| In the megalopolis, the layout of urban planning was already shaping up, and it was indicated that developing the traffic research and enlarging the Intelligent Transportation System (ITS) applying was one of the most effective method to solve the traffic crux instead of just building the roads now and then. The ITS’s construction basis and the system’s using condition was mostly depend on the traffic information extraction and applying.Rapid transit is a complex system impacted by many factors, and it had strong randomness. But for the concrete traffic status, it had some regularity and correlation. This dissertation researched the traffic statue time series characteristic and correlation analysis using the related theory and method in the statistics and systems engineering, and the data came from Beijing expressway detection system. The target was to find the changing rule of the expressway and to provide the basis in expressway control and management.Facing the mass dynamic traffic data, it had three core aspects in expressway traffic statue analysis, one was how to choice the traffic statue parameters, the second was how to choice the analyzing methods and the last one was how to apply the research results. Considering the core aspects, the main study contexts were described as follow:(1)Considering the expressway traffic statue grading demand, the classic volume-occupancy model was used to build the two stage-four phase traffic statue based on the elastic constant theory. The traffic statue was divided into congested flow and non-congested flow firstly, then the non-congested flow could be divided into smooth flow and coherent-moving flow and the congested flow divided into synchronized flow and jam, thus the traffic statue dividing had quantization basis. According to the traffic condition and the traffic statue dividing standard, the traffic statue fuzzy-inference decision method was deduced based on the similar speed and occupancy, it was proved this method was more suitable in true traffic condition.(2)The comprehensive and systematic time series analyzing method was studied for the traffic statue time series had statistic characteristics and pattern, additional, it had the tendency, periodicity, mutability, chaos characteristic and so on. Thus, the traffic statue changing rule could be recognized more clearly. Using the mutability and chaos analyzing results, the complex traffic statue time series could be decomposed into simple traffic statue time series, then the trend fitting and model composing were used. Comparing with the long-term fitting model, the MAE and MAPE of the fractal trend composing model was decreased18.34%and the RSE was decreased18.42%, also.(3)For the degree of traffic statue time series association analyzing, the statistic correlation, fuzzy correlation and grey correlation were used firstly, then the single-section and two-section correlation methods were deduced considering the retardance, the precision of forecast could be improve if the results were used. For the degree of space physics property association analyzing, the comprehensive index of path capacity and path mileage was proposed, this meant using the connectivity to build the space weight matrix, then the space autocorrelation theory was used to calculate the time series global space autocorrelation Morans’s I and local space autocorrelation Morans’s Ii, This made the traffic statue time series space attribution more clearly in different section’s.(4)For the details of the traffic statue correlation, considering the random events’ frequency, using the space weight matrix, the space Markov-chain was used to extract the traffic statue time series correlation pattern. Comparing with the traditional ARIMA forecast and Markov-chain forecast, the correlation model using space Markov-chain could reflect the traffic statue time series dependency on different section, and it had more precision in short-term traffic flow forecast for different traffic stage.
|
PDF Full Text Request