Research On City Short-term Traffic Flow Forecast Algorithm Based On Joint Temporal And Spatial Features

Recently,with the rapid development of our country’s economy,the use of automobiles has gradually increased,and urban road resources have become increasingly scarce.Traffic congestion has become an important issue affecting social development.The prediction of urban traffic flow can effectively reduce problems such as traffic congestion,traffic accidents,and environmental pollution.Most of the existing research methods use a single model to analyze the traffic flow.However,there are problems that the analysis of the time and space characteristics of the traffic data is not sufficient,and the change of a single characteristic of the data will easily have an important impact on the prediction accuracy of the entire model.In order to fully explore the temporal and spatial characteristics of traffic flow,this thesis proposes corresponding traffic prediction algorithms for cities with regular and irregular layouts.By integrating different deep learning models into an end-to-end network,the temporal and spatial characteristics of data are realized.The main research contents are as follows:(1)This thesis proposes an urban short-term traffic flow prediction algorithm based on the CNN-Res Net-LSTM model for urban areas with regular layouts.First,divide the urban area into several grids according to the latitude and longitude;secondly,integrate the convolutional neural network,residual neural unit and long-short-term memory neural network into an end-to-end network framework to extract the spatiotemporal characteristics of the data respectively,that is,use Convolutional neural network extracts the spatial characteristics of traffic flow,and uses long and short-term memory neural network to extract the time characteristics of traffic flow;finally,combined with external data to predict the city’s short-term traffic flow.By comparing with the classic deep learning model,the number of model parameters,Root Mean Squared Error(RMSE),and Mean Absolute Percentage Error(MAPE)values are all reduced,which optimizes the model structure and improves the accuracy of prediction.(2)This thesis proposes an urban short-term traffic flow prediction algorithm based on the convolution mechanism of the attention spatio-temporal graph for the construction of irregular urban areas,which integrates the graph convolution network and the long and short-term memory neural network into an end-to-end network.First,use graph convolutional network and long short-term memory neural network to extract the spatial and temporal characteristics of traffic flow respectively;secondly,connect with spatial attention mechanism and temporal attention mechanism to capture the dynamic correlation of data;finally,use New York Experiments with city bicycle data,compared with classic forecasting models,the designed model improves the forecasting accuracy rate and reduces the average absolute percentage error and root mean square error of the model.In summary,this thesis designs and develops a prototype system for urban shortterm traffic flow prediction.Based on the detailed analysis of the system architecture and user needs,it realizes the functions of urban traffic data preprocessing,traffic flow prediction and visualization.This thesis has 45 pictures,11 tables,and 80 references.