Font Size: a A A

Short-term Traffic Status Prediction Of Freeway Based On Big Data

Posted on:2020-09-14Degree:MasterType:Thesis
Country:ChinaCandidate:Z P ZhengFull Text:PDF
GTID:2392330623963677Subject:Electronic and communication engineering
Abstract/Summary:PDF Full Text Request
The continuous development of expressways has brought great convenience to people's travel.However,with the improvement of people's living standard,vehicles are becoming more and more popular,which causes the congestion of freeway.The congestion of freeway not only wastes lots of travelers' time and affects travel experience,but also can pollute the environment.In order to solve the congestion problem,effective induction and control of traffic is needed.One of the key technologies is to realize short-term prediction of traffic parameters.Through a large number of historical traffic parameters' data,this paper analyzes the characteristics of three traffic parameters of traffic engineering and concldues that the traffic parameters have the characteristics of periodicity and randomness.Considering the periodicity and randomness of traffic parameters,this paper establishes two models for traffic parameters' s prediction of one location node and multiple location nodes of freeway road.Firstly,this paper studies the traffic flow parameters' prediction of one location node.Traditional traffic parameters' prediction methods include parametric models,non-parametric models,combined models,hybrid models,etc,but these models only consider the traffic parameters' data a time series problem,ignoring the periodic characteristics of traffic parameters.Based on the characteristics of the traffic parameters,this paper proposes a mode decomposition based hybrid model for traffic parameters' prediction.The periodic sequence and the two random sequences are obtained through mode decomposition and mode combination of the original sequence.Then,according to the complexity of the three subsequences and the advantages of the existing prediction models,the models of BP,?-(1,LSTM are established to predict three sub-sequence respectively.The final result is obtained by combining the results of three sub-sequences.Comparing with singe prediciton models,the experiment results show the proposed model achieves higher accuracy.Based on one location node's prediction model,this paper also studies the traffic flow parameters' prediction of multiple location nodes.The prediction problem gets more difficult because of much more randomness that the multiple location nodes consists of the spatial and temporal factors.The existing prediction models are mainly considering the correlation of each location node and choose to reduce the data dimensions.However,these models have the problem of parameter selection,which needs to get a trade-off between computational complexity and prediction accuracy,and it is difficult to achieve good prediction results to meet actual needs.This paper proposes a convolutional neural network based on mode decomposition.Similarly,The periodic sequence and the random sequence are obtained through mode decomposition and mode combination.Next,considering the advantages convolutional neural network in processing multi-dimensional data,the two convolutional neural network are established to predict the two sub-sequences.The experiment results show the proposed model achieves higher accuracy.The one location node's prediction model and the multiple location nodes' prediction model constitute the microscopic and medium-scopic system of the traffic parameters' prediction.The model of one location node can provide information for travelers who will pass through a specific location and the model of multiple location nodes can reflect the changes status of road in the future.
Keywords/Search Tags:Intelligent Transportation System, Mode Decomposition, Hybrid Model, Convolutional Neural Network, Traffic Parameters Prediction
PDF Full Text Request
Related items