Research On Prediction Model Based On Traffic Time Series Data

With the progress of the society,the level of economic growth,the factors of production increase liquidity,traffic as the carrier of production factors flow,is the important driving force of the development of the city,so to forecast the traffic time series,have important meaning,this article mainly research the traffic time series for the Traffic Index data and South Luogu Lane passenger flow data,using statistical,machine learning and other methods to analyze the characteristics.In addition,the characteristics of applied traffic time series are extracted by Deep Learning method to realize the prediction of traffic time series data.The main research contents are as follows:1.The application of Sequence to Sequence Learning(Seq2Seq)model in the field of Natural Language Processing(NLP)was innovated.Based on the research status of traffic index prediction at home and abroad,this paper analyzes and compares the traditional prediction methods and Deep Learning prediction methods,excavates the characteristics of traffic index;Analyzes the advantages and disadvantages of Seq2 Seq model,and uses this model to predict the traffic index;Uses ‘Random Search Method' to select the parameters of the model.The experimental results prove that Seq2 Seq model is effective for traffic index prediction.2.Proposed a Prediction Model based on Pattern-Features(PMPF).According to the change law of traffic index,it is found that there is correlation between the data;the Pattern-features of traffic index data are extracted through the Convolutional Neural Network(CNN)and input into the decoder layer of Seq2 Seq model as auxiliary information;In view of the problems in PMPF prediction,the Linear Regressive(LR)is used to fine tune the model.Experiments shows that LR+PMPF model has good performance.3.Proposed a tourist flow forecasting model based on Graph Convolutional Network(GCN)-RNN model.According to the historical passenger flow of South Luogu Lane and the passenger flow of the surrounding bus and subway stations,the passenger flow in the scenic spot is predicted;The geographical relationship between the scenic spot and the surrounding stations is a ‘graph 'structure,and the passenger flow of the stations not only affects the passenger flow in the scenic spot,but also affects each other,Therefore,GCN is used to extract the temporal characteristics of the passenger flow in the scenic spot and the stations;Takes the spatial features extracted by GCN as the input of RNN model,the temporal features of scent spot passenger and surrounding stations passenger flow are extracted.The applicability of GCN-RNN model is proved by experiments.