Research On Short-term Traffic Flow Forecasting Method Of Urban Roads Based On Multi-factor Fusion

In recent years,intelligent transportation system as an efficient technical means has been widely used in the field of urban traffic management.Real-time and accurate traffic information can assist intelligent transportation systems to achieve effective traffic management and control.Mastering and accurately predicting traffic flow information in real time to meet the data needs of intelligent transportation systems for traffic information can provide a good foundation for the improvement of urban road network operating efficiency and service levels.This article is based on the analysis of the temporal and spatial characteristics of urban short-term traffic flow and its influencing factors,combined with deep learning theory to build a short-term traffic flow prediction model,to provide real-time and accurate traffic information support for urban traffic management and control.Taking urban roads as the research object,this paper first sorts out the relevant literature on short-term traffic flow prediction methods and objects,the extraction method of traffic flow characteristics in short-term prediction,and the relevant literature on the factors affecting the traffic flow characteristics.Secondly,it analyzes the spatiotemporal characteristics of traffic flow and its influencing factors to lay a theoretical foundation for model construction.Next,in order to more deeply capture the spatiotemporal characteristics of traffic flow,this paper proposes a short-term traffic flow prediction model based on combined deep learning(CDL model for short).This model combines the advantages of spatio-temporal convolutional blocks and long-term and short-term memory networks.It uses time-series graph-structured traffic flow data as model inputs to perform multi-step spatio-temporal predictions of short-term traffic flows in urban road networks.Then,considering that the spatiotemporal characteristics of traffic flow are disturbed by weather and other factors,based on the CDL model,this paper integrates the cycle factor,time factor,weather factor and road segment characteristic factor that affect the characteristics of traffic flow and propose a short-term traffic flow prediction model based on multi-factor fusion(abbreviated as MFCDL model).Finally,based on the taxi operation trajectory data of Xi’an High-tech Zone,a model effect verification experiment is designed and implemented,which proves the applicability and reliability of the CDL model and MFCDL model proposed in this paper in short-term traffic flow prediction scenarios.The experimental results show that the overall prediction performance of the CDL model is better than the traditional time series model and basic deep learning model,which verifies that the reasonable use of the combined model can fully exploit the spatiotemporal characteristics of traffic flow and effectively improve the prediction accuracy of the model.Secondly,the overall prediction performance of the MFCDL model is better than other prediction models that do not consider multiple factors,and it is verified that the effective use of influencing factors of traffic flow characteristics can improve the prediction accuracy of the model.Thirdly,through the sensitivity analysis of the four fusion factors of the MFCDL model,it can be seen that the cycle factor has a significant impact on improving the prediction accuracy of the model.As the prediction time step increases,the degree of influence of nearby components on the prediction accuracy of the model gradually decreases,and the degree of influence of weekly components gradually increases.At the same time,the degree of influence of time factors and weather factors on the prediction accuracy of the model gradually increases,and the degree of influence of the characteristic factors of road sections gradually decreases.This paper proposes a reasonable and feasible model method for the problem of short-term traffic flow prediction of urban roads,and explores and analyzes the impact of multi-factors on the prediction effect of the model.Providing an analytical idea for the interpretability research of deep learning methods applied to traffic prediction scenarios,and also verifies the applicability and practicability of deep learning in different short-term traffic flow prediction scenarios,and provides research for short-term traffic flow prediction Certain theoretical support and application practice.