Research On Short-term Traffic Flow Forecast Of Urban Road Network In Plum Rain Based On Deep Learning

Short-term traffic flow forecast is one of the popular research field of Intelligent Transportation Systems(ITS),and also a key basic technology for achieving efficient dynamic traffic guidance systems and active traffic control systems.Accurate short-term traffic flow prediction can provide a basis for managers to formulate effective traffic control strategies,and provide optimal travel route selection for users,to alleviate urban traffic congestion effectively and improve the efficiency of urban road network operation.Due to the low quality of traffic flow data collected by traditional traffic flow detectors,short-term traffic flow prediction methods have defects such as low prediction accuracy,weak real-time performance,and insufficient processing capacity for complex road networks.Especially in the complex natural environments such as plum rain,the prediction accuracy is lower and becomes an issue to solve in the practical process of short-term traffic flow prediction methods.This study provides a short-term traffic flow prediction method for the road network during the plum rain based on deep learning.Firstly,collect traffic trajectory data of the road network in the plum rain of the floating car dynamic traffic data with GPS.Secondly,doing data cleaning of traffic trajectory data by the four-step preprocessing algorithm.At the same time,this study provides an improved point-to-line map matching algorithm and matches vehicle and path valid traffic trajectory data with electronic maps by Google Geocoding API.Then,extract the rate of time-series target all sections of the road network and map speed parameter on the road network to maintain spatiotemporal characteristics of traffic flow.Finally,in order to deep mine the spatiotemporal characteristics of the road network traffic flow in the plum rain,this paper optimizes the traditional CNNs,and proposes a DCNNs model with multiple convolutional layers and pooling layers for the spatial characteristics of the traffic flow on the road network with plum rain.To merge the model with GRU neural network model sensitive to time-series,and propose DCNNs-GRU deep network model to realize short-term traffic flow prediction under the plum rain weather conditions.To verify the effectiveness of the model,this paper takes the road network of Fuzhou as a test area,by on the scientific research platform of Beidou Navigation and Smart Transportation Collaborative Innovation Center in Fujian Province to analyze speed parameters of massive traffic trajectory data in Fuzhou City and DCNNs-GRU deep network short-term traffic flow prediction.Compare the results with classic short-term traffic flow prediction model such as ARIMA,LSTM and GRU,the experiments show that the short-term traffic flow prediction method based on the DCNNs-GRU deep network is more feasible and effective to extract the spatio-temporal characteristics of the road network traffic flow during the plum rain,and prediction accuracy is also better than the classic short-term traffic flow prediction model.