Long-term Prediction Of Traffic Volume Based On Combined Model Of Markov Chains

Long-term traffic volume prediction has great significance in urban traffic network planning and transportation policymaking.However,traffic volume is affected by climate,economy,travel and other factors.The change of traffic volume is a non-stationary random process with volatility and non-strict periodicity.So this causes that long-term traffic volume is difficult to predict.After analyzing the characteristics of traffic volume in urban road,this thesis introduces Markov Chains theory to construct three kinds of combined models to predict traffic volume 24 hours in advance.The main research of this thesis is as follows.Firstly,this thesis summarizes three characteristics of traffic volume through analyzing the real traffic volume data.Among them,randomness is the premise of introducing Markov Chains theory to predict traffic volume.Volatility is a basis of interval prediction.And non-strict periodicity provides the possibility for the long-term traffic volume prediction.Then,this thesis introduces and uses the data compensation algorithm and filtering algorithm to deal with missing data and noise data in real traffic volume data.Secondly,this thesis constructs three kinds of combined forecasting models based on Markov Chains theory.The combined forecasting model based on Mean Markov Chains constructs traffic volume states by means.According to the state transition of traffic volume data,a traffic volume curve is gotten by predicting traffic volume 24 hours in advance.But considering the strong volatility of traffic volume,this thesis presents the idea of interval prediction.The clustering algorithm is used to reconstruct the state of traffic volume.So the combined forecasting model based on Clustering Markov Chains predicts a region enclosed by an upper curve and a lower curve.This region quite reflects a possible fluctuation range of traffic volume clearly.But this range is a bit wide.Therefore,this thesis introduces the weighting idea to reconstruct the traffic volume forecasting module.And the combined forecasting model based on Clustering Weighted Markov Chains predicts a region with the shorter mean interval length.Finally,this thesis uses the measured traffic data from Shenzhen to evaluate the prediction results of the above three models.The evaluation results show that the combined forecasting model based on Clustering Weighted Markov Chains has better prediction effect.At the same time,compared with other methods in the references,this method can not only guarantee the prediction accuracy of traffic volume,but also give a possible fluctuation range.