic characteristics of traffic flow under differe nt states,and provide an important basis and a more complete idea for the construction of short-term traffic flow prediction model under the influence of weather factors.On this basis,this paper takes the periodic characteristics of road network traffic flow,longterm time correlation,spatial correlation and the influence mechanism of weather factors on traffic flow operation as the basis,and further constructs the short-term traffic flow prediction model based on deep learning based on attention mechanism,graph neural network and gated recurrent unit.Finally,a real large-scale network traffic flow data is used to verify the proposed temporal and spatial traffic flow prediction model WGG.The results show that the proposed model has a greater improvement in the prediction performance compared with the existing models,and the proposed model has a higher prediction accuracy and practical significance.