Evaluation And Analysis Of Highway Traffic State Under Huge Data

With the increase of the traffic load and the traffic demand sustainability,the traffic system is faced with serious challenges.Traffic state evaluation is the basis of the traffic system behaviors and management decisions.As the expanded field of reliability technology,the traffic reliability is an important part for the system performance evaluation.Thus the traffic state consists of two parts: the level and the stability.Concise information is used to make decisions for managers and travelers.Due to the "Internet +" and the era of big data,the intelligent transportation is developing rapidly.As a result,there are still opportunities and challenges for traffic management and decision-making.In this paper,taking the Shaanxi Province Transportation Science and Technology Project "evaluation method and application of provincial highway operation service " as the basis,aiming at the actual demand of management and decision-making,researches on the theory and application of highway traffic assessment are included.The content,methodology and conclusion of this paper are all the improvement,complementary and useful exploration for the existing methods.(1)Data analysis of characteristics and qualityIn order to improve the data quality and the specific mining goals,data preprocessing is carried out before researches.Data quality control is the key and basic step for the reasonable accurate research conclusion.To keep important information,the asymmetric cleaning principle is put forward in highway travel time data,combined with the fourth quantile and statistical principle.Besides,the statistics are usually used to describe the data dispersion tendency and degree.To describe the data distribution features,the measure is established based on nonparametric test in kind of time,space,and vehicle.The significant differences between cars and buses are shown in travel time distribution.It is reasonable that vehicles are classified according to the traffic survey standard.(2)Highway travel time reliability estimation methodDensity estimation is considered as an important beginning on travel time reliability researches.The accuracy of traditional parameter methods depends on the distribution hypothesis in advance.There is not a unified conclusion for travel time distribution model in the literature of home and abroad.With the situation of uncertain probability distribution clusters,the model of highway travel time reliability base on wavelet density estimation is proposed in this paper.The model can be applied in roads with different grades or traffic conditions flexibility.Compared with two nonparametric models and five parameter models,the empirical results show that Wavelet Density Estimation(WDE)is much superior to other methods on highway travel time distribution estimation.The result shows that the proposed method is effective.In addition,the influence of ETC on highway travel time reliability is discussed.(3)Highway travel time short-term forecasting methodThe reliability of travel time prediction is the focus of managers and travelers.A large number of modeles are studied to improve the accuracy in travel time short-term forecast with two ways mainly,including model combination and the data fusion.Most researches are focused on point prediction without the confidence level and forecast range.Due to the lack of reliability in point prediction,the model of travel time interval prediction is constructed based on Bootstrap.K nearest neighbor and wavelet neural network models are improved and compared with the prediction error.As the less error,K nearest neighbor algorithm is introduced into the Bootstrap strategy.Interval prediction performances are compared on the different Bootstrap method with there indexes.Finally,the proposed model is validated by using expressway of Shaanxi in the case study.The results show that Percentile Bootstrap-KNN travel time interval prediction is more reliable,and better than KNN model.(4)Highway traffic state evaluation and analysisThe current standard of traffic state or service level is based on the density parameter.It is difficulty to collect the parameter of density directly.With the situation of blurred traffic state,the highway traffic state assessment method is puts forward for managers and travelers based on travel time statistical data,instead of traffic flow density.Considering the differences of parameter weight and sample size,the fuzzy cmeans clustering is improved under the application of traffic status discriminant.The results show that the proposed method are improved both in convergence speed and misjudgment rate,and can effectively describe the highway traffic state.Finally,the present situation of the Shaanxi province highway network status monitoring is summarized.And the example verification is carried out on the highway sections in Shaanxi province.The threshold of traffic reliability evaluation standard is calculated based on historical data.The decision-making process for traveler is established under scenarios.