Research On Travel Time Prediction And Travel Time Reliability Based On Multi-Dimensional Data

ABSTRACT:With the continuous development of the transportation industry and the extension of the road network,the development level of the expressway and the public transportation demand show a double increasing trend,and the traffic information service is becoming more personalized,differentiated and refined.How to excavate and release the information of travel time,which is influenced by multi-dimensional factors such as vehicle types,meteorological scenes and period,becomes a new research field.In the fast-paced modern life,travelers pay more attention to time values and they are increasingly concerned about the delay and reliability of the travel time.Releasing predicted travel time and its reliability can provide support for the driver's path decision.Based on this,this paper used the toll collection data and meteorological monitoring data to take the Liaoning expressway as the experimental section carrying out the research of travel time and its reliability,which provides the decision basis for route decision making and relevant department management and operation.The major work and achievements are conclude as follows:(1)The influence of vehicle type,time and meteorological factors on the travel time was analyzed.Then the logical model of multi-dimensional data warehouse with the theme of travel time was designed,and the structure of data warehouse was constructed.Aiming at the integration problem of non-homologous data,a ratio-spatio matching method was proposed to realize the integration of toll collection data and meteorological monitoring data.The data cleaning,data conversion method were used to clear the abnormal data and unified the data format,improving the toll colletion data warehouse.(2)The problem of travel time sparse data and abnormal data processing was studies.The "upstream and downsteam data constructing method" is put forward creatively,which solves the sparse problem of charging data.Based on the related research,this paper put forward the data filtering method of "improved quartile method",which effectively eliminates the outliers in data set.After processing,the data information can be more complete and reflects the real situation.The process of travel time extraction was designed to prepare for the travel time prediction.OLAP technique was used to extract the multi-dimensional travel time information,and the influence of time,vehicle type and meteorological factors on the travel time is analyzed quantitatively.The rationality of fractal dimension was verified.(3)The autocorrelation and partial autocorrelation characteristics of the travel time series was analyzied,BIC criterion was used to determine the order of the model,the least squares method wass used to estimate the parameters,and the ARMA travel time prediction model can be establish.Then ARMAX prediction model was established by adding additional traffic flow series as the regression variables,which helps to improve the prediction effect of traditional ARMA model.The examples show that ARMAX's travel time prediction is effective and can meet the actual needs.Besides,prediction hysteresis problem of traditional ARMA model can be improves.The maximum percentage error is 5%lower than that of traditional ARMA model.(4)The distribution characteristics of historical travel time are studied.A variety of type of probability distribution were used to fitting the distribution,and K-S hypothesis test and fitting superiority are used to prove that logarithmic normal distribution is the optimal expression model of travel time reliability.Based on this model,the calculation method of the measure of the travel time reliability is established.The coefficient of variation,the buffer index,the planning time index and the crowding frequency are used to characterize the reliability of the historical travel time.The influence degree of the multi-dimensional factors such as vehicle type,time and meteorology on the reliability was studied by examples.This paper proposed two indexes including predicted travel time probability reliability and the buffer index of predicted travel time.The proposed indexes combine the future travel time with the historical statistical travel time to realize the evaluation of the reliability of the future travel time and supplement the travel time reliability index system.The examples show that the proposed predicted travel time reliability indexes are practical and important in route decision making.