Research On Short-term Traffic Flow Forecast Technology Of Intersections Based On Neural Network

With the progress of the construction of integrated three-dimensional transportation network,the large-scale application of sensing technology in the field of transportation leads to the rapid growth of traffic data,which will become the guarantee of modern high-quality integrated three-dimensional transportation network.Short-term traffic flow prediction is one of the problems that must be solved in the modernization of comprehensive three-dimensional transportation network,which is of great significance to the realization of traffic guidance and path planning.It is also the technical guarantee for the network traffic to be more intelligent,convenient,smooth and reliable.In order to study the problem of short-term traffic flow prediction at intersections,this paper uses neural network theory to design a traffic flow prediction model based on multi-dimensional space-time characteristics.On the basis of modeling recent trend,daily cycle trend and weekly cycle trend,the model designed in this paper also combines the traffic flow characteristics of high correlation intersections in the surrounding area and characteristics of taxi activity in the surrounding area to capture and enrich the temporal and spatial characteristics of traffic flow at the same time.In reality,the bad environment of the intersection traffic scene makes the traffic data inevitably lack of data,which has a negative impact on the analysis of the spatio-temporal characteristics of traffic flow and the prediction performance of the model.Therefore,this paper designs a traffic data missing value filling model based on time convolution network,which repairs the traffic data fragment defect,and makes the traffic flow in the intersection scene more stable The analysis is more accurate.In this paper,the traffic flow prediction technology of intersection is mainly focused on the following aspects.(1)The traffic flow data is resampled at a time interval of 5 minutes.The noise points are removed from the traffic flow data.The traffic flow sequence is smoothed to remove the random factors.The standardized data converts the eigenvalues to 0 to 1,which is convenient for analysis,modeling and comparison,and provides reference for the follow-up research Support.(2)Based on the temporal and spatial characteristics of traffic flow at intersections,this paper proposes and constructs a short-term traffic flow prediction model based on temporal and spatial characteristics.Firstly,the similarity algorithm and correlation strength table are used to calculate the correlation strength of the intersection of the traffic network,and the correlation degree is analyzed.Combined with the analysis results,the traffic flow characteristics of the intersection with the highest correlation are selected and integrated into the neural network model for analysis.In addition,the surrounding taxi activity is also an important spatial feature,and its information is integrated into the neural network model as a feature.In order to fully mine the temporal and spatial characteristics,the model uses the attention based bilstm neural network as the core algorithm.Finally,the multi-dimensional space-time characteristics are used to fully mine the change law of traffic flow,and realize the accurate prediction of short-term traffic flow.(3)the phenomenon of missing data can not be avoided,which affects the deep analysis and mining of the essential characteristics of traffic data and the construction of short-term traffic flow prediction model.In order to avoid the impact of missing data of traffic flow data on the results of traffic analysis,this paper designs a traffic flow missing value filling model based on time convolution neural network,which integrates the past association information and future association information of missing segments of traffic flow,and realizes the repair of missing segments of traffic flow series.In the experiment,the short-term traffic flow prediction model and missing value filling model designed in this paper have good performance.The short-term traffic flow prediction model can accurately predict the traffic flow in the intersection scene.Filling model can repair the missing segments of traffic data with high quality,and the repaired data has higher value for traffic flow prediction.