Research On Analysis And Prediction Of Spatiotemporal Big Data

In recent years,with the development of Internet and storage technology,the scale of spatio-temporal data is increasing,forming spatio-temporal big data with the characteristics of big data.Among them,traffic big data is the most typical spatio-temporal big data.The intelligent transportation system plays an important role as a part of the smart city.It uses close cooperation between vehicles and the environment to alleviate traffic jams and improve the efficiency of the transportation network.Accurate traffic forecasting is essential for intelligent transportation systems in smart cities.The main research goal of this thesis is to make more accurate predictions of traffic flow and provide important decision indicators for downstream tasks.Due to the complex spatial correlation,short-term proximity correlation and long-term periodic correlation of traffic big data,traffic prediction is challenging.Although existing methods have considered these factors in modeling.Most solutions use convolutional or graph convolutional neural networks to model spatial correlation.However,the convolution operator may not be sufficient to model non-Euclidean pairwise correlation.This thesis proposes an end-to-end neural network model based on attention mechanism to predict traffic flow.First,a jump-connected recurrent neural network is used to capture the long-term periodic correlation,and the periodic feature of each road is learned.Secondly,a novel spatial attention mechanism is proposed.It can capture these dependencies more easily because every node attends to all other nodes in the network,which brings regularization effect to the model and avoids overfitting between nodes.Finally,a time attention mechanism is used to capture the time correlation between traffic flow at different moments in the immediate time and the current moment.Through experiments on multiple real-world traffic data sets,we analyze the effect of different modules on the model prediction accuracy,and compares the prediction performance of the proposed model with other benchmark models.Experimental results show that the prediction accuracy of the model in this thesis is superior to other comparative models,which proves that the model can model the spatial and temporal correlations of the traffic road network more effectively.