Short-term traffic flow prediction method based on global-local residual combination model

Abstract

The invention discloses a short-term traffic flow prediction method based on a global-local residual combination model, wherein the method comprises the following steps: collecting urban road network traffic flow data, transmitting the data to a traffic big data cluster in real time, carrying out data preprocessing of original data, and reducing the data redundancy; converting the preprocessed data into space-time raster data according to the latitude and longitude of the road network; performing standardization processing on the space-time raster data, and dividing the space-time raster data into a training set and a test set; constructing the global-local space-time residual combination model; and training the constructed global-local space-time residual combination model to predict the space-time raster data of the urban road network at the next moment. On the basis of time-space analysis of the road network, the global and local spatial features of the urban road network are analyzed at the same time, and the global and local spatial features and long-term features of the road network are captured more effectively, so that the short-term traffic flow prediction precision is improved.

Description

technical field [0001] The invention belongs to the field of intelligent traffic and traffic flow prediction, and specifically relates to a short-term traffic flow prediction method based on a global-local residual combined model. Background technique [0002] With the acceleration of my country's urbanization process, the construction of transportation infrastructure is gradually difficult to meet the growing number of cars, resulting in increasingly serious traffic congestion. As one of the important components of the intelligent transportation subsystem, short-term traffic flow forecasting can provide traffic management departments with traffic information in the future and help them formulate scientific and reasonable traffic guidance and traffic scheduling. Real-time and accurate acquisition of urban traffic flow can not only help travelers make road planning in advance, but also help traffic management departments formulate corresponding traffic management methods for ...

AI Technical Summary

Problems solved by technology

[0003] There are many models for traffic flow prediction. With the rise of deep learning, the deep neural network model with neural network as the core has become popular. Urban road network traffic flow has complex spatiotemporal characteristics, and only considering a certain feature will consider the model. The prediction accuracy is not high. There are many models that analyze the spatio-temporal characteristics of traffic flow, but they lack the capture of their long-term temporal characteristics and global spatial characteristics. As a result, such models cannot adapt to the traffic flow prediction from the macro perspective of urban road networks. global considerations

Images