Short-Term Traffic Flow Prediction Based On Multi-Characteristic Analysis

With the acceleration of social and economic development and the further progress of urbanization,the convenience of urban transportation has become a key factor restricting the development of urbanization.How to improve the efficiency of urban transportation has become an important subject in the urbanization process.Intelligent transportation system plays an important role in improving urban traffic efficiency.The key to building intelligent transportation system is to realize real time and accurate short-term traffic flow prediction.The key to realize short-term traffic flow prediction is to learn its complex spatial correlation,temporal correlation and randomness of traffic flow.In this paper,using Convolution Neural Network(CNN)to deal with spatial correlation between different regions,considering that the large urban area leads to a relatively deep network layer,Residual Network(ResNet)composed of Residual unit is added into the Convolutional Neural Network;Using three Gated Recurrent Unit(GRU)to deal with recent time dependence,daily period dependence and weekly period dependence,considering that each historical period data to forecast the influence degree of the time period is different,this paper take three Attention Mechanism into GRU;Using a two-layer full connection network to deal with the randomness of short-term flow combined with additional information such as weather data.After that,the prediction model is established by combining these three modules.What's more,in order to verify the influence of spatial correlation on prediction model,we proposed an urban functional area identification model to identify different functional regions.Finally,we based on the history of New York City taxi order data and weather data to validate our proposed model,the experimental results show that the prediction precision of the model is obviously superior to the mainstream of existing prediction methods.Besides,compared with predecessors put forward more advanced methods ST–ResNet,our model's RMSE has reduced 1.95 and MAE has reduced 1.3,which has higher application value.